SFC WG T. Ao
Internet-Draft Individual
Intended status: Informational G. Mirsky
Expires: December 6, 2019 ZTE Corp.
June 4, 2019
Analysis of the SFC scalability
draft-ao-sfc-scalability-analysis-05
Abstract
SFC is an ordered set of service function, should be scalable to meet
broad range of requirements. The scalability of SFC can be
interpreted as ability of the SFC to accommodate one or more SFs
joining the SFC , or leaving the SFC without significant impact to
SFC performance.
This document presents four aspects on SFC scalability, and provide
analysis of the data plane and the control plane to implement the
scalable SFC.
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 December 6, 2019.
Copyright Notice
Copyright (c) 2019 IETF Trust and the persons identified as the
document authors. All rights reserved.
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
Ao & Mirsky Expires December 6, 2019 [Page 1]
Internet-Draft Analysis of the SFC scalability June 2019
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 2
3. Four Use cases for scale-out/scale-in . . . . . . . . . . . . 3
3.1. Join . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3.2. Redundancy . . . . . . . . . . . . . . . . . . . . . . . 3
3.2.1. SF Redundancy . . . . . . . . . . . . . . . . . . . . 3
3.2.2. SFC Redundancy . . . . . . . . . . . . . . . . . . . 4
3.3. By-pass . . . . . . . . . . . . . . . . . . . . . . . . . 5
3.4. Failure or Remove . . . . . . . . . . . . . . . . . . . . 5
4. Data Plane Requirements . . . . . . . . . . . . . . . . . . . 6
5. Control Plane Requirements . . . . . . . . . . . . . . . . . 6
5.1. Centralized CP . . . . . . . . . . . . . . . . . . . . . 6
5.2. Distributed CP . . . . . . . . . . . . . . . . . . . . . 7
6. Security Considerations . . . . . . . . . . . . . . . . . . . 7
7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 8
8. Information References . . . . . . . . . . . . . . . . . . . 8
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 8
1. Introduction
Service Function Chain (SFC) is the chain with a series of ordered
Service Functions(SF). The SFC maybe changed because of load balance
, failure, or other management requirement. We call it SFC
scalability. The SFC being scalable means that the Service Functions
can be added or removed from the path of this SFC without impact on
other SFCs and minimal impact in the SFC being modified. With this
capability, SFC is more flexible and elastic to adapt all kinds of
requirements.
In this document, we will present four use cases on SFC scale-out and
scale-in, and analysis some requirements to support SFC scalability.
2. Terminology
SFC(Service Function Chain): An ordered set of some abstract SFs.
SFC Scale-out: One or more SFs are added into the path of the SFC for
the sake of load balance, protection or other new services
requirement.
Ao & Mirsky Expires December 6, 2019 [Page 2]
Internet-Draft Analysis of the SFC scalability June 2019
SFC Scale-in: One or more SFs are removed from the path of the SFC
for the sake of the SFs are by-passed or the SFs are failed.
3. Four Use cases for scale-out/scale-in
Following describes four use cases to illustrate the scalability of
the SFC.
3.1. Join
This is SFC horizontal scale-out use case. One or more new SFs must
be added to a certain SFC for the traffic that has been classified to
require application of new SF(s). This case is the reverse scenario
to the by-pass. In this case one or more SFs that were by-passed
need to be re-inserted into the SFC. And the SFC itself can be
characterized as being scaled out.
There are two sub-cases of an SF joining the SFC. One when both the
SF and corresponding SFF are new to the SFC. The second is when the
SF attaches to an existing SFF. In the first scenario, control plane
needs to notify the upstream SFF to modify its next hop to point to
the new SFF and configure the new SFF's forwarding information. In
the second scenario control plane needs to configure the existing
SFF's forwarding information. In this scenario, SFF forwards the
packets not only according to the SFPID but also according to the
metadata in the SFC header.
3.2. Redundancy
3.2.1. SF Redundancy
This is an example of SFC vertical scale-out use case. One or more
SFs are added into the SFC to meet the redundancy or load balance
requirements for some certain SFs. This case is different from the
Join case (section 3.1) in which the SF in this case is the same with
one of the SF that is on the path of the SFC. The new SF have the
same function with the existing SF, so that the new SF is added into
the SFC to protect the existing corresponding SF and to load balance
the existing corresponding SF. Figure 1 is the illustration about SF
redundancy. In this figure, SF2' is the redundency of SF2, so that
when SF2 is down, SF2' can keep working.
Ao & Mirsky Expires December 6, 2019 [Page 3]
Internet-Draft Analysis of the SFC scalability June 2019
+-----------+
| SFC | +----+ +----+ +----+
|Classifier |---->|SFF1|----->|SFF2|------->|SFF3|
| Node | | | | | | |
+-----------+ +----+ +----+ +----+
| | |
| -------- |
| | | +-----------+
+----+ +----+ +----+ | SFC Proxy |
| SF1| | SF2| |SF2'| +-----------+
+----+ +----+ +----+
Figure 1
In this case, control plane need to notify the upstream SFF that a
new SF joins the SFC as a redundancy SF for protection or load
balance, and its next hop should be a protection group or ECMP group.
For the purpose of load balance to ensure proper forwarding, the Flow
Id field MUST be presented in the NSH as expression of entropy so
that SFF can select an SF from the group according to the Flow Id.
In the above figure, SFF2 knows that it is connecting a group of SFs
and when it foward the packet, it would use Flow id in NSH.
3.2.2. SFC Redundancy
This is also an example of SFC vertical scal-out use case, namely
Reduncancy. In this case, SFC is scaled out to two SFP paths. One
SFP is redundant to another SFP, and the two SFPs are for protection
or load balance. They belongs to a SFC, but have different SFP. The
two SFPs are forming a group. Figure 2 is the illustration about the
SFC redundancy. In this figure, we can see that SF1', SF2', SFC
proxy' are the backup of the SF1, SF2, SFC Proxy seperately. The two
SFPs are a group for the Classifier. All these nodes can be joint at
some nodes and can be disjoint as well. In the figure 2, all the
nodes are disjoint.
Ao & Mirsky Expires December 6, 2019 [Page 4]
Internet-Draft Analysis of the SFC scalability June 2019
+-----+ +-----+ +-----------+
| SF1 | | SF2 | | SFC Proxy |
+-----+ +-----+ +-----------+
+-----------+ | | |
| SFC | +-----+ +---- + +-----+
|Classifier |---->|SFF1 |----->|SFF2 |------->|SFF3 |
| Node | | | | | | |
+-----------+ +-----+ +-----+ +-----+
|
|
| +-----+ +-----+ +-----+
----------->|SFF1'|----->|SFF2'|------->|SFF3'|
| | | | | |
+-----+ +-----+ +-----+
| | |
+-----+ +-----+ +-----------+
| SF1'| | SF2'| | SFC Proxy'|
+-----+ +-----+ +-----------+
Figure 2
In this case, control plane need to notify the Classifier that the
SFC is a group which contains two SFPs. The group can be used as
protection or load balance. For the purpose of load balance, to
ensure proper forwarding, the Flow Id field MUST be presented in the
NSH as expression of entropy so that the forwarder in the classifier
can select an SFP from the group according to the Flow Id. For the
case of joint, the joint node also need to have capability to forward
the traffic accroding to the Flow ID.
3.3. By-pass
This is an example of horizontal scale-in case. In this scenario
some SFs are not removed from the SFC but just by-passed by the
traffic so that the packets will not be processed by these SFs. Use
cases for this scenario are described in [draft-ietf-sfc-long-lived-
flow-use-cases] and [draft-ietf-sfc-offloads] . In these two drafts,
the SF is offloaded because it is not necessary to steer the traffic
to the SFs to improve the forwarding performance.
The corresponding solution is also provided in the above drafts.
3.4. Failure or Remove
This is a vertical SFC scale-in case. This happens only when the SFC
is being protected or load balanced. When SF of one SFC has failed
or needs to be removed because it is no longer needed to do the
pretection, the ability of the SFC to scale-in is excercised.
Ao & Mirsky Expires December 6, 2019 [Page 5]
Internet-Draft Analysis of the SFC scalability June 2019
In this case, the upstream SFF MUST be notified that its next hop has
been changed to the next SF of the SF.
From the cases described we can conclude that no matter if is SFC
scale-out case or scale-in cases, there are some requirements to SFC
control protocol. And for some cases, there are requirements to data
plane as well.
4. Data Plane Requirements
For the cases of load balancing or protection switchover of SFC
scalability, it is highly beneficial to have an entropy field in the
SFC header NSH. The entropy may be presented in the dedicated field
named as Flow ID which be part of SFC encapsulation.
This means that SFF not only forwards the traffic based on different
SFPID, but also MAY use Flow ID to select particular SF out of set of
SFs of the same type.
According to the NSH draft in draft--ietf-sfc-nsh-27, we propose to
extend NSH to include the entropy field. Two options can be
considered. One is to use existing field, for example, some reserved
bits. Suggested extended field in NSH Service Path Header is showed
in Figure 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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Service Path Identifier (SPI) | Service Index |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Reserved | Flow ID |
+-----------------------------------------------+---------------+
Figure 3
Another is to extend a new metadata to meet the requirement. Which
has been described in the section 8 of the draft-quinn-sfc-nsh-tlv-04
.
5. Control Plane Requirements
5.1. Centralized CP
SFC Controller is required to:
a) Send a message to SFF that the joined SF connected to set the
correct SFPID and its next hop.
b) Send register message to upstream SFF or classifier with some
information. The information not only includes next hop locator, but
Ao & Mirsky Expires December 6, 2019 [Page 6]
Internet-Draft Analysis of the SFC scalability June 2019
also includes an indicator if the next hop is a new joined SF or a
group that a new SF that added into. If the indicator is a new
joined SF, it means the new SF will join the SFC. If the indicator
is a group, it means a new SF or a new SFP will be added into this
group for load balance or protection.
c) Send de-register message to upstream SFF or classifier with some
information. The information not only includes next hop locator, but
also includes an indicator that if the next hop is by-passed, or the
next hop is removed from a group. If the indicator is the by-passed
SF, it means the current SF is by-passed or is leaving from the SFC.
If the indicator is a group SF, it means the current SF or SFP will
be removed from a protection group that is for load balance or
protection.
5.2. Distributed CP
Distributed SFC CP can be used in Plug-and-Play scenario.
Distributed SFC CP required:
a) The SF that needs to join into the SFC or be by-passed by the SFC
should explicitly notify the SFF it is associated with.
b) Once get the connection notification from the SF, the associated
SFF should send a register message to the upstream SFF with some
information. Such information not only includes next hop locator,
but also includes an indicator that if the next hop is a new joined
SF or the next hop is a new SF that added into a group. If the
indicator is a new joined SF, it means a new SF will join the SFC.
If the indicator is a group, it means a new SF will be added into a
group for load balance or protection.
c) The SFF send de-register message to upstream SFF with some
information. Such information not only includes next hop locator,
but also includes an indicator that if the next hop is the next SF
because the current SF is by-passed, or the next hop is the SF that
is removed from a group. If the indicator is the by-passed SF, it
means the current SF is by-passed or is leaving from the SFC. If the
indicator is group SF, it means the current SF will be removed into a
protection group that is for load balance or protection.
6. Security Considerations
For the scalability of the SFC, security is very important to be
considered. Before allow the SF to join to the SFC, it is required
to make sure the SF's security first.
Ao & Mirsky Expires December 6, 2019 [Page 7]
Internet-Draft Analysis of the SFC scalability June 2019
7. IANA Considerations
TBD
8. Information References
[I-D.ietf-sfc-architecture]
Halpern, J. and C. Pignataro, "Service Function Chaining
(SFC) Architecture", draft-ietf-sfc-architecture-11 (work
in progress), July 2015.
[I-D.ietf-sfc-long-lived-flow-use-cases]
Krishnan, R., Ghanwani, A., Halpern, J., Kini, S., and D.
Lopez, "SFC Long-lived Flow Use Cases", draft-ietf-sfc-
long-lived-flow-use-cases-03 (work in progress), February
2015.
[I-D.ietf-sfc-nsh]
Quinn, P., Elzur, U., and C. Pignataro, "Network Service
Header (NSH)", draft-ietf-sfc-nsh-28 (work in progress),
November 2017.
[I-D.ietf-sfc-offloads]
Kumar, S., Guichard, J., Quinn, P., Halpern, J., and S.
Majee, "Service Function Simple Offloads", draft-ietf-sfc-
offloads-00 (work in progress), April 2017.
[RFC7498] Quinn, P., Ed. and T. Nadeau, Ed., "Problem Statement for
Service Function Chaining", RFC 7498,
DOI 10.17487/RFC7498, April 2015,
<https://www.rfc-editor.org/info/rfc7498>.
Authors' Addresses
Ting Ao
Individual
No.889, BiBo Road
Shanghai 201203
China
Phone: +86 17721209283
Email: 18555817@qq.com
Ao & Mirsky Expires December 6, 2019 [Page 8]
Internet-Draft Analysis of the SFC scalability June 2019
Greg Mirsky
ZTE Corp.
1900 McCarthy Blvd. #205
Milpitas, CA 95035
USA
Email: gregimirsky@gmail.com
Ao & Mirsky Expires December 6, 2019 [Page 9]