BMWG G. Fioccola
Internet-Draft E. Vasilenko
Intended status: Informational P. Volpato
Expires: 25 April 2024 Huawei Technologies
L. Contreras
Telefonica
B. Decraene
Orange
23 October 2023
Benchmarking Methodology for IPv6 Segment Routing
draft-vfv-bmwg-srv6-bench-meth-08
Abstract
This document defines a methodology for benchmarking Segment Routing
(SR) performance for Segment Routing over IPv6 (SRv6). It builds
upon RFC 2544, RFC 5180, RFC 5695 and RFC 8402.
Status of This Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at https://datatracker.ietf.org/drafts/current/.
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."
This Internet-Draft will expire on 25 April 2024.
Copyright Notice
Copyright (c) 2023 IETF Trust and the persons identified as the
document authors. All rights reserved.
Fioccola, et al. Expires 25 April 2024 [Page 1]
Internet-Draft BM for SRv6 October 2023
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents (https://trustee.ietf.org/
license-info) in effect on the date of publication of this document.
Please review these documents carefully, as they describe your rights
and restrictions with respect to this document. Code Components
extracted from this document must include Revised BSD License text as
described in Section 4.e of the Trust Legal Provisions and are
provided without warranty as described in the Revised BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1. Requirements Language . . . . . . . . . . . . . . . . . . 4
2. SRv6 Forwarding . . . . . . . . . . . . . . . . . . . . . . . 4
3. Test Methodology . . . . . . . . . . . . . . . . . . . . . . 6
3.1. Test Setup . . . . . . . . . . . . . . . . . . . . . . . 6
3.2. Locator and Endpoint behaviors methods . . . . . . . . . 7
3.3. Frame Formats and Sizes . . . . . . . . . . . . . . . . . 8
3.4. Protocol Addresses . . . . . . . . . . . . . . . . . . . 10
3.5. Trial Duration . . . . . . . . . . . . . . . . . . . . . 10
3.6. Traffic Verification . . . . . . . . . . . . . . . . . . 10
3.7. Buffer tests . . . . . . . . . . . . . . . . . . . . . . 11
4. Reporting Format . . . . . . . . . . . . . . . . . . . . . . 12
5. SRv6 Forwarding Benchmarking Tests . . . . . . . . . . . . . 13
5.1. Throughput . . . . . . . . . . . . . . . . . . . . . . . 14
5.1.1. Throughput of a Source Node . . . . . . . . . . . . . 14
5.1.2. Throughput of a transit Segment Endpoint Node . . . . 15
5.1.3. Throughput of a Segment Endpoint Node with
decapsulation . . . . . . . . . . . . . . . . . . . . 15
5.1.4. Throughput of a Transit Node . . . . . . . . . . . . 16
5.2. Buffers size . . . . . . . . . . . . . . . . . . . . . . 16
5.3. Latency . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.4. Frame Loss . . . . . . . . . . . . . . . . . . . . . . . 17
5.5. System Recovery . . . . . . . . . . . . . . . . . . . . . 17
5.6. Reset . . . . . . . . . . . . . . . . . . . . . . . . . . 18
6. Security Considerations . . . . . . . . . . . . . . . . . . . 18
7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 19
8. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 19
9. References . . . . . . . . . . . . . . . . . . . . . . . . . 19
9.1. Normative References . . . . . . . . . . . . . . . . . . 19
9.2. Informative References . . . . . . . . . . . . . . . . . 20
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 22
Fioccola, et al. Expires 25 April 2024 [Page 2]
Internet-Draft BM for SRv6 October 2023
1. Introduction
Segment Routing (SR), defined in [RFC8402], leverages the source
routing paradigm. The headend node steers a packet through an SR
Policy [RFC9256], instantiated as an ordered list of segments. A
segment, referred to by its Segment Identifier (SID), can have a
semantic local to an SR node or global within an SR domain. SR
supports per-class explicit routing while maintaining per-class state
only at the ingress nodes to the SR domain.
However, there is no standard method defined to compare and contrast
the foundational SR packet forwarding capabilities of network
devices. This document aims to extend the efforts of [RFC1242],
[RFC2544], [RFC5180] and [RFC5695] to SR network.
The SR architecture can be instantiated on two data-plane: SR over
MPLS (SR-MPLS) and SR over IPv6 (SRv6). This document is limited to
SRv6.
This document is limited to Headend encapsulations (H.Encaps.xxx) and
segment Endpoints (End, End.X). It is expected that future documents
may cover the benchmarking of SRv6 applications with decapsulation
(End.Dxxx), Binding (End.Bxxx), Fast ReRoute
[I-D.ietf-rtgwg-segment-routing-ti-lfa], Compressed SID
[I-D.ietf-spring-srv6-srh-compression], etc.
SR can be applied to the IPv6 architecture with a new type of routing
header called the SR Header (SRH) [RFC8754]. An instruction is
associated with a segment and encoded as an IPv6 address. An SRv6
segment is also called an SRv6 SID. An SR Policy is instantiated as
an ordered list of SRv6 SIDs in the routing header. The active
segment is indicated by the Destination Address (DA) of the packet.
SRv6 involves 3 types of forwarding plane operations (PUSH/ NEXT/
CONTINUE) as further described in Section 2. SR list for PUSH
operation is typically constructed by the source node with a SR
Policy, see [RFC9256]. NEXT operation is done by a SR segment
endpoint node. CONTINUE operation happens for a SR transit node.
[RFC5695] describes a methodology specific to the benchmarking of
MPLS forwarding devices, by considering the most common MPLS packet
forwarding scenarios and corresponding performance measurements.
[RFC5180] provides benchmarking methodology recommendations that
address IPv6-specific aspects, such as evaluating the forwarding
performance of traffic containing extension headers.
Fioccola, et al. Expires 25 April 2024 [Page 3]
Internet-Draft BM for SRv6 October 2023
The purpose of this document is to describe a methodology specific to
the benchmarking of Segment Routing over IPv6. The methodology
described is a complement for [RFC5180] and [RFC5695].
1.1. Requirements Language
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 [RFC2119],
RFC 8174 [RFC8174].
2. SRv6 Forwarding
SR leverages the source routing paradigm. In SRv6, a SID is
allocated as an IPv6 address. For the IPv6 data plane, a new type of
IPv6 Routing Extension Header, called Segment Routing Header (SRH)
has been defined [RFC8754]. The SRH contains the Segment List as an
ordered list of IPv6 addresses: each address in the list is a SID.
Hence SRv6 Segment list typically contains more SIDs. A dedicated
field, referred to as Segments Left, is used to maintain the pointer
to the active SID of the Segment List.
There are three different categories of nodes that may be involved in
segment routing networks.
The SR source node is the headend node and steers a packet into an SR
Policy. It can be a host originating an IPv6 packet or an SR domain
ingress router encapsulating a received packet into an outer IPv6
packet and inserts the SRH in the outer IPv6 header. It sets the
first SID of the SR Policy as the IPv6 Destination Address of the
packet.
The SR transit node forwards packets destined for a remote segment as
a normal IPv6 packet based on the IPv6 destination address, because
the IPv6 destination address does not locally match with a segment.
According to [RFC8200] the only node allowed to inspect the Routing
Extension Header (and therefore the SRH) is the node corresponding to
the destination address of the packet.
The SR segment endpoint node receives packets whose IPv6 destination
address is locally configured as a segment. It creates Forwarding
Information Base (FIB) entries for its local SIDs. For each SR
packet, it inspects the SRH, may prepare some actions (like
forwarding through a particular interface), then replaces the IPv6
destination address with the new active segment.
The operations applied by the SRv6 packet processing are different at
the SR source, transit, and SR segment endpoint nodes.
Fioccola, et al. Expires 25 April 2024 [Page 4]
Internet-Draft BM for SRv6 October 2023
The processing of the SR source node corresponds to the sequence of
insertion of the SRH, composed of SIDs stored in reverse order, and
setting of the IPv6 Destination Address as the first SID of the SR
Policy. It can be performed by encapsulating a packet into an outer
IPv6 packet with an SRH.
The processing of the SR segment endpoint node corresponds to the
detection of the new active segment, which is the next segment in the
Segment List and the related modification of the IPv6 destination
address of the outer IPv6 header. Then packets are forwarded on the
basis of the IPv6 forwarding table.
The processing of the SR transit node corresponds to normal
forwarding of the packets containing the SR header. In SRv6, the
transit nodes do not need to be SRv6 aware, as every IPv6 router can
act as an SRv6 transit node since any IPv6 node will maintain a plain
IPv6 FIB entry for any prefix, no matter if the prefix represents a
segment or not.
[RFC9256] specifies the concepts of SR Policy and steering into an SR
Policy. The header of a packet steered in an SR Policy is augmented
with the ordered list of segments associated with that SR Policy. SR
Policy state is instantiated only on the headend node, which steers a
flow into an SR Policy. Intermediate and endpoint nodes do not
require any state to be maintained. SR Policies can be instantiated
on the headend dynamically and on demand basis. SR policy may be
installed by PCEP [RFC8664], BGP
[I-D.ietf-idr-segment-routing-te-policy], or via manual configuration
on the router. PCEP and BGP signaling of SR Policies can be the case
of a controller-based deployment.
In addition to the basic SRv6 packet processing, the SRv6 Network
Programming model [RFC8986] describes a set of functions that can be
associated to segments and executed in a given SRv6 node.
Examples of such functions are described in [RFC8986], but, in
practice, any behavior and function can be associated with a local
SID in a node, to apply any special processing on the packet. The
definition of a standardized set of segment routing functions
facilitates the deployment of SR domains with interoperable equipment
from multiple vendors.
According to [RFC8986], 128 bit SID can be logically split into three
fields and interpreted as LOCATOR:FUNCTION:ARGS (in short
LOC:FUNCT:ARG) where LOC includes the L most significant bits, FUNCT
the following F bits and ARG the remaining A bits, where L+F+A=128.
The LOC corresponds to an IPv6 prefix (for example with a length of
48, 56, or 64 bits) that can be distributed by the routing protocols
Fioccola, et al. Expires 25 April 2024 [Page 5]
Internet-Draft BM for SRv6 October 2023
and provides the reachability of a node that hosts some functions.
All the different functions residing in a node have a different FUNCT
code, so that their SIDs will be different. The ARG bits are used to
provide information (arguments) to a function. From the routing
point of view, the solution is scalable, as a single prefix is
distributed for a node, which implements a potentially large number
of functions and related arguments.
3. Test Methodology
3.1. Test Setup
The test setup in general is compliant with section 6 of [RFC2544]
but augmented by the methodology specified in section 4 of [RFC5695]
using many interfaces. It is needed to test the packet forwarding
engine that may have different performance based on the number of
interfaces served. The Device Under Test (DUT) may have
oversubscribed interfaces, then traffic for such interfaces should be
proportionally decreased according to the specific DUT
oversubscription ratio. All interfaces served by a particular packet
forwarding engine should be loaded in reverse proportion to the
claimed oversubscription ratio. Tests SHOULD be done with
bidirectional traffic that better reflects the real environment for
SRv6 nodes. It is OPTIONAL to choose a non-equal proportion for
upstream and downstream traffic for some specific aggregation nodes.
The RECOMMENDED topology for SRv6 Forwarding Benchmarking should be
the same used for MPLS benchmarking, as described in section 4 of
[RFC5695]. A simplified view is reported below for reference.
+----------+
+---------| |<---------+
| +-------| Tester |<-------+ |
| | +-----| |<-----+ | |
| | | +----------+ | | |
| | | | | |
| | | +--------+ | | |
| | +----->| |-------+ | |
| +------->| DUT |---------+ |
+--------->| |-----------+
+--------+
Figure 1: Test environment for SRv6 Forwarding Benchmarking
Differently from [RFC5695], this document prefers the use of the term
"interface" instead of "port" as an interface may be either virtual
or physical.
Fioccola, et al. Expires 25 April 2024 [Page 6]
Internet-Draft BM for SRv6 October 2023
The RECOMMENDED topology for SRv6 Forwarding Benchmarking should be
the same as MPLS and it is described in section 4 of [RFC5695].
Interface numbers involved in the tests and their oversubscription
ratio MUST be reported. This document is benchmarking only "source
routing". Hence, SIDs represent only Headend encapsulation
(H.Encaps.xxx) or segment Endpoint (End, End.X) that may be carried
in IGP extensions. In general, Functions of the last SID (called
"behavior" in [RFC8986]) may be used to encode services (similar to
L2/L3 VPNs and much more) but it is out of the scope of this
document.
It is OPTIONAL to test SRH in the combination with any other
extension headers (fragmentation, hop-by-hop, destination options,
etc.) but in all tests, the SRH header should be present for the test
to be relevant for SRv6. It is RECOMMENDED to follow section 5.3 of
[RFC5180] to introduce other extension headers in proportion 1%, 10%,
50% that may better reflect real use cases.
Segment Routing may also be implemented as a software network
function in an NFV Infrastructure and, in this case, additional
considerations should be done. [ETSI-GR-NFV-TST-007] describes test
guidelines for NFV capabilities that require interactions between the
components implementing NFV functionality.
Special capabilities SHOULD NOT exist in the DUT/SUT specifically for
benchmarking purposes.
3.2. Locator and Endpoint behaviors methods
As specified in [RFC8986], topological segments have the structure
that consists of Locator and Endpoint behavior (H.Encaps, End, End.X,
etc), the latter may have a few different flavors (PSP, USP, USD).
Different combinations of behavior and flavor are recommended for
every test.
It is RECOMMENDED that the DUT and test tool support at least one
option for SID stack construction:
* IS-IS Extensions to Support Segment Routing over IPv6 Dataplane
[RFC9352]
* OSPFv3 Extensions for SRv6 [I-D.ietf-lsr-ospfv3-srv6-extensions]
* Segment Routing Policy Architecture [RFC9256].
A routing protocol (OSPF or IS-IS) SHOULD be used for the
construction of the simplest SRH with 1 SID. It is RECOMMENDED that
SR policy should be used for the construction of SRH with 2 SIDs. It
Fioccola, et al. Expires 25 April 2024 [Page 7]
Internet-Draft BM for SRv6 October 2023
is OPTIONAL to test longer SRH if there is an interest. For longer
SRH lists, it is suggested to consider DUT maximum, or "network
diameter" plus one SID for the service, where "network diameter" is
the maximum number of SR hops that the target SRv6 network have, it
is typically between 3 to 7 hops.
It is RECOMMENDED that the top SID on the list should have an End.X
flavor type to emulate traffic engineering scenario. In all cases,
SID stack configuration SHOULD happen before packet forwarding would
be started. Control plane convergence speed is not the subject of
the present tests.
The Locator and Endpoint construction method and SR policy
construction method used MUST be reported according to Section 4.
3.3. Frame Formats and Sizes
SRv6 tests will use Frame characteristics similarly to section 4.1.5
of [RFC5695], except the need for a bigger MTU to accommodate SRH.
It is assumed that MTU is big enough to accommodate all frame sizes
proposed below. Fragmentation is not an option for SRv6.
It is to be noted that [RFC5695] requires exactly a single entry in
the MPLS label stack in an MPLS packet that is not enough to simulate
a typical SR SID list. The number of entries in SRH MUST be
reported.
According to section 4.1.4.2 of [RFC5695], the payload is RECOMMENDED
to have an IP packet (IPv6 or IPv4 with UDP or TCP) to better
represent the real environment. The minimal Ethernet payload (46B)
could not accommodate the whole IPv6 stack (not enough room for TCP
or UDP), hence only IPv4 is possible to use if the test for minimal
Ethernet payload is needed. It is possible to choose the bigger
payload size for the IPv6 only environment.
It is assumed that the test would be for Ethernet media only. Other
media is possible (see section 4.1.5.2 of [RFC5695] for the POS
example). Some layer 2 technologies (like POS/PPP) have bit- or
byte- stuffing then [RFC4814] may help to calculate real performance
more accurately or else 1-2% error is expected. The most popular
layer 2 technology for SR is Ethernet, it does not have stuffing.
Fioccola, et al. Expires 25 April 2024 [Page 8]
Internet-Draft BM for SRv6 October 2023
RECOMMENDED frame sizes are presented below. Any other frame sizes
may be added if suspected of abnormal behavior. For example, some
architectures may allocate buffer memory in big fixed chunks that may
drop performance if frame sizes are chosen just a few octet more than
the fixed chunk size (the second chunk would have a very low memory
utilization).
The resulting Ethernet frame structure is depicted in the next
figure.
<---18B---><-40B-><8+n*16B><--------46-1500-9000B----------->
+---------+-------+-------+---------+-----------------------+
| | Outer | | Inner | | |
| Layer 2 | IPv6 | SRH | Layer 3 | Layer 4 | High layers |
+---------+-------+-------+---------+-----------------------+
Figure 2: Ethernet Frame Structure
RECOMMENDED payload sizes (encapsulated packet with L3 headers and
above) are the following:
* Ethernet Minimal: 46
* DUT Minimal Wire Speed: typically 128-256 (it depends on the DUT
specification)
* Ethernet Typical: 1500
* DUT Maximum: 9000 (or any claimed maximum)
It is evident from the figure 2 that in all cases we need to add
40+8+n*16 bytes for the needed frame size, where
40 octets are added for the outer (tunnel) IPv6 header
8 octets are added for the SRH header itself
n is the number of segments multiplied on 16 octet SID size.
The typical frame size values are listed above for the DUT minimal
wire speed and maximum, but they can be modified according to the DUT
characteristics. The minimum wire speed frame size can be considered
based on the DUT specification but, in some cases, many tests may be
needed in the search for the real minimum wire speed frame size.
VLAN tag may additionally increase the frame size. VLAN tag tests
are OPTIONAL.
Fioccola, et al. Expires 25 April 2024 [Page 9]
Internet-Draft BM for SRv6 October 2023
3.4. Protocol Addresses
IANA reserved an IPv6 address block 2001:0002::/48 for use with IPv6
benchmark testing (see section 8 of [RFC5180]). IPv6 source and
destination addresses for the test streams SHOULD belong to the IPv6
range assigned by IANA. It is not principal what Locator blocks
would be chosen for tests. It may be /52, /56, /64, or even bigger.
It is possible to test a few different Locator blocks if there is a
need.
As it is discussed in section 3.1, there is a need to load the whole
forwarding engine (on all interfaces). [RFC4814] discusses the
importance to have many flows with address randomization for
acceptable hash-based load balancing that is implemented in all
forwarding engines. Note that IPv6 flow label randomization can be
used, according to [RFC6438]. In the context of this document, it
may also be relevant for SIDs, because SIDs may be used for hash to
choose the next link (depending on DUT default or desired
configuration). It is important to check what exactly is used for
the hash load balancing algorithm on the DUT to keep these numbers
sufficiently random and at volume. It is very often that IP
addresses and transport protocol ports are used instead of SIDs.
3.5. Trial Duration
The test portion of each trial must take into account the respective
protocol configuration. IGP protocols typically have a shorter hold
time, while some BGP default configurations may be up to 180 seconds.
It is needed to check the default hold time of the DUT for the
respective protocol used.
In general, the test portion of each trial SHOULD be no less than 250
seconds, which is a reasonable value based on common hold time
values. But a test can also adapt to the real setup and select a
different value if default configuration has been changed. The test
portion of each trial can be chosen at least 10 seconds longer than
the hold time to verify that the DUT can maintain a stable control
plane when the data-forwarding plane is under stress.
3.6. Traffic Verification
Traffic verification is following section 10 of [RFC2544] and section
4.1.8 of [RFC5695]. The text is copied here for your convenience.
"The test equipment SHOULD discard any frames received during a test
run that are not actual forwarded test frames. For example, keep-
alive and routing update frames SHOULD NOT be included in the count
of received frames. In all cases, sent traffic MUST be accounted
Fioccola, et al. Expires 25 April 2024 [Page 10]
Internet-Draft BM for SRv6 October 2023
for, whether it was received on the wrong port, the correct port, or
not received at all. In all cases, the test equipment SHOULD verify
the length of the received frames and check that they match the
expected length.
Preferably, the test equipment SHOULD include sequence numbers (or
signature) in the transmitted frames and check for these numbers on
the received frames. If this is done, the reported results SHOULD
include in addition to the number of frames dropped, the number of
frames that were received out of order, the number of duplicate
frames received and the number of gaps in the received frame
numbering sequence".
Many test tools may, by default, only verify that they have received
the embedded signature on the receive side. However, some SRv6 tests
assumes headers modifications (add or delete SRH, replace destination
address, adjust "segments left"). All packets SHOULD be checked of
the correct headers values on the receiving side.
3.7. Buffer tests
Back-to-back frame test was initially discussed in section 26.4
[RFC2544] and later improved in [RFC9004] which is considered the
comprehensive reference for Back-to-back frame test. Modern
forwarding engines are typically flexible in the buffer distribution
between different interfaces. Hence, like for all other benchmarking
tests, it is important to stress the forwarding engine on all
interfaces. It should be necessary to perform throughput tests first
because only frame sizes that stress DUT below wire-speed can be used
for back-to-back tests. Buffers would be filled with the rate equal
to the difference between the theoretical maximum frame rate (wire-
speed) and DUT measured throughput for the respective frame size.
The test time could be much shorter than recommended in [RFC9004]
because typical SR DUT is hardware-based with claimed buffers between
30ms to 100ms. It is better to consult with the vendor to find a
good starting search point. If DUT is software-based then [RFC9004]
recommendation for 2-30 seconds is applied.
Queuing SHOULD NOT have weighted random early detection (WRED) or any
other mechanism that may start dropping packets before the buffer is
filled. Queuing SHOULD be configured for the tail drop which is
typically a non-default configuration. Back-to-back frame test is
rather complex and expensive (50 runs for every frame size). Hence,
it is OPTIONAL for SRv6.
Fioccola, et al. Expires 25 April 2024 [Page 11]
Internet-Draft BM for SRv6 October 2023
4. Reporting Format
There are a few parameters that must be changed in section 5 of
[RFC5695] for SRv6 tests.
Reporting parameter preserved from [RFC5695]:
* Throughput in bytes per second and frames per second
* Frame sizes in Octets (see Section 3.3)
* Interface speed (10/50/100/400/800/etc GE)
* Interface encapsulation (Ethernet or Ethernet VLAN)
* Interface media type (probably Ethernet)
Parameters changed from [RFC5695]:
* SRv6 Forwarding Operations (PUSH/ NEXT/ CONTINUE).
* Label Distribution protocol and IGP are the same in the context of
SRv6. Hence, it can be called "Locator and Endpoint behaviors
methods".
New parameters that MUST be reported are:
* Interface numbers involved for ingress and egress in the tests and
their respective oversubscription ratio.
* Upstream/downstream traffic proportion (equal bidirectional or
some other split).
* Number of Segments considered in the SRH.
* Behavior (H.Encaps, etc.) and Flavor (PSP, USP, USD) used for
tests (according to [RFC8986]).
* SR Policy construction method (PCEP, BGP, manual configuration).
* Type of the payload (IPv6/IPv4, UDP/TCP).
* Time to recover from the overload state
* Time to recover from the reset state and reset type (particular
module in reset)
Fioccola, et al. Expires 25 April 2024 [Page 12]
Internet-Draft BM for SRv6 October 2023
* Tested buffers size in frames with respective frame size (for the
optional back-to-back test); it is possible to record calculated
buffer time for wire-speed throughput in milliseconds.
Some parameters may be the same for all tests (like Media type or
Ethernet encapsulation) then it may be reported one time.
5. SRv6 Forwarding Benchmarking Tests
In general, tests are compliant with [RFC2544] but the important
correction discussed in section 6 of [RFC2544] is applied: interfaces
chosen for every test MUST stress all interfaces served by one
forwarding engine. It is better to check the DUT specification for
the relationship between interfaces and the forwarding engine to
minimize the number of interfaces involved. But it is possible to
understand the worst case by looking at the throughput and latency
from the trial tests. If any doubt exists about how full is the
offered load for the forwarding engine then it is better to stress
all interfaces of the line card or all interfaces for the whole
router with a centralized forwarding engine. A partial load on the
forwarding engine would show optimistic results. Controllable
traffic distribution between many interfaces (as specified in section
4 of [RFC5695]) would need separate SID announcements for separate
interfaces.
The performance of modern packet forwarding engines may be huge that
may need to involve many testers to sufficiently load the DUT as
presented on figure 3. Then results correlation and recalculation of
the real performance would be an additional burden.
+----------+
+-------| Tester1 |<-------+
| +-----| |<-----+ |
| | +----------+ | |
| | | |
| | +--------+ | |
| +----->| |-------+ |
+------->| DUT |---------+
+------->| |---------+
| +----->| |-------+ |
| | +--------+ | |
| | | |
| | +----------+ | |
| +-----| Tester2 |<-----+ |
+-------| |<-------+
+----------+
Figure 3: Many testers
Fioccola, et al. Expires 25 April 2024 [Page 13]
Internet-Draft BM for SRv6 October 2023
As specified in section 6 of [RFC5695], the traffic is sent from test
tool Tx interface(s) to the DUT at a constant load for a fixed-time
interval, and is received from the DUT on test tool Rx interface(s).
If any frame loss is detected, then a new iteration is needed where
the offered load is decreased and the sender will transmit again. An
iterative search algorithm MUST be used to determine the maximum
offered frame rate with a zero frame loss (No-Drop Rate - NDR). Each
iteration should involve varying the offered load of the traffic,
while keeping the other parameters (test duration, number of
interfaces, number of addresses, frame size, etc.) constant, until
the maximum rate at which none of the offered frames are dropped is
determined.
The test can be repeated with a varying number of Segments pushed on
ingress in order to measure the resulting maximum number. It can
also be tested the maximum number of Segments that are correctly
load-balanced in transit by only changing the Nth label in the stack
and detect when load-balancing fails.
Therefore, the two main parameters that can be evaluated are:
Maximum offered frame rate,
Maximum number of Segments that can be pushed and hashed by the SR
node for load-balancing.
5.1. Throughput
This section contains a description of the tests that are related to
the characterization of a DUT's SRv6 traffic forwarding throughput.
The list of segments for SRv6 is represented as a list of IPv6
addresses, included in the SRH. There are three distinct types of
nodes that are involved in segment routing networks that may
represent four different cases.
5.1.1. Throughput of a Source Node
Objective: To obtain the DUT's Throughput during the packet
processing of a Source Node. It is when the Source SR node, which
corresponds to the headend node, encapsulates a received packet into
an outer IPv6 packet and inserts the SR Header (SRH) as a Routing
Extension Header in the outer IPv6 header. The Segment List in the
SRH is composed of SIDs and the Source SR node sets the first SID of
the SR Policy as the IPv6 Destination Address of the packet. The
RECOMMENDED headend behavior is H.Encaps, in case of interest for
another behavior (H.Encaps.Red or H.Encaps.L2 or H.Encaps.L2.Red) it
is OPTIONAL to test it with proper reporting.
Fioccola, et al. Expires 25 April 2024 [Page 14]
Internet-Draft BM for SRv6 October 2023
Procedure: Similar to section 6.1 of [RFC5695] or section 26.1 of
[RFC2544] with extension to test SID list longer than 1 SID (more
than 2 are RECOMMENDED). SID list can be from 1 to N SIDs. N could
be specified a priori or measured as part of the test. The test tool
must advertise and learn the IP prefix(es) and SID(s) on respective
sides, as per Section 3.4, and must use one option for SID stack
construction, as per Section 3.2, on its receive and transmit
interfaces towards the DUT.
Reporting Format: A table, similar to [RFC5180], with all parameters
specified in Section 4.
5.1.2. Throughput of a transit Segment Endpoint Node
Objective: To obtain the DUT's Throughput during the packet
processing of a Segment Endpoint Node. It is when the SR Segment
Endpoint node receives packets whose IPv6 destination address is
locally configured as a segment. The SR Segment Endpoint node
inspects the SR header: it detects the new active segment, i.e. the
next segment in the Segment List, modifies the IPv6 destination
address of the outer IPv6 header and forwards the packet on the basis
of the IPv6 forwarding table. The RECOMMENDED endpoint behavior is
End.X, in case of interest for another behavior (End, End.T, End.BM,
End.B6.Encaps, End.B6.Encaps.Red) it is OPTIONAL to test it with
proper reporting. SRH SL (Segment Left) is assumed to be bigger than
zero for this test. Moreover, it is assumed that DUT would not need
to delete headers (no PSP, USD, or USP).
Procedure: Similar to section 6.1 of [RFC5695] or section 26.1 of
[RFC2544] with extension to test SID list longer than 1 SID (more
than 2 are RECOMMENDED). SID list can be from 1 to N SIDs. N could
be specified a priori or measured as part of the test. The test tool
must advertise and learn the IP prefix(es) and SID(s) on respective
sides, as per Section 3.4, and must use one option for SID stack
construction, as per Section 3.2, on its receive and transmit
interfaces towards the DUT.
Reporting Format: A table, similar to [RFC5180], with all parameters
specified in Section 4.
5.1.3. Throughput of a Segment Endpoint Node with decapsulation
Objective: To obtain the DUT's Throughput during the packet
processing of a Segment Endpoint Node that needs decapsulation. It
is when the SR Segment Endpoint node receives packets whose IPv6
destination address is locally configured as a segment and SL
(segment left) in the SRH header is decremented to zero. The SR
Segment Endpoint node inspects the SR header: it detects the new
Fioccola, et al. Expires 25 April 2024 [Page 15]
Internet-Draft BM for SRv6 October 2023
active segment, i.e. the next segment in the Segment List, modifies
the IPv6 destination address of the outer IPv6 header, decapsulate
the packet, and forwards the packet on the basis of the IPv6
forwarding table. The RECOMMENDED endpoint decapsulation behavior is
End with USD flavor, in case of interest for another flavor (PSP,
USP) it is OPTIONAL to test it with proper reporting.
Procedure: Similar to section 6.1 of [RFC5695] or section 26.1 of
[RFC2544] with extension to test SID list longer than 1 SID (more
than 2 are RECOMMENDED). SID list can be from 1 to N SIDs. N could
be specified a priori or measured as part of the test. The test tool
must advertise and learn the IP prefix(es) and SID(s) on respective
sides, as per Section 3.4, and must use one option for SID stack
construction, as per Section 3.2, on its receive and transmit
interfaces towards the DUT.
Reporting Format: A table, similar to [RFC5180], with all parameters
specified in Section 4.
5.1.4. Throughput of a Transit Node
Objective: To obtain the DUT's Throughput during the packet
processing of a Transit Node. It is when a Transit node forwards the
packet containing the SR header as a normal IPv6 packet because the
IPv6 destination address does not locally match with a segment.
Procedure: Similar to section 6.1 of [RFC5695] or section 26.1 of
[RFC2544] with extension to test SID list longer than 1 SID (more
than 2 are RECOMMENDED). SID list can be from 1 to N SIDs. N could
be specified a priori or measured as part of the test. The test tool
must advertise and learn the IP prefix(es) and SID(s) on respective
sides, as per Section 3.4, and must use one option for SID stack
construction, as per Section 3.2, on its receive and transmit
interfaces towards the DUT.
Reporting Format: A table, similar to [RFC5180], with all parameters
specified in Section 4.
5.2. Buffers size
Back-to-back frame test is OPTIONAL and SHOULD be performed only
after throughput tests because it SHOULD use only frame sizes that
DUT is not capable to forward wire-speed, as explained in
Section 3.7.
Objective: To determine the buffer size as defined in section 6 of
[RFC9004] for each of the SRv6 forwarding operations.
Fioccola, et al. Expires 25 April 2024 [Page 16]
Internet-Draft BM for SRv6 October 2023
Procedure: Should be inherited from [RFC9004] with 2 SIDs RECOMMENDED
(many SIDs are possible). Despite the simple general idea for
filling the buffer until tail drop, [RFC9004] has many details for
procedure, precautions, and calculations that would be too lengthy to
copy here.
Reporting Format: A table, similar to [RFC5180], with all parameters
specified in Section 4.
5.3. Latency
Objective: To determine the latency as defined in section 6.2 of
[RFC5695] and section 26.2 of [RFC2544] for each of the SRv6
forwarding operations. It is RECOMMENDED to test all four test types
discussed in Section 5.1.
Procedure: Similar to Section 5.1. It is OPTIONAL to improve the
procedure according to section 7.2 of [RFC8219] with calculations for
typical and worst-case latency.
Reporting Format: A table, similar to [RFC5180], with all parameters
specified in Section 4.
5.4. Frame Loss
Objective: To determine the frame-loss rate (as defined in section
6.3 of [RFC5695] or and section 26.3 of [RFC2544]) for each of the
SRv6 forwarding operations of a DUT throughout the entire range of
input data rates and frame sizes. The primary idea is to see what
would be the frame loss under the overload conditions. It may be
that overloaded forwarding engine would forward less traffic than in
the situation close to the overload. Throughput may drop below the
possible maximum. As for the section 26.3 of [RFC2544], it is
RECOMMENDED to have the data for all tested frame sizes with 10% load
step above the wire-speed throughput measured in Section 5.1. It is
RECOMMENDED to test all four test types discussed in Section 5.1.
Procedure: Similar to Section 5.1.
Reporting Format: A table, similar to [RFC5180], with all parameters
specified in Section 4.
5.5. System Recovery
Objective: To characterize the speed at which a DUT recovers from an
overload condition for each of the SRv6 forwarding operations. It is
RECOMMENDED to test all four test types discussed in Section 5.1.
Fioccola, et al. Expires 25 April 2024 [Page 17]
Internet-Draft BM for SRv6 October 2023
Procedure: Similar to section 6.4 of [RFC5695] or section 26.5 of
[RFC2544]. Send a stream of frames at a rate 110% of the recorded
throughput rate or the maximum rate for the media, whichever is
lower, for at least 60 seconds. At Timestamp A reduce the frame rate
to 50% of the above rate and record the time of the last frame lost
(Timestamp B). The system recovery time is determined by subtracting
Timestamp B from Timestamp A. The test SHOULD be repeated a number
of times and the average of the recorded values being reported.
Reporting Format: A table, similar to [RFC5180], with all parameters
specified in Section 4.
5.6. Reset
All type of reset tests are OPTIONAL.
Objective: To characterize the speed at which a DUT recovers from a
device or software reset for each of the SRv6 forwarding operations.
According to section 1.3 of [RFC6201] it is possible to measure frame
loss or time stamps (depending on the test tool capability).
According to section 4 of [RFC6201] reset could be: 1) hardware, 2)
software, or 3) power interruption. All resets may be partial, i.e.
only for a particular part of hardware (line card) or software
(module). Especial interest may be to test redundant power supplies
or routing engines to make sure that reset does not affect the
traffic. Hardware reset may be soft (command for reset) or hard
(physical removal and insertion of the module). These types of reset
SHOULD be treated as different. It is OPTIONAL to test all four test
types discussed in Section 5.1, typically they would give the same
result.
Procedure: It is inherited from [RFC6201] (see it for more details).
It is simple in essence: create the traffic, initiate a reset,
measure the time for the traffic lost.
Reporting Format: A table, similar to [RFC5180], with all parameters
specified in Section 4.
6. Security Considerations
Benchmarking methodologies are limited to technology characterization
in a laboratory environment, with dedicated address space and
constraints. Special capabilities SHOULD NOT exist in the DUT/SUT
specifically for benchmarking purposes. Any implications for network
security arising from the DUT/SUT SHOULD be identical in the lab and
production networks. The benchmarking network topology is an
independent test setup and MUST NOT be connected to devices that may
forward the test traffic into a production network or misroute
Fioccola, et al. Expires 25 April 2024 [Page 18]
Internet-Draft BM for SRv6 October 2023
traffic to the test management network.
There are no specific security considerations within the scope of
this document.
7. IANA Considerations
This document has no IANA actions.
8. Acknowledgements
The authors would like to thank Al Morton, Gabor Lencse, Boris
Khasanov for the precious comments and suggestions.
9. References
9.1. Normative References
[RFC1242] Bradner, S., "Benchmarking Terminology for Network
Interconnection Devices", RFC 1242, DOI 10.17487/RFC1242,
July 1991, <https://www.rfc-editor.org/info/rfc1242>.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<https://www.rfc-editor.org/info/rfc2119>.
[RFC2544] Bradner, S. and J. McQuaid, "Benchmarking Methodology for
Network Interconnect Devices", RFC 2544,
DOI 10.17487/RFC2544, March 1999,
<https://www.rfc-editor.org/info/rfc2544>.
[RFC4814] Newman, D. and T. Player, "Hash and Stuffing: Overlooked
Factors in Network Device Benchmarking", RFC 4814,
DOI 10.17487/RFC4814, March 2007,
<https://www.rfc-editor.org/info/rfc4814>.
[RFC5180] Popoviciu, C., Hamza, A., Van de Velde, G., and D.
Dugatkin, "IPv6 Benchmarking Methodology for Network
Interconnect Devices", RFC 5180, DOI 10.17487/RFC5180, May
2008, <https://www.rfc-editor.org/info/rfc5180>.
[RFC5695] Akhter, A., Asati, R., and C. Pignataro, "MPLS Forwarding
Benchmarking Methodology for IP Flows", RFC 5695,
DOI 10.17487/RFC5695, November 2009,
<https://www.rfc-editor.org/info/rfc5695>.
Fioccola, et al. Expires 25 April 2024 [Page 19]
Internet-Draft BM for SRv6 October 2023
[RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
May 2017, <https://www.rfc-editor.org/info/rfc8174>.
[RFC8219] Georgescu, M., Pislaru, L., and G. Lencse, "Benchmarking
Methodology for IPv6 Transition Technologies", RFC 8219,
DOI 10.17487/RFC8219, August 2017,
<https://www.rfc-editor.org/info/rfc8219>.
[RFC8402] Filsfils, C., Ed., Previdi, S., Ed., Ginsberg, L.,
Decraene, B., Litkowski, S., and R. Shakir, "Segment
Routing Architecture", RFC 8402, DOI 10.17487/RFC8402,
July 2018, <https://www.rfc-editor.org/info/rfc8402>.
[RFC8754] Filsfils, C., Ed., Dukes, D., Ed., Previdi, S., Leddy, J.,
Matsushima, S., and D. Voyer, "IPv6 Segment Routing Header
(SRH)", RFC 8754, DOI 10.17487/RFC8754, March 2020,
<https://www.rfc-editor.org/info/rfc8754>.
[RFC8986] Filsfils, C., Ed., Camarillo, P., Ed., Leddy, J., Voyer,
D., Matsushima, S., and Z. Li, "Segment Routing over IPv6
(SRv6) Network Programming", RFC 8986,
DOI 10.17487/RFC8986, February 2021,
<https://www.rfc-editor.org/info/rfc8986>.
9.2. Informative References
[ETSI-GR-NFV-TST-007]
ETSI, "ETSI GR NFV-TST 007: Network Functions
Virtualisation (NFV) Release 3; Testing; Guidelines on
Interoperability Testing for MANO", 2020,
<https://www.etsi.org/deliver/etsi_gr/NFV-
TST/001_099/007/03.01.01_60/gr_NFV-TST007v030101p.pdf>.
[I-D.ietf-idr-segment-routing-te-policy]
Previdi, S., Filsfils, C., Talaulikar, K., Mattes, P., and
D. Jain, "Advertising Segment Routing Policies in BGP",
Work in Progress, Internet-Draft, draft-ietf-idr-segment-
routing-te-policy-25, 26 September 2023,
<https://datatracker.ietf.org/doc/html/draft-ietf-idr-
segment-routing-te-policy-25>.
[I-D.ietf-lsr-ospfv3-srv6-extensions]
Li, Z., Hu, Z., Talaulikar, K., and P. Psenak, "OSPFv3
Extensions for SRv6", Work in Progress, Internet-Draft,
draft-ietf-lsr-ospfv3-srv6-extensions-15, 21 June 2023,
<https://datatracker.ietf.org/doc/html/draft-ietf-lsr-
ospfv3-srv6-extensions-15>.
Fioccola, et al. Expires 25 April 2024 [Page 20]
Internet-Draft BM for SRv6 October 2023
[I-D.ietf-rtgwg-segment-routing-ti-lfa]
Litkowski, S., Bashandy, A., Filsfils, C., Francois, P.,
Decraene, B., and D. Voyer, "Topology Independent Fast
Reroute using Segment Routing", Work in Progress,
Internet-Draft, draft-ietf-rtgwg-segment-routing-ti-lfa-
11, 30 June 2023, <https://datatracker.ietf.org/doc/html/
draft-ietf-rtgwg-segment-routing-ti-lfa-11>.
[I-D.ietf-spring-srv6-srh-compression]
Cheng, W., Filsfils, C., Li, Z., Decraene, B., and F.
Clad, "Compressed SRv6 Segment List Encoding in SRH", Work
in Progress, Internet-Draft, draft-ietf-spring-srv6-srh-
compression-08, 12 September 2023,
<https://datatracker.ietf.org/doc/html/draft-ietf-spring-
srv6-srh-compression-08>.
[RFC6201] Asati, R., Pignataro, C., Calabria, F., and C. Olvera,
"Device Reset Characterization", RFC 6201,
DOI 10.17487/RFC6201, March 2011,
<https://www.rfc-editor.org/info/rfc6201>.
[RFC6438] Carpenter, B. and S. Amante, "Using the IPv6 Flow Label
for Equal Cost Multipath Routing and Link Aggregation in
Tunnels", RFC 6438, DOI 10.17487/RFC6438, November 2011,
<https://www.rfc-editor.org/info/rfc6438>.
[RFC8200] Deering, S. and R. Hinden, "Internet Protocol, Version 6
(IPv6) Specification", STD 86, RFC 8200,
DOI 10.17487/RFC8200, July 2017,
<https://www.rfc-editor.org/info/rfc8200>.
[RFC8664] Sivabalan, S., Filsfils, C., Tantsura, J., Henderickx, W.,
and J. Hardwick, "Path Computation Element Communication
Protocol (PCEP) Extensions for Segment Routing", RFC 8664,
DOI 10.17487/RFC8664, December 2019,
<https://www.rfc-editor.org/info/rfc8664>.
[RFC9004] Morton, A., "Updates for the Back-to-Back Frame Benchmark
in RFC 2544", RFC 9004, DOI 10.17487/RFC9004, May 2021,
<https://www.rfc-editor.org/info/rfc9004>.
[RFC9256] Filsfils, C., Talaulikar, K., Ed., Voyer, D., Bogdanov,
A., and P. Mattes, "Segment Routing Policy Architecture",
RFC 9256, DOI 10.17487/RFC9256, July 2022,
<https://www.rfc-editor.org/info/rfc9256>.
Fioccola, et al. Expires 25 April 2024 [Page 21]
Internet-Draft BM for SRv6 October 2023
[RFC9352] Psenak, P., Ed., Filsfils, C., Bashandy, A., Decraene, B.,
and Z. Hu, "IS-IS Extensions to Support Segment Routing
over the IPv6 Data Plane", RFC 9352, DOI 10.17487/RFC9352,
February 2023, <https://www.rfc-editor.org/info/rfc9352>.
Authors' Addresses
Giuseppe Fioccola
Huawei Technologies
Palazzo Verrocchio, Centro Direzionale Milano 2
20054 Segrate (Milan)
Italy
Email: giuseppe.fioccola@huawei.com
Eduard Vasilenko
Huawei Technologies
17/4 Krylatskaya str.
Moscow
Email: vasilenko.eduard@huawei.com
Paolo Volpato
Huawei Technologies
Palazzo Verrocchio, Centro Direzionale Milano 2
20054 Segrate (Milan)
Italy
Email: paolo.volpato@huawei.com
Luis Miguel Contreras Murillo
Telefonica
Spain
Email: luismiguel.contrerasmurillo@telefonica.com
Bruno Decraene
Orange
France
Email: bruno.decraene@orange.com
Fioccola, et al. Expires 25 April 2024 [Page 22]