PWE3 Working Group Moran Roth (Ed.) Internet Draft Ronen Solomon Document: draft-ietf-pwe3-fc-encap-08.txt Corrigent Systems Expires: February 2009 Munefumi Tsurusawa KDDI August 18, 2008 Encapsulation Methods for Transport of Fibre Channel frames Over MPLS Networks Status of this Memo By submitting this Internet-Draft, each author represents that any applicable patent or other IPR claims of which he or she is aware have been or will be disclosed, and any of which he or she becomes aware will be disclosed, in accordance with Section 6 of BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. Abstract A Fibre Channel pseudowire (PW) is used to carry Fibre Channel frames over an MPLS network. This enables service providers to offer "emulated" Fibre Channel services over existing MPLS networks. This document specifies the encapsulation of Fibre Channel PDUs within a pseudowire. It also specifies the procedures for using a PW to provide a Fibre Channel service. Roth, et al. Expires - January 2009 [Page 1]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 Table of Contents 1. Specification of Requirements..................................3 2. Introduction...................................................3 2.1. Transparency..............................................4 2.2. Bandwidth Efficiency......................................4 2.3. Traffic Engineering.......................................5 3. Reference Model................................................5 4. Encapsulation..................................................7 4.1. The Control Word..........................................7 4.2. MTU Requirements..........................................8 4.3. Mapping of FC traffic to PW PDU...........................8 4.4. PW failure mapping.......................................10 5. Signaling of FC Pseudo Wires..................................10 5.1. Interface Parameters for FC PW...........................11 5.1.1. SR Poll Timeout (T1)...................................11 5.1.2. SR Response Timeout (T2)...............................11 5.1.3. SR Poll Retries (N2)...................................11 5.1.4. SR Window Size (k).....................................11 5.1.5. Fragmentation Indicator................................12 6. Congestion Control............................................12 6.1. Rate Control.............................................12 6.1.1. Protocol Mechanism.....................................13 6.1.2. Data Sender Protocol...................................13 6.1.3. Data Receiver Protocol.................................16 6.2. Selective Retransmission overview........................16 6.2.1. FC Encapsulation Header................................18 6.2.2. Encapsulation Header field parameters..................19 6.2.3. Selective reject (SR-SREJ) frame.......................20 6.2.4. Exception condition reporting and recovery.............22 6.3. Selective Retransmission procedures......................24 6.3.1. SR mode of operation...................................24 6.3.2. SR procedure for addressing............................24 6.3.3. SR procedure for the use of the Poll/Final bit.........24 6.3.4. Procedures for information transfer....................24 6.3.5. List of SR system parameters...........................32 7. Timing Consideration..........................................33 8. Security Considerations.......................................35 9. Applicability Statement.......................................35 10. IANA Considerations..........................................36 11. References...................................................37 12. Informative references.......................................38 13. Author's Addresses...........................................39 14. Contributing Author Information..............................39 Roth, et al. Expires - January 2009 [Page 2]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 1. Specification of Requirements The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [BCP14]. 2. Introduction As metro transport networks migrate towards a packet-oriented network infrastructure, the PSN is being extended in order to allow all services to be transported over a common network infrastructure. This has been accomplished for services such as Ethernet [RFC4448], Frame Relay [RFC4619], ATM [RFC4717] and SONET/SDH [RFC4842] services. Another such service, which has yet to be addressed, is the transport of Fibre Channel (FC) frames over the PSN. This will allow network service providers to transparently carry FC services over the packet- oriented network, along with the aforementioned data and TDM services. During recent years applications such as Storage Area Networks (SAN) extension and disaster recovery have become a prominent business opportunity for network service providers. In order to meet the intrinsic service requirements that characterize FC-based applications, such as transparency and low latency, various methods for encapsulating and transporting FC frames over backbone networks have been developed [FC-BB]. FC/IP, as described in [RFC3821] and [FC-BB], defines the mechanisms that allow the interconnection of islands of FC SANs over IP Networks. It provides a method for encapsulating FC frames employing FC Frame Encapsulation, as defined in [RFC3643], and addresses specific FC concerns related to tunneling FC over a pure IP network. Fibre Channel pseusowire (FC PW) is being proposed to provide a method for transporting FC frames over an MPLS network. It defines the encapsulation of FC Protocol Data Units (PDU) into an MPLS pseudowire, as well as procedures for using PW encapsulation to enable FC services such as SAN extension and disaster recovery over an MPLS PSN. FC PW complements the currently available standardized methods for transporting FC frames over a PSN. Specifically, FC/IP addresses "only the requirements necessary to properly utilize a pure IP network as a conduit for FC Frames", whereas FC PW addresses the requirements necessary to transport FC over an MPLS PSN. An example of such a network might be a packet-oriented multi-service transport Roth, et al. Expires - January 2009 [Page 3]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 network, where MPLS is used as the universal method for encapsulating and transporting all type of services, including mission critical FC applications as well as other TDM and data services. Hence, a key benefit of FC PW is that it will enable the extension of FC applications to the carrier space. The following sections describe some of the key carrier requirements for transporting FC frames over an MPLS PSN. 2.1. Transparency Transparent emulation of an FC link is a key requirement for transporting FC frames over a carrier's network. Conventionally, the coupling (or pairing) of FC entities with those pertaining to specific encapsulation methods requires the protocol-specific entity to terminate the FC Entity. This, in most cases, would require global address synchronization to be performed by the operator. In addressing this requirement, and providing full transparency, FC PW defines a port-mode FC encapsulation into a PW. This requires the creation of an FC pseudowire emulating an FC Link between two FC ports, appearing architecturally as being wired to those ports, similar to the approach defined for FC over GFPT in [FC-BB]. This results in transparent forwarding of FC frames over the MPLS PSN from both the FC Fabric and the operator's point of view. 2.2. Bandwidth Efficiency This is an important requirement for transporting FC over an MPLS PSN, where the protocol overhead has to be minimized in order to guarantee an end-to-end performance consistent with, e.g., SONET networks. FC PW defines a minimal overhead of 16 bytes, required due to the inclusion of the FC Encapsulation Header (4 bytes, refer to section 6.2.1), as well as the Control Word (4 bytes), PW label (4 bytes) and MPLS label (4 bytes). This can be contrasted with the overhead required by other methods such as those defined in [FC-BB]. Moreover, the ability to characterize services by specific bandwidth attributes, such as Committed Information Rate (CIR) and Excess Information Rate (EIR), effectively enables network operators to take full advantage of the statistical multiplexing capabilities of a packet-oriented network. This allows the multiplexing of best effort and premium services over the same media, effectively optimizing bandwidth utilization while still providing bandwidth guarantees and high service availability, as required by premium services such as FC PW. Roth, et al. Expires - January 2009 [Page 4]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 2.3. Traffic Engineering The transport of FC frames over a PSN network requires the operator not only to optimize the use of bandwidth resources, but also to define an explicit path over which availability and performance can be guaranteed. This capability is offered by other interconnect technologies such as ATM or SONET network technologies. FC PW defines the mapping of FC frames into a PW, implicitly assuming the use of MPLS-TE for the explicit provisioning of an FC PW over the MPLS PSN. This enables the operator to guarantee the performance and availability of the emulated FC link. FC requires a reliable transmission mechanism between FC entities. This implicitly assumes a lossless media with high availability. This, however, cannot always be guaranteed in best effort networks where FC frames are at times transported over sub-optimal paths. Bearing this in mind, FC PW relies on MPLS-TE to create an emulated FC link over a packet-oriented network, effectively enabling network operators to establish an explicit path to enhance frame transmission performance. 3. Reference Model FC PW allows FC Protocol Data Units (PDUs) to be carried over an MPLS network. In addressing the issues associated with carrying a FC PDU over an MPLS network, this document assumes that a pseudowire has been set up by some means outside of the scope of this document. This MAY be achieved via manual configuration, or using the signaling protocol as defined in [RFC4447]. FC PW emulates a single FC link between exactly two endpoints. This document specifies the emulated PW encapsulation for FC. Figure 1 describes the reference models which are derived from [RFC3985] to support the FC PW emulated services. For the purpose of the discussion in this document PE1 will be defined as the ingress router, and PE2 as the egress router. A layer 2 PDU will be received at PE1, encapsulated at PE1, transported, decapsulated at PE2, and transmitted out on the attachment circuit of PE2. Roth, et al. Expires - January 2009 [Page 5]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 |<-------------- Emulated Service ----------------->| | | | |<------- Pseudo Wire ------->| | | | | | | | |<-- MPLS Tunnel -->| | | | V V V V | V AC +----+ +----+ AC V +-----+ | | PE1|===================| PE2| | +-----+ | |----------|............PW1..............|----------| | | CE1 | | | | | | | | CE2 | | |----------|............PW2..............|----------| | +-----+ ^ | | |===================| | | ^ +-----+ ^ | +----+ +----+ | | ^ | | Provider Edge 1 Provider Edge 2 | | | | | | Customer | | Customer Edge 1 | | Edge 2 | | | | Native FC service Native FC service Figure 1: PWE3 FC Interface Reference Configuration The following reference model describes the termination point of each end of the PW within the PE: +-----------------------------------+ | PE | +---+ +-+ +-----+ +------+ +------+ +-+ | | |P| | | |PW ter| | MPLS | |P| | |<==|h|<=| NSP |<=|minati|<=|Tunnel|<=|h|<== From PSN | | |y| | | |on | | | |y| | C | +-+ +-----+ +------+ +------+ +-+ | E | | | | | +-+ +-----+ +------+ +------+ +-+ | | |P| | | |PW ter| | MPLS | |P| | |==>|h|=>| NSP |=>|minati|=>|Tunnel|=>|h|==> To PSN | | |y| | | |on | | | |y| +---+ +-+ +-----+ +------+ +------+ +-+ | | +-----------------------------------+ Figure 2: PW reference diagram The Native Service Processing (NSP) function includes native FC traffic processing that is required either for the proper operation Roth, et al. Expires - January 2009 [Page 6]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 of the FC link, or for the FC frames that are forwarded to the PW termination point. The NSP function is outside of the scope of PWE3 and is defined by [FC-BB]. 4. Encapsulation This specification provides port to port transport of FC encapsulated traffic. The following FC connections (as specified in [FC-BB]) are supported over the MPLS network: - N-Port to N-Port - N-Port to F-Port - E-Port to E-Port FC Primitive Signals and FC-Port Login handling by the NSP function within the PE is defined in [FC-BB]. 4.1. The Control Word The Generic PW Control Word, as defined in "PWE3 Control Word" [RFC4385] MUST be used for FC PW to facilitate the transport of short packets (by setting the Length field as detailed below), and convey the flag bit defined below. The structure of the Control Word is as follows: 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ |0 0 0 0| PT |A|0 0| Length | Sequence Number | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ Figure 3 - Control Word structure for the one-to-one mapping mode The first four bits of the PW Control Word MUST be set to 0 by the ingress PE to indicate PW data. The Flags bits are in use to convey the value of two flags, as specified below. PT - Payload Type indication. This field identifies the payload type carried within the PW PDU. The following types are defined: PT = 0: FC data frame. PT = 1: FC login frame. PT = 2: FC Primitive Sequence. PT = 6: FC Control Frame (refer to [FC-BB]). A - The Address bit identifies the frame as either a command or a Roth, et al. Expires - January 2009 [Page 7]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 response. This field is used in conjunction with the Poll Bit of the Selective Retransmission protocol. Messages containing commands MUST set this bit to 1. Messages containing responses MUST set this bit to 0. This bit MUST be set to 0 for FC Control frames as indicated by Payload Type value of 6. Further details regarding the use of this flag are provided in section 6. The FRG bits are not used for FC PW. These bits may be used in the future for FC specific indications as defined in [RFC4385]. The length field MUST be used for packets shorter than 64 bytes. Its processing must follow the rules defined in [RFC4385]. The sequence number is not used for FC PW and MUST be set to 0 by the ingress PE, and MUST be ignored by the egress PE. Refer to section 6 for the sequencing mechanism used for FC PW. 4.2. MTU Requirements The MPLS PSN MUST be able to transport the largest Fibre Channel encapsulation frame, including the overhead associated with the tunneling protocol. The maximum frame size without PW and MPLS labels (refer to Figure 4) is 2164 bytes. The MPLS PSN SHOULD accommodate frames of up to 2500 bytes to support future expansion of FC frames. Fragmentation, described in [RFC4623], SHALL NOT be used for FC PW, therefore the network MUST be configured with a minimum MTU that is sufficient to transport the largest encapsulation frame. 4.3. Mapping of FC traffic to PW PDU FC frames and Primitive Sequences are transported over the PW. All packet types are carried over a single PW. The FC header MUST contain a FC PW Control Word and a FC Encapsulation Header. The Encapsulation Header is described in section 6. Each FC frame is mapped to a PW PDU, including the Start Of Frame (SOF) delimiter, frame header, CRC field and the End Of Frame (EOF) delimiter, as shown in figure 4. SOF and EOF frame delimiters are encoded as specified in [FC-BB]. FC Primitive Sequences are encapsulated in a PW PDU containing the encoded K28.5 character [FC-BB], followed by the encoded 3 data characters, as shown in Figure 5. A PW PDU may contain one or more FC encoded ordered sets [FC-BB]. The length field in the FC PW Control Roth, et al. Expires - January 2009 [Page 8]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 Word is used to indicate the packet length when the PW PDU contains a small number of Primitive Sequences. 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +---------------------------------------------------------------+ | FC PW Control Word | +---------------------------------------------------------------+ | FC Encapsulation Header | +---------------+-----------------------------------------------+ | SOF Code | Reserved | +---------------+-----------------------------------------------+ | | +----- FC Frame ----+ | | +---------------------------------------------------------------+ | CRC | +---------------+-----------------------------------------------+ | EOF Code | Reserved | +---------------+-----------------------------------------------+ Figure 4 - FC frame encapsulation within PW PDU Idle Primitive Signals are carried over the PW in the same manner as Primitive Sequences. Note that in both cases a PE is not required to transport all the ordered sets received. The PE MAY implement repetitive signal suppression functionality as part of the NSP functionality. This is out of the scope of this document (refer to [FC-BB] for further details). 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +---------------------------------------------------------------+ | FC PW Control Word | +---------------------------------------------------------------+ | FC Encapsulation Header | +---------------+---------------+---------------+---------------+ | K28.5 | Dxx.y | Dxx.y | Dxx.y | +---------------+---------------+---------------+---------------+ | | +---- ----+ | | +---------------+---------------+---------------+---------------+ | K28.5 | Dxx.y | Dxx.y | Dxx.y | +---------------+---------------+---------------+---------------+ Figure 5 - FC Ordered Sets encapsulation within PW PDU Roth, et al. Expires - January 2009 [Page 9]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 The egress PE extracts the Primitive Sequence and Idle Primitive Signals from the received PW PDU. It continues transmitting the same ordered set until a FC frame or another ordered set is received over the PW. FC Control frames are transported over the PW, by encapsulating each frame in a PW PDU. The FC header MUST contain a FC PW Control Word, with PT = 6, and an all zeros FC Encapsulation Header (Selective Retransmission does not apply to FC Control frame transmission). FC Control Frame payload is out of scope of this document and is defined in [FC-BB]. 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +---------------------------------------------------------------+ | FC PW Control Word | +---------------------------------------------------------------+ | FC Encapsulation Header | +---------------------------------------------------------------+ | | +----- FC Control Frame ----+ | | +---------------------------------------------------------------+ Figure 6 - FC Control frame encapsulation within PW PDU 4.4. PW failure mapping PW failure mapping, which are detected through PW signaling failure, PW status notifications as defined in [RFC4447], or through PW OAM mechanisms MUST be mapped to emulated signal failure indications. The FC link failure indication is performed by the NSP, as defined by [FC-BB], and is out of the scope of this document. 5. Signaling of FC Pseudo Wires [PWE3-CONTROL] specifies the use of the MPLS Label Distribution Protocol, LDP, as a protocol for setting up and maintaining pseudo wires. This section describes the use of specific fields and error codes used to control FC PW. The PW Type field in the PWid FEC element and PW generalized ID FEC elements MUST be set to "FC Port Mode" as requested in section 8 below. Roth, et al. Expires - January 2009 [Page 10]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 The Control Word is REQUIRED for FC pseudo-wires. Therefore the C-Bit in the PWid FEC element and PW generalized ID FEC elements MUST be set. If the C-Bit is not set the pseudo-wire MUST not be established and a Label Release MUST be sent with an "Illegal C-Bit" status code [PWE3-CONTROL]. 5.1. Interface Parameters for FC PW 5.1.1. SR Poll Timeout (T1) The Selective Retransmission (SR) Poll Timeout (Parameter ID = TBA by IANA) is defined in section 6.3.5. The parameter length is 4 bytes. The parameter value indicates the poll timeout in units of 1 millisecond. The two PE on the edges of a FC PW MUST agree on the same value of this parameter for the PW to be set up successfully. 5.1.2. SR Response Timeout (T2) The Selective Retransmission Response Timeout (Parameter ID = TBA by IANA) is defined in section 6.3.5. The parameter length is 4 bytes. The parameter value indicates the response timeout in units of 1 microsecond. The restrictions specified in section 6.3.5 MUST be enforced for proper operation of the SR mechanism. The two PE on the edges of a FC PW MUST agree on the same value of this parameter for the PW to be set up successfully. 5.1.3. SR Poll Retries (N2) The Selective Retransmission Poll Retries (Parameter ID = TBA by IANA) is defined in section 6.3.5. The parameter length is 4 bytes. The parameter value is an integer indicating the number of poll retries. The two PE on the edges of a FC PW MUST agree on the same value of this parameter for the PW to be set up successfully. 5.1.4. SR Window Size (k) The Selective Retransmission Window Size (Parameter ID = TBA by IANA) is defined in section 6.3.5. The parameter length is 4 bytes. The parameter value is an integer indicating the maximum number of outstanding packets. Roth, et al. Expires - January 2009 [Page 11]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 The two PE on the edges of a FC PW MUST agree on the same value of this parameter for the PW to be set up successfully. 5.1.5. Fragmentation Indicator The Fragmentation Indicator (Parameter ID = 0x09) is specified in [RFC4446] and its usage is defined in [RFC4623]. Since fragmentation is not used in FC PW, the fragmentation indicator parameter MUST be omitted from the Interface Parameter Sub-TLV. 6. Congestion Control FC PW traffic can be transmitted over networks that may experience congestion due to statistical multiplexing. When congestion conditions are experienced frames may be discarded within the MPLS PSN. Congestion control mechanism is required to prevent congestion collapse and provide fairness among the different connections. Fairness is usually defined with respect to TCP flow control [RFC2914]. The FC PW relies on a congestion control mechanism that provides TCP-friendly behavior by controlling the transmission rate into the PSN by a rate shaper, whose output rate is a function of network congestion. Frame loss within the MPLS PSN also requires a reliable transmission mechanism in the PE to support faithful emulation of FC service, providing in-order, no-loss transport of FC traffic between CE1 and CE2. Reliable transmission is provided by a sliding-window selective retransmission (SR) mechanism to allow efficient retransmission of lost frames. This was standardized for FC transport in [FC-BB]. The SR mechanism also provides congestion indication (i.e. Frame loss events) to the rate control mechanism. 6.1. Rate Control The rate control mechanism provides adaptive shaper control in response to network congestion indications. The rate shaper is configured with BW attributes, such as CIR and EIR, assigned to the FC PW service. The rate control operation is based on [RFC3448]. In the following sections the applicability of [RFC3448] to FC PW is analyzed, and rate control operation is detailed. [RFC3448] is a receiver-based congestion control mechanism, where the congestion control information (i.e., the loss event rate) is calculated by the receiver. In FC PW, on the other hand, the congestion control information is calculated by the sender. This Roth, et al. Expires - January 2009 [Page 12]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 approach is more appropriate for the point-to-point nature of FC PW. This sender-based approach is also mentioned in [RFC3448] as a possible variant of the protocol. 6.1.1. Protocol Mechanism In accordance with [RFC3448] the actual allowed sending rate is directly computed by a throughput equation, as a function of lost frames and round trip time. In general, the congestion control mechanism works as follows: o The receiver detects lost frames and feeds this information back to the sender as part of the SR mechanism. o The sender calculates the frame loss probability and measures the round-trip time (RTT) as defined in [FC-BB]. o The lost frame probability and RTT are then fed into the throughput equation, calculating the acceptable transmission rate. o The sender then adjusts its transmission rate to match the calculated rate in accordance with the service BW attributes (CIR, EIR). As the CIR is guaranteed, the throughput equation controls only the excess transmission rate. The parameters of the throughput equation are set as follows: o The retransmission timeout (t_RTO) is replaced by the T1 timer of the SR mechanism as defined in section 6.3. o The number of frames acknowledged by a single SR acknowledgment frame (b) is set to b = 1, as recommended in [RFC3448]. Different implementation MAY use delayed acknowledgement by increasing the value of b. Frame loss probability (p) and RTT (R) are calculated as specified in Section 6.1.2. 6.1.2. Data Sender Protocol The data sender sends a stream of data frames to the data receiver at a controlled rate. When a feedback frame is received from the data receiver, the data sender calculates the frame loss probability and changes its sending rate accordingly. If the sender does not receive Roth, et al. Expires - January 2009 [Page 13]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 a feedback frame during a timeout period, it reduces its sending rate. This is achieved by the SR T1 timer. We specify the sender-side protocol in the following steps: o Sender initialization. o The sender behavior when a feedback frame is received. o The sender calculation of the frame loss probability. o The sender behavior when a feedback frame is not received for a timeout period. The sender rate shaper is initialized to transmit at the CIR. The SR mechanism is also initialized by resetting the sequence numbers. The sender calculates RTT (R), based on delay measurement frames transmitted by the NSP (as defined in [FC-BB]). These frames MUST be sent at least every 100 milliseconds, and are used to measure round trip samples that are averaged to obtain RTT (refer to [RFC3448] section 4.3 for details). If an RTT measurement is missed (either due to a loss of a delay measurement frame, or to an RTT larger than the measurement period), PE1 SHOULD shut the PW down, as specified in [FC-BB]. The sender calculates the frame loss probability based on feedback frames generated by the receiver. A feedback frame with accordance to the SR mechanism defined in [FC-BB] is one of the following: o Receiver Ready (RR) - a frame that includes the N(R) counter to acknowledge the sender frames up to frame N(R). o Receiver Not Ready (RNR) - a frame that includes the N(R) counter to acknowledge the sender frames up to frame N(R), and pause the sender from sending additional frames. o Selective Reject (SREJ) - a frame that includes lost frames indication (sequence numbers). When the sender receives a feedback frame it re-calculates the frame loss probability. RR and RNR will effectively decrease the frame loss probability due to no frame loss. On the other hand, reception of a SREJ frame tends to increase the frame loss probability. An implementation MAY consider sending feedback frames, in a controlled network environment, with expedite forwarding (EF) CoS to assure delivery. Roth, et al. Expires - January 2009 [Page 14]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 After the frame loss probability is updated, the sender calculates a new transmission rate for the rate shaper. The transmission rate is calculated as: Rate = MIN( PIR, MAX(CIR,X) ), where CIR is the Committed Information Rate, PIR is the Peak Information Rate (PIR = CIR+EIR), X is the outcome of the throughput equation as specified in [RFC3448], and MIN/MAX are functions returning the smallest/largest value among their operands, respectively. Note that the transmission rate as controlled by the above function, is bounded in the range [CIR,PIR]. No feedback in accordance with [RFC3448] is defined by the timer T1. When the sender does not receive a feedback for such an interval it halves its transmission rate as defined in [RFC3448]. The transmission rate equation as specified above MUST still be applied to guarantee that the CIR is the lower limit for the throughput. The procedure controlling timer T1 (refer to Section 6.3) guarantees that transmission rate is not halved during idle periods, as the timer is not activated during these periods. The maximum burst size allowed MUST be limited to a round-trip time's worth of packets, to achieve efficient transmission while conforming to [RFC3448]. In case the transmission rate is equal to CIR for a period greater than RTT, and transmitted frames are still lost in transit, as indicated to the sender by receiving SREJ frames, the sender MUST signal PW status of "Unable to maintain minimum transmission rate" (refer to Section 9 - "IANA Considerations" for details) as specified in [RFC4447], and MUST stop transmission over the PW for a duration of 10 seconds (this period allows a transient network problem to resolve itself, and guarantees that no more than one HELLO message [FC-SW] is lost, and the link between the two FC devices is not affected). The sender and receiver MUST discard all frames residing in the buffers associated with the congested PW. The sender MUST also discard all frames received from the attached FC device. If within 10 seconds after transmission was restarted severe congestion conditions are encountered, as described above (i.e., CIR cannot be maintained), the sender MUST shut the PW down, as described in Section 6.5 of [RFC3985]. If the PW has been set up using the PWE3 control protocol [RFC4447], the regular PW teardown procedures SHOULD be used. The PW MUST NOT be automatically restarted, and administrative intervention is required. Upon PW shutdown the sender and receiver MUST discard all frames associated with the PW. Note that congestion may be avoided by employing connection admission control (CAC) mechanism, which assures that congestion conditions will not be reached when a PW is transmitting at its configured CIR. Roth, et al. Expires - January 2009 [Page 15]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 6.1.3. Data Receiver Protocol The data receiver receives a stream of data frames from the data sender, generates SR feedback frames (SR-RR, SR-RNR and SR-SREJ), and sends them to the data sender. The details of feedback frames generation and transmission are specified in section 6.3. 6.2. Selective Retransmission overview The selective retransmission mechanism provides efficient retransmission of lost frames to enable faithful emulation of FC service, with no frame loss experienced by the CE. The proposed selective retransmission mechanism was standardized for FC transport in [FC-BB], and is specified in details in this standard. The SR protocol is an efficient sliding window full-duplex protocol that supports both the flow control and error recovery functions. SR has been adopted from ITU's Link Access Protocol B (LAPB) that was derived from ISO/IEC's High-level Data Link Control (HDLC) balanced classes. Use of LAPB in SR is limited to a subset of the synchronous modulo 32768 super sequence numbering service option. SR works between two PE devices (see figure 7). SR flow control works by streaming multiple messages within an allowed window, bounded by the system parameter k, and awaits acknowledgements before sending more messages. Acknowledgements indicate which messages were correctly received and there is a provision for requesting retransmission of selected messages in the current window. Fibre Channel Sequences and Exchanges are not visible to the SR flow control protocol which sees the PW packets constructed from the FC frames. Some benefits of the SR protocol are summarized below: a) it is used for reliable transport of all Class 2, 3, 4, and F frames between two PE devices; b) it optimizes buffer management at the PE devices; c) it acts as a congestion avoidance technique to match the capacity of the sender to the capacity of the network that carries the payload; d) it ensures correct delivery of messages (i.e., an error control and recovery function); and Roth, et al. Expires - January 2009 [Page 16]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 e) it provides a continuous stream of traffic across the MPLS PSN thus leading to a higher throughput (i.e., optimizes bandwidth utilization at each BBW device). Note that the synchronization of the Sender PE and the Receiver PE at the PW message level, which is required for correct SR operation is performed through PW signaling. The four different SR messages described in section 6.2.1 have a correspondence to the LAPB frame types. Note that only the information transfer SR-I message is flow-controlled while all other messages are control messages of the protocol. The SR protocol specifies the maximum number (k) of outstanding messages at any given time. k is a system parameter that is not negotiated and is fixed in a given implementation. The value of this system parameter depends on the MPLS PSN delay characteristics and the number of buffers available. Typically, the value of k is expected to be far below the maximum number of 32767. +--------------------+ +--------------------+ | PE | | PE | | | | | | +--------------+ | | +--------------+ | | | Flow Control |<---------------->| Flow Control | | | | Protocol | | | | Protocol | | | +--------------+ | | +--------------+ | | | | | | | | | | | | | | +--------------+ | | +--------------+ | | | PW Interface | | | | PW Interface | | | +--------------+ | | +--------------+ | | | | | +--------------------+ +--------------------+ | - -- | | -- / \ / \ | | / \/ -- \ | | \ / | | \ MPLS \ | +---------/ PSN /-----------+ / / ------------- Figure 7 - SR flow control protocol between two PEs Roth, et al. Expires - January 2009 [Page 17]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 6.2.1. FC Encapsulation Header The FC Encapsulation Header defines two types of field formats that are used to perform information transfer (i.e., I-format frames), and supervisory functions (i.e., S-format frames). SR makes use of four different types of messages: a) I-format (1): SR-I frame. This frame is used to perform an information transfer. The Encapsulation Header of an I-format frame is shown in figure 8. The I-frame Encapsulation Header contains the following fields: (1) N(S): Transmitter send sequence number. (2) N(R): Transmitter receive sequence number. (3) P: Poll bit (1 = Poll). A detailed description of the different fields and explanation of The functionality involved is provided in section 6.2.2. An SR-I frame is a command message (i.e., the A-bit in the Control Word is set to 1), and carries an encapsulated FC frame. b) S-format (3): SR-RR, SR-RNR, SR-SREJ frames. These frames are used to perform supervisory control functions of the Selective Retransmission mechanism, such as acknowledge SR-I messages, request retransmission of SR-I messages, and to request a temporary suspension of transmission of SR-I messages. The Encapsulation Header of an S-format frame is shown in figure 9. The S-frame Encapsulation Header contains the following fields: (1) N(R): Transmitter receive sequence number. (2) S: Supervisory function bits to define the frame type. S = 00: SR-RR. S = 01: Reserved. S = 10: SR-RNR. S = 11: SR-SREJ. (3) P: Poll/Final bit (refer to section 6.2.2 for detailed description). (4) Reserved: MUST be set to 0 by the ingress PE, and MUST be ignored by the egress PE. A detailed description of the different fields and explanation of Roth, et al. Expires - January 2009 [Page 18]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 The functionality involved is provided in section 6.2.2. An SR-RR frame carries no payload, and may be either a command or response message (the A-bit in the Control Word is set to 1 for a Command, and to 0 for a Response). It indicates Ready to Receive SR-I messages (negates busy condition) and acknowledges previous SR-I messages. An SR-RNR frame carries no payload, and may be either a command or response message. It indicates Receiver not Ready to accept more SR-I messages (busy condition) and acknowledges previous SR-I messages. An SR-SREJ frame may be either a command or response message, and carries a payload that indicate SR-I frames in need of selective retransmission. 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-----------------------------+-+-----------------------------+ |0| N(S) |P| N(R) | +-+-----------------------------+-+-----------------------------+ Figure 8 - FC Encapsulation Header format for I-frame 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+---+-----------------------+-+-----------------------------+ |1|0| S | Reserved |P| N(R) | +-+-+---+-----------------------+-+-----------------------------+ Figure 9 - FC Encapsulation Header format for S-frame 6.2.2. Encapsulation Header field parameters The following describes the different fields of the Encapsulation Header and details how these fields are handled. a) Modulus of SR - Each SR-I message is sequentially numbered and may have the value 0 through modulus minus 1, where "modulus" is equal to 32768 (i.e., the maximum value of the sequence numbers). The sequence numbers cycle through the entire range. b) Send state variable V(S) - The send state variable V(S) denotes the sequence number of the next-in-sequence SR-I message to be transmitted. V(S) may take on the values 0 through modulus minus 1. The value of V(S) is incremented by 1 with each successive SR-I Roth, et al. Expires - January 2009 [Page 19]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 message transmission, but cannot exceed the N(R) of the last received SR-I or supervisory message by more than the maximum number of outstanding SR-I messages k. The value of k is defined in section 6.3.5. c) Send sequence number N(S) - Only SR-I messages contain N(S), the send sequence number of the transmitted SR-I message. At the time that an in-sequence SR_I message is designated for transmission, the value of N(S) is set to the value of the send state variable V(S). d) Receive state variable V(R) - The receive state variable V(R) denotes the sequence number of the next-in-sequence SR-I message expected to be received. V(R) may take on the values 0 through modulus minus 1. The value of V(R) is incremented by 1 by the receipt of an error-free, in-sequence SR-I message whose send sequence number N(S) equals the receive state variable V(R). e) Rceive sequence number N(R) - All SR-I messages and supervisory messages, except SR-SREJ messages with the F bit set to 0, SHALL contain N(R), the expected send sequence number of the next received SR-I message. At the time that a message of the above types is designated for transmission, the value of N(R) is set to the current value of the receive state variable V(R). N(R) indicates that the PE transmitting the N(R) has correctly received all SR_I messages numbered up to and including N(R)-1. f) Functions of the Poll/Final bit (P-bit) - All messages contain P- bit, the Poll/Final bit. In command messages, the P-bit is referred to as the Poll bit. In response messages it is referred to as the Final bit. The Poll bit set to 1 is used by the PE to solicit (i.e., poll) a response from the remote PE. The Final bit set to 1 is used by the PE to indicate the response message transmitted by the remote PE, as a result of the soliciting (i.e., poll) command. The use of the P/F bit is further described in section 6.3.3. 6.2.3. Selective reject (SR-SREJ) frame The SR-SREJ supervisory message is used by a PE to request retransmission of one or more, not necessarily contiguous, SR-I messages. The N(R) field SHALL contain the sequence number of the earliest SR-I message to be retransmitted and the information field (see figure 9) SHALL contain, in ascending order (i.e., 32767 is Roth, et al. Expires - January 2009 [Page 20]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 higher than 32766 and 0 is higher than 32767 for modulo 32768), the sequence numbers of additional SR-I message(s), if any, that needs to be retransmitted. The payload field SHALL be encoded such that there is a 2-byte field for each standalone SR-I message in need of retransmission, and a 4- byte span list for each sequence of two or more contiguously numbered SR-I messages in need of retransmission, as depicted in figure 9. Standalone SR-I messages are identified in the payload field by the appropriate N(R) value preceded by a 0 bit in the 2-byte field used. Span lists are identified in the payload field by the N(R) value of the first SR-I message in the span list preceded by a 1 bit in the 2- byte field used, followed by the N(R) value of the last message in the span list preceded by a 1 bit in the 2-byte field used. The maximum payload size of a SR-SREJ message is 2148 bytes corresponding to a maximum possible encoding of 1074 standalone SR-I messages or a maximum possible encoding of 537 span list sets. If the P-bit in an SR-SREJ message is set to 1, then SR-I messages numbered up to N(R)-1 inclusive, N(R) being the value in the Encapsulation Header field, shall be considered as acknowledged. If the P-bit in an SR-SREJ message is set to 0, then the N(R) in the Encapsulation Header field of the SR-SREJ message does not indicate acknowledgement of SR-I messages. 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +---------------------------------------------------------------+ | FC PW Control Word | +---------------------------------------------------------------+ | FC Encapsulation Header | +-+-----------------------------+-+-----------------------------+ |0| N(R) of standalone SR-I |1| N(R) of first SR-I in span | +-+-----------------------------+-+-----------------------------+ |1| N(R) of last SR-I in span |0| N(R) of standalone SR-I | +-+-----------------------------+-+-----------------------------+ |1| N(R) of first SR-I in span |1| N(R) of last SR-I in span | +-+-----------------------------+-+-----------------------------+ | | . . . . +---------------------------------------------------------------+ Figure 10 - SR-SREJ frame format example Roth, et al. Expires - January 2009 [Page 21]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 6.2.4. Exception condition reporting and recovery The error recovery procedures that are available to effect recovery following the detection/occurrence of an exception condition are described in this section. Exception conditions described are those situations that may occur as the result of transmission errors, PE device malfunction, or operational situations. a) Busy condition - The busy condition results when the PE is temporarily unable to continue to receive SR-I messages due to internal constraints (e.g., receive buffering limitations). Upon entering the busy condition, a PE transmits an SR-RNR message. SR-I messages pending transmission may be transmitted from the busy PE prior to or following the SR-RNR message. An indication that the busy condition has cleared is communicated by the transmission of SR-RR or SR-SREJ. b) N(S) sequence error condition - The information field of all received SR-I messages whose N(S) is not in the range V(R) and V(R)+k-1 inclusive, SHALL be discarded. The information field of all SR-I messages received by the PE whose N(S) is in the range V(R) and V(R) + k -1 inclusive, SHALL be saved in the receive buffer. An N(S) sequence error exception condition occurs in the receiver when a received SR-I message contains an N(S) that is not equal to the receive state variable V(R) at the receiver. The receiver SHALL not acknowledge (i.e., increment its receive state variable) the SR-I message causing the sequence error, or any SR-I message that may follow, until an SR-I message with the correct N(S) is received. A PE that receives one or more valid SR-I messages having sequence errors or subsequent supervisory messages (i.e., SR-RR, SR-RNR, or SR-SREJ) shall accept and handle the N(R) field and the P-bit. The means specified below shall be available for initiating the retransmission of lost or errored SR-I messages following the occurrence of an N(S) sequence error condition. (1) SR-SREJ recovery - The SR-SREJ message shall be used to initiate more efficient error recovery by selectively requesting the retransmission of one or more, not necessarily contiguous, lost or errored SR-I message(s) following the detection of sequence errors, rather than requesting the retransmission of all SR-I messages. Roth, et al. Expires - January 2009 [Page 22]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 When a PE receives an out-of-sequence message, the SR-I message shall be saved in a receive buffer. The SR-I message shall be delivered to the upper layer only when all SR-I messages numbered below N(S) are correctly received. If message number N(S)-1 has not been received previously, then an SR-SREJ response message with the P-bit set to 0 shall be transmitted containing the sequence numbers of the block of consecutive missing SR-I messages ending at N(S)-1. On receiving such an SR-SREJ message the PE shall retransmit all requested SR-I messages. After retransmitting these SR-I messages, the BBW may transmit new SR_I messages, if they become available. When a PE receives a command message with the P-bit set to 1, if there are out-of-sequence SR-I messages saved in the receive buffer, it shall transmit an SR-SREJ message, with the F bit set to 1, containing a complete list of missing sequence numbers. The PE that receives the SR-SREJ message shall retransmit all requested SR-I messages, except those that were transmitted subsequent to the last command message with the P bit set to 1. (2) Time-out recovery - If a PE, due to a transmission error, does not receive, or receives and discards, a single SR-I message or the last SR-I message in a sequence of SR-I messages, it shall not detect a N(S) sequence error condition and, therefore, shall not transmit an SR-SREJ message. The PE that transmitted the unacknowledged SR-I message(s) shall, following the completion of a system specified time-out period (see section 6.3.4 items b) and j) below), send a supervisory command message (i.e., SR-RR or SR-RNR) with the P-bit set to 1. SR-I messages shall be retransmitted on the receipt of an SR-RR response message with the F bit set to 1 or an SR-SREJ message. c) Invalid message condition - Any message that is invalid shall be discarded, and no action is taken as the result of that message. An invalid message is defined as one that contains: (1) the Control Word with an invalid encoding; or (2) the Encapsulation Header with an invalid encoding. Roth, et al. Expires - January 2009 [Page 23]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 6.3. Selective Retransmission procedures 6.3.1. SR mode of operation The SR protocol shall be limited to a subset of the synchronous modulo 32768 super sequence numbering service option operation of the LAPB protocol. The SR protocol is initialized upon PW set-up following a successful signaling session. 6.3.2. SR procedure for addressing The Address Bit field in the Control Word (see figure 3) identifies a message as either a command or a response. This field is used in conjunction with the P-bit (Poll/Final). 6.3.3. SR procedure for the use of the Poll/Final bit The PE receiving a supervisory command (i.e., SR-RR, SR-RNR, SR- SREJ), or SR-I message with the P bit set to 1 shall set the F bit to 1 in the next response message it transmits. The response message returned by the PE to an SR-I message with the P bit set to 1, shall be an SR-RR, SR-SREJ, or SR-RNR response with the F bit set to 1. The response message returned by the PE to a supervisory command with the P bit set to 1, shall be an SR-RR, SR-RNR, or SR-SREJ response with the F bit set to 1. The P bit may be used by the PE in conjunction with the timer recovery condition (see section 6.3.4. item j) below). 6.3.4. Procedures for information transfer a) Procedures for SR-I messages The procedures that apply to the transmission of SR-I messages in each direction using multi-selective reject are described below. b) Sending new SR-I messages When the PE has a new SR-I message to transmit (i.e., an SR-I message not already transmitted), it shall transmit it with a N(S) equal to its current send state variable V(S), and a N(R) equal to its current receive state variable V(R). At the end of the transmission of the SR-I message, it shall increment its send state variable V(S) by 1. Roth, et al. Expires - January 2009 [Page 24]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 If the SR timer T1 is not running at the time of transmission of the SR-I message, it shall be started. If the SR send state variable V(S) is equal to the last value N(R) received plus k, where k is the maximum number of outstanding SR-I frames (see section 6.3.5), the PE shall not transmit any new SR-I frames. If the remote PE is busy, the PE shall not transmit any new SR-I messages. When the PE is in the busy condition, it may still transmit SR-I messages, provided that the remote PE is not busy. c) Receiving an in-sequence SR-I message When the PE is not in a busy condition and receives a valid SR-I message whose send sequence number N(S) is equal to its receive state variable V(R), the PE shall accept the information field of this message and increment by one the receive state variable V(R). If the SR-I message, whose N(S) is equal to the incremented value of V(R), is present in the receive buffer, then the PE shall remove it from the receive buffer, deliver it to the upper layer and increment V(R) by one. The PE shall repeat this procedure until V(R) reaches a value such that the SR-I message whose N(S) is equal to V(R) is not present in the receive buffer. The PE shall then take one of the following actions: (1) if the PE is now in the busy condition, it shall transmit an SR-RNR message with N(R) equal to the value of the SR receive variable V(R) (see item i) below); or (2) if the PE is still not in a busy condition: - if the P bit is set to 1, then the PE shall transmit a response message with the F bit set to 1, as specified in item l) below; - if an SR-I message is available for transmission the PE shall act as described in item b) above, sending new SR-I messages and acknowledging the received SR-I message by setting N(R) in the Encapsulation Header field of the next transmitted SR-I message to the value of the SR receive state variable V(R), or the PE shall acknowledge the received SR-I message by transmitting an SR-RR message with the N(R) equal to the value of the SR receive state variable V(R); or Roth, et al. Expires - January 2009 [Page 25]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 - the PE shall transmit an SR-RR message with N(R) equal to the value of the SR receive state variable V(R). When the PE is in a busy condition, it may ignore the information field contained in any received SR-I message. d) Reception of invalid messages When the PE receives an invalid message (see 6.2.4. item c), it shall discard the message. e) Reception of out-of-sequence SR-I messages When the PE is not in a busy condition and it receives a valid SR-I message whose send sequence number N(S) is out-of-sequence, (i.e., not equal to the receive state variable V(R)), then it shall perform one of the following actions: 1) if N(S) is less than V(R) or greater than or equal to V(R) + k, then it shall discard the information field of the SR-I message. If the P bit of the SR-I message is set to 1, then the PE shall transmit a response message with the F bit set to 1, as specified in item l) below; or 2) if N(S) is greater than V(R) and less than V(R) + k, then it shall save the SR-I message in the receive buffer. It shall then perform one of the following actions: - if the P bit of the SR-I message is set to 1, then the PE shall transmit a response message with the F bit set to 1, as specified in item l) below; - if the PE is now in a busy condition, it shall transmit an SR-RNR message with N(R) equal to the value of the receive variable V(R), as specified in item i) below; or - if the SR-I message numbered N(S)-1 has not yet been received, then the PE shall transmit an SR-SREJ response message with the F bit set to 0. The PE shall create a list of contiguous sequence numbers N(X), N(X)+1, N(X)+2,..., N(S)-1, where N(X) is greater than or equal to V(R) and none of the SR-I messages N(X) to N(S)-1 have been received. The N(R) field of the SR-SREJ message shall be set to N(X) and the information field set to the list N(X)+1,...,N(S)-1. If the list of sequence numbers is too large to fit into the information field of the SR-SREJ Roth, et al. Expires - January 2009 [Page 26]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 message, then the list shall be truncated to fit in one SR-SREJ message, by including only the earliest sequence numbers. When the PE is in the busy condition, it may ignore the information field contained in any received SR-I message. f) Receiving acknowledgement When correctly receiving an SR-I message or a supervisory message (i.e., SR-RR, SR-RNR, or SR-SREJ with the F bit set to 1), even in the busy condition, the PE shall consider the N(R) contained in this message as an acknowledgement for all the SR-I messages it has transmitted with a N(S) up to and including the received N(R)-1. The PE shall stop the timer T1 if the received supervisory message has the F bit set to 1 or if there is no outstanding poll condition and the N(R) is higher than the last received N(R), actually acknowledging some SR-I messages. If timer T1 has been stopped by the receipt of an SR-I message, an SR-RR command message, an SR-RR response message with the F bit set to 0, or an SR-RNR message, and if there are outstanding SR-I messages still unacknowledged, the PE shall restart timer T1. If timer T1 has been stopped by the receipt of an SR-SREJ message with the F bit set to 1, the PE shall follow the retransmission procedure specified in item g.2) below. If timer T1 has been stopped by the receipt of an SR-RR message with the F bit set to 1, the PE shall follow the retransmission procedure specified in item k) below. g) Receiving an SR-SREJ response message 1) Receiving an SR-SREJ response message with the F bit set to 0 When receiving an SR-SREJ response message with the F bit set to 0, the PE shall retransmit all SR-I messages, whose sequence numbers are indicated in the N(R) field and the information field of the SR-SREJ message, in the order specified in the SR-SREJ message. Retransmission shall conform to the following: - if the PE is transmitting a supervisory or SR-I message when it receives the SR-SREJ message, it shall complete that transmission before commencing transmission of the requested SR-I messages; or - if the PE is not transmitting any message when it receives the SR-SREJ message, it shall commence transmission of the requested SR-I messages immediately. Roth, et al. Expires - January 2009 [Page 27]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 If there is no outstanding poll condition, then a poll shall be sent, either by transmitting an SR-RR command, or SR-RNR command if the PE is in the busy condition, with the P bit set to 1 or by setting the P bit in the last retransmitted SR-I message and timer T1 shall be restarted. If there is an outstanding poll condition, then timer T1 shall not be restarted. 2) Receiving an SR-SREJ response message with the F bit set to 1 When receiving an SR-SREJ response message with the F bit set to 1, the PE shall retransmit all SR-I messages, whose sequence numbers are indicated in the N(R) field and the information field of the SR-SREJ message, in the order specified in the SR-SREJ message, except those messages that were sent after the message with the P bit set to 1 was sent. Retransmission shall conform to the following: - if the PE is transmitting a supervisory message or SR-I message when it receives the SR-SREJ message, it shall complete that transmission before commencing transmission of the requested SR-I messages; or - if the PE is not transmitting any message when it receives the SR-SREJ message, it shall commence transmission of the requested SR-I messages immediately. If any messages are retransmitted, then a poll shall be sent, either by transmitting an SR-RR command, or SR-RNR command if the PE is in the busy condition, with the P bit set to 1 or by setting the P bit in the last retransmitted SR-I message. Timer T1 shall be restarted. h) Receiving an SR-RNR message After receiving an SR-RNR message, the PE shall stop transmission of SR-I messages until an SR-RR or SR-SREJ message is received. The PE shall start timer T1, if necessary, as specified in section 6.3.5. When timer T1 runs out before receipt of a busy clearance indication, the PE shall transmit a supervisory message (i.e., SR-RR, SR-RNR), with the P bit set to 1 and shall restart timer T1, in order to Roth, et al. Expires - January 2009 [Page 28]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 determine if there is any change in the receive status of the remote PE. The remote PE shall respond to the P bit set to 1 with a supervisory response message (i.e., SR-RR, SR-RNR, SR-SREJ) with the F bit set to 1 indicating continuation of the busy condition (i.e., SR-RNR message) or clearance of the busy condition (i.e., SR-RR, SR- SREJ). Upon receipt of the remote PE response, timer T1 shall be stopped. The PE shall process the supervisory response message as follows: 1) if the response is an SR-RR message, the busy condition shall be assumed to be cleared and the PE may retransmit messages as specified in item k) below. New SR-I messages may be transmitted as specified in item b) above; 2) if the response is an SR-SREJ message, the busy condition shall be assumed to be cleared and the PE may retransmit messages as specified in item g.2) above. New SR-I messages may be transmitted as specified in item b) above; or 3) if the response is an SR-RNR message, the busy condition shall be assumed to still exist and the PE, after a period of time (e.g., the duration of timer T1), shall repeat the enquiry of the remote PE receive status. If timer T1 runs out before a status response is received, the enquiry process above shall be repeated. If N2 attempts to get a status response fail, the PE MAY declare the PW as down. If, at any time during the enquiry process, an unsolicited SR-RR or SR-SREJ message is received from the remote PE, it shall be considered to be an indication of clearance of the busy condition. Should the unsolicited SR-RR message be a command message with the P bit set to 1, the appropriate response message with the F bit set to 1 shall be transmitted (see item l) below) before the PE may resume transmission of SR-I messages. The PE shall not clear the outstanding poll condition. The PE shall not stop timer T1. If an unsolicited SR- SREJ message is received, then the PE shall perform retransmissions as specified in item g.1) above. i) BBW busy condition When the PE enters a busy condition, it shall transmit an SR-RNR message at the earliest opportunity. The SR-RNR message shall be a command frame with the P bit set to 1 if an acknowledged transfer of the busy condition indication is required, otherwise the SR-RNR message may be a command or response message. While in the busy condition, the PE shall accept and process supervisory messages, Roth, et al. Expires - January 2009 [Page 29]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 accept and process the N(R) field of SR-I, SR-RR, and SR-SREJ messages with the F bit set to 1, and return an SR-RNR response with the F bit set to 1 if it receives a supervisory command or SR-I command message with the P bit set to 1. Received SR-I messages may be discarded or saved as specified in items c) and e) above, however, SR-RR or SR-SREJ messages shall not be transmitted. To clear the busy condition, the PE shall transmit an SR-RR message, with the N(R) field set to the current receive state variable V(R). The SR-RR message shall be a command message with the P bit set to 1 if an acknowledged transfer of the busy-to-non-busy transition is required, otherwise the SR-RR message may be either a command or response message. j) Awaiting acknowledgement If the timer T1 runs out while waiting for the acknowledgement of an SR-I message from the remote PE, the PE shall restart timer T1 and transmit an appropriate supervisory command message (i.e., SR-RR, SR- RNR) with the P bit set to 1. The PE may transmit new SR-I messages after sending this enquiry message. If the PE receives an SR-SREJ response message with the F bit set to 1, the PE shall restart timer T1 and retransmit SR-I messages as specified in item g.2) above. If the PE receives an SR-SREJ response message with the F bit set to 0, the PE shall retransmit SR-I messages as specified in item g.2) above. If the PE receives an SR-RR response message with the F bit set to 1, the PE shall restart timer T1 and retransmit SR-I messages as specified in item k) below. If the PE receives an SR-RR response message with the F bit set to 0, or an SR-RR command message or SR-I message with the P bit set to 0 or 1, the PE shall not restart timer T1, but shall use the received N(R) as an indication of acknowledgement of transmitted SR-I messages up to and including SR-I message numbered N(R)-1. If timer T1 runs out before a supervisory response message with the F bit set to 1 is received, the PE shall retransmit an appropriate supervisory command message (i.e., SR-RR, SR-RNR) with the P bit set to 1. After N2 such attempts, the PE MAY declare the PW as down. k) Receiving an SR-RR response messages with the F bit set to 1 Roth, et al. Expires - January 2009 [Page 30]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 When receiving an SR-RR response message with the F bit set to 1, the PE shall process the N(R) field as specified in item f) above. If there are outstanding SR-I messages that are unacknowledged and no new SR-I messages have been transmitted subsequent to the last message with the P bit set to 1, then the PE shall retransmit all outstanding SR-I messages except those that were sent after the message with the P bit set to 1 was sent. Retransmission shall conform to the following: 1) if the PE is transmitting a supervisory or SR-I message when it receives the SR-RR message, it shall complete that transmission before commencing transmission of the requested SR-I messages; 2) if the PE is not transmitting any message when it receives the SR-RR message, it shall commence transmission of the requested SR-I messages immediately. If any messages are retransmitted, then a poll shall be sent, either by transmitting an SR-RR command, or SR-RNR command if the PE is in the busy condition, with the P bit set to 1 or by setting the P bit in the last retransmitted SR-I message. The timer T1 shall be stopped. If any SR-I messages are outstanding, then timer T1 shall be started. l) Responding to command messages with the P bit set to 1 When receiving an SR-RR, SR-RNR, or-SR_I command message with the P bit set to 1, the PE shall generate an appropriate response message as follows: 1) if the PE is in the busy condition, it shall transmit an SR-RNR response message with the F bit set to 1; 2) if there are some out-of-sequence messages in the receive buffer, then it shall transmit an SR-SREJ message with the F bit set to 1; N(R) shall be set to the receive state variable V(R) and the information field set to the sequence numbers of all missing SR-I messages, except V(R). If the list of sequence numbers is too large to fit in the information field of the SR-SREJ message, then the list shall be truncated by including only the earliest sequence numbers; or 3) if there are no out-of-sequence messages in the receive buffer, then an SR-RR response message with the F bit set to 1 shall be sent. Roth, et al. Expires - January 2009 [Page 31]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 6.3.5. List of SR system parameters a) SR Poll Timeout (Timer T1) The same value of the timer T1 shall be made known and agreed to by the two PEs. The period of timer T1, at the end of which retransmission of a message may be initiated (see 6.3.4), shall take into account whether T1 is started at the beginning or the end of the transmission of a message. The proper operation of the procedure requires that the transmitter's timer T1 be greater than the maximum time between transmission of a message (i.e., SR_I, or supervisory command) and the reception of the corresponding message returned as an answer to that message (i.e., acknowledging message). Therefore, the receiver should not delay the response or acknowledging message returned to one of the above messages by more than a value T2, where T2 is a system parameter (see item b) below). The PE shall not delay the response or acknowledging message returned to one of the above remote PE messages by more than a period T2. b) SR Response Timeout (Timer T2) The same value of the parameter T2 shall be made known and agreed to by the two PEs. The period of parameter T2 shall indicate the amount of time available at the PE before the acknowledging message shall be initiated in order to ensure its receipt by the remote PE, prior to timer T1 running out at the PEs (parameter T2 < timer T1). The period of parameter T2 shall take into account the following timing factors: 1) the transmission time of the acknowledging message; 2) the propagation time over the access link; 3) the stated processing times at the PEs; and 4) the time to complete the transmission of the message(s) in the PE transmit queue that are neither displaceable nor modifiable in an orderly manner. Given a value for timer T1 for the PEs, the value of parameter T2 shall be no larger than T1 minus 2 times the propagation time over Roth, et al. Expires - January 2009 [Page 32]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 the access data link, minus the message processing time at the PE, minus the message processing time at the remote PE, and minus the transmission time of the acknowledging message by the PE. c) SR Poll Retries (N2) The same value of the N2 system parameter shall be made known and agreed to by the two PEs. The value of N2 shall indicate the maximum number of attempts made by the PE to complete the successful transmission of a message to the remote PE. d) SR Window Size (k) The same value of the k system parameter shall be made known and agreed to by the two PEs. The value of k shall indicate the maximum number of sequentially numbered SR-I messages that the PEs may have outstanding (i.e., unacknowledged) at any given time. The value of k shall never exceed 32767 for modulo 32768 operation. 7. Timing Consideration Correct Fibre Channel information exchange requires that the inherent latency between CE1 and CE2 (refer to Figure 1) be: a) no more than one-half of the R_T_TOV (Receiver Transmitter Timeout Value, default value: 100 milliseconds, defined in [FC-FS]) of the attached devices for Primitive Sequences; b) no more than one-half of the E_D_TOV (Error Detect Timeout Value, default value: 2 seconds, defined in [FC-FS]) of the attached devices for frames; and c) within the R_A_TOV (Resource Allocation Timeout Value, default value: 10 seconds, defined in [FC-FS]) of the attached fabric(s), if any. Requirement a) above, controlling the latency associated with FC Primitive Sequence transport is addressed by [FC-BB], stating that in case Primitive Sequences are received from the CE or the remote PE, while the device is unable to forward these Primitive Sequence due to backpressure indication, it shall flush its respective buffer (PSN- facing if the Primitive Sequences were received from the CE, CE- Roth, et al. Expires - January 2009 [Page 33]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 facing if they were received from the remote PE), and shall forward the Primitive Sequences. Another case is when there is no specific backpressure indication, rather Primitive Sequences are being delayed due to retransmission of dropped frames. In this case also, the PE shall flush its PSN-facing buffer and shall forward the Primitive Sequences. Requirements b) and c) above apply to the latency associated with transporting FC frames, and the system MUST comply with the lower of the two timeouts. The mechanism controlling this latency is described below for the PE1 --> PE2 direction. A duplicate mechanism MUST be used to control the PE2 --> PE1 direction. PE2 (the receiver) maintain a timer Td, set to the minimum timeout set by requirements b) and c), with a safety margin to allow a deviation of the estimated RTT from the actual RTT. PE2 SHOULD set Td = 0.8 x (0.5 x MIN(E_D_TOV,R_A_TOV) - 2 x RTT), where RTT is an average value calculated as specified in Section 6.1.2, and the factor 0.8 provides safety margins for RTT fluctuations. In case the calculated Td < 2 x RTT for two consecutive calculation periods, the FC PW MUST be shut down (this avoids getting into conditions where Td expiration is too frequent). To guarantee correct operation of this mechanism FC PW SHOULD NOT be used in environments where the RTT may have high variability, i.e., environments where the estimated RTT may not be off by more than 5% of MIN(E_D_TOV,R_A_TOV). If the FC PW is used in an environment where this limit is exceeded, the safety margins MUST be increased to encompass twice the expected maximum variability in the RTT. Upon Td expiration PE2 declares WAN Down as defined in [FC-BB], send the Not Operational (NOS) Primitive Sequence to CE2, and flush its buffers. The timer Td is started when either of the following two conditions occurs: a) SR-RNR is sent by PE2 toward PE1, i.e., PE2 indicates backpressure, and stop PE1 from transmitting; b) SR-SREJ is sent by PE2 toward PE1, requiring retransmission. PE2 notes the sequence number (Nx) of the first frame to be received following the timer initiation (for SR_RNR this will be the next frame to be transmitted by PE1, i.e., the sequence number of the last frame received by PE2 before starting the timer plus 1. For SR_SREJ Roth, et al. Expires - January 2009 [Page 34]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 this frame will be the first sequence number in the retransmission list). The timer continues counting and is initialized in one of two cases: a) PE2 receives the frame with sequence number (Nx + k), where k is the Selective Retransmission window size; b) PE2 identifies idle period (all sent frames were acknowledged, and the CE-facing buffer is empty). 8. Security Considerations FC PW does not enhance or detract from the security properties of the underlying MPLS PSN, rather it relies upon the PSN's mechanisms for encryption, integrity, and authentication whenever required. The level of security provided may be less than that of a native FC service. FC PW shares susceptibility to a number of pseudowire-layer attacks and implementations SHOULD use whatever mechanisms for confidentiality, integrity, and authentication are developed for general PWs. These methods are beyond the scope of this document. The protocols used to implement security in a Fibre Channel fabric are defined in [FC-SP]. These protocols work at higher layers of the FC hierarchy and are transparent to the FC PW. 9. Applicability Statement FC PW allows the transport of point-to-point Fibre Channel links while saving network bandwidth. - The pair of CE devices operates as if they were directly connected by an FC link. In particular they react to Primitive Sequences on their local ACs in the standard way. - The FC PW carries only FC data frames and a single copy of a Primitive Sequence. Idle Primitive Signals encountered between FC data frames, and long streams of the same Primitive Sequence are suppressed over the PW thus saving bandwidth. FC PW traffic can traverse controlled (i.e., providing committed information rate for the service) networks and uncontrolled (i.e., providing excess information rate for the service) networks. In case of FC PW traversing an uncontrolled network, it MUST provide TCP- Roth, et al. Expires - January 2009 [Page 35]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 friendly behavior under network congestion (refer to Congestion Control section for further details). Faithfulness of a FC PW may be increased if the carrying MPLS PSN is Diffserv-enabled and implements a per-domain behavior (PDB, defined in [RFC3086]) that guarantees low loss, low re-ordering events and low delay. The NSP may include mechanisms to reduce the effect of these events on the FC service. These mechanisms are out of the scope of this document. This document does not provide any mechanisms for protecting FC PW against PSN outages. As a consequence, resilience of the emulated service to such outages is defined by the PSN behavior. However, the NSP MAY implement a mechanism to convey the PW status to the CE, to enable faster handling of the PSN outage. Moreover, the NSP MAY implement egress buffer and packet reordering mechanism to increase the emulated service resiliency to fast PSN rerouting events. As a function of the NSP this is out of the scope of this document. 10. IANA Considerations IANA is requested to assign a new PW type as follows: PW type Description Reference -------- -------------- ---------- 0x001F FC Port Mode [FC-encap] The above value is suggested as the next available value and the reference [FC-encap] refers to this document. IANA is requested to add the following entries to the Pseudowire Interface Parameters Sub-TLV type Registry: Parameter ID Length Description Reference --------- --------- ------------------------ ---------- 0x12 4 SR Poll Timeout (T1) [FC-encap] 0x13 4 SR Response Timeout (T2) [FC-encap] 0x14 4 SR Poll Retries (N2) [FC-encap] 0x15 4 SR Window Size (k) [FC-encap] The parameters are defined in sections 5.1.1 through 5.1.4. The reference [FC-encap] refers to this document. IANA is requested to assign an LDP (Label Distribution Protocol) Status Code from the portion of the LDP Status Code Name Space that is assigned by IETF consensus as follows: Roth, et al. Expires - January 2009 [Page 36]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 Range/Value E Description Reference ------------ ----- ---------------------- ---------- 0x0000003B 0 Unable to maintain [FC-encap] minimum transmission rate The above value is suggested as the next available value and the reference [FC-encap] refers to this document. 11. References [RFC3985] Bryant, S., et al, "Pseudo Wire Emulation Edge-to-Edge (PWE3) Architecture", RFC 3985, March 2005. [RFC3916] Xiao, X., et al, "Requirements for Pseudo Wire Emulation Edge-to-Edge (PWE3)", RFC 3916, September 2004. [RFC3086] Nichols, K., et al, "Definition of Differentiated Services Per Domain Behaviors and Rules for their Specification)", RFC 3086, April 2001. [RFC3448] Floyd, S., et al, "TCP Friendly Rate Control (TFRC): Protocol Specification", draft-ietf-dccp-rfc3448bis-06, April 2008. [RFC4446] Martini, L., "IANA Allocations for Pseudowire Edge to Edge Emulation (PWE3)", RFC 4447, April 2006. [RFC4447] Martini, L., et al, "Pseudowire Setup and Maintenance using the Label Distribution Protocol (LDP)", RFC 4447, April 2006. [RFC4385] Bryant, S., et al, "Pseudowire Emulation Edge-to-Edge (PWE3) Control Word for use over an MPLS PSN", RFC 4385, February 2006. [RFC4623] Malis, A., Townsley, M., "PWE3 Fragmentation and Reassembly", RFC 4623, August 2006. [FC-BB] "Fibre Channel Backbone-4" (FC-BB-4), ANSI INCITS 419:2008, to appear. RFC Editor: Please contact authors to obtain the correct date for the "to appear" in the above reference prior to publication. Roth, et al. Expires - January 2009 [Page 37]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 [FC-SW] "Fibre Channel - Switch Fabric - 4" (FC-SW-4), ANSI INCITS 418:2006, April 2006. [FC-FS] "Fibre Channel - Framing and Signaling - 2" (FC-FS-2), ANSI INCITS 424:2007, February 2007. [BCP14] Bradner, S., "Key words for use in RFCs to Indicate requirement Levels", BCP 14, RFC 2119, March 1997. [FC-SP] "Fibre Channel - Security Protocols" (FC-SP), ANSI INCITS 426:2007, February 2007. 12. Informative references [RFC3668] Bradner, S., "Intellectual Property Rights in IETF Technology", RFC 3668, February 2004. [RFC3821] M. Rajogopal, E. Rodriguez, "Fibre Channel over TCP/IP (FCIP)", RFC 3821, July 2004. [RFC3643] R. Weber, et al, "Fibre Channel (FC) Frame Encapsulation", RFC 3643, December 2003. [RFC2914] Floyd, S., "Congestion Control Principles", RFC 2914, September 2000. [RFC2581] Allman, M., et al, "TCP Congestion Control", RFC 2581, April 1999. [RFC4448] Martini, L., et al, "Encapsulation Methods for Transport of Ethernet over MPLS Networks", RFC 4448, April 2006. [RFC4842] Malis, A., et al, "SONET/SDH Circuit Emulation Over Packet (CEP)", RFC 4842, April 2007. [RFC4619] Martini, L., et al, "Encapsulation Methods for Transport of Frame Relay over MPLS Networks", RFC 4619, September 2006. [RFC4717] Martini, L., et al, "Encapsulation Methods for Transport of ATM over MPLS Networks", RFC 4717, December 2006. Roth, et al. Expires - January 2009 [Page 38]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 13. Author's Addresses Moran Roth Corrigent Systems 101, Metro Drive San Jose, CA 95110 Phone: +1-408-392-9292 Email: moranr@corrigent.com Ronen Solomon Corrigent Systems 126, Yigal Alon st. Tel Aviv, ISRAEL Phone: +972-3-6945316 Email: ronens@corrigent.com Munefumi Tsurusawa KDDI R&D Laboratories Inc. Ohara 2-1-15, Fujimino-shi, Saitama, Japan Phone: +81-49-278-7828 14. Contributing Author Information David Zelig Corrigent Systems 126, Yigal Alon st. Tel Aviv, ISRAEL Phone: +972-3-6945273 Email: davidz@corrigent.com Leon Bruckman Corrigent Systems 126, Yigal Alon st. Tel Aviv, ISRAEL Phone: +972-3-6945694 Email: leonb@corrigent.com Luis Aguirre-Torres Corrigent Systems 101 Metro Drive Ste 680 San Jose, CA 95110 Roth, et al. Expires - January 2009 [Page 39]
INTERNET DRAFT draft-ietf-pwe3-fc-encap-08.txt August 2008 Phone: +1-408-392-9292 Email: Luis@corrigent.com Intellectual Property Statement The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79. Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf-ipr@ietf.org. Disclaimer of Validity This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Copyright Statement Copyright (C) The IETF Trust (2008). This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights. Roth, et al. Expires - January 2009 [Page 40]
