Internet DRAFT - draft-dunbar-idr-5g-edge-compute-bgp-usage
draft-dunbar-idr-5g-edge-compute-bgp-usage
Network Working Group L. Dunbar
Internet Draft Futurewei
Intended status: Informational K. Majumdar
Expires: January 9, 2023 Microsoft
H. Wang
Huawei
G. Mishra
Verizon
July 11, 2022
BGP Usage for 5G Edge Computing Service Metadata
draft-dunbar-idr-5g-edge-compute-bgp-usage-00
Abstract
This draft describes the problems in the 5G Edge computing
environment and how BGP can be used to propagating additional
information, so that the ingress routers in the 5G Local Data
Network can make path selection not only based on the routing
distance but also the running environment of the destinations.
The goal is to improve latency and performance for 5G EC
services.
Status of this Memo
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.
This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79. This document may not be
modified, and derivative works of it may not be created,
except to publish it as an RFC and to translate it into
languages other than English.
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."
xxx, et al. Expires January 9, 2023 [Page 1]
Internet-Draft AppMetaData for 5G EC Service
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
This Internet-Draft will expire on April 7, 2021.
Copyright Notice
Copyright (c) 2022 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
(http://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 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.............................................. 3
1.1. 5G Edge Computing Background......................... 3
1.2. 5G Edge Computing Network Properties................. 4
1.3. Problem#1: ANYCAST in 5G EC Environment.............. 5
1.4. Problem #2: Unbalanced Anycast Distribution due to UE
Mobility.................................................. 7
1.5. Problem 3: Application Server Relocation............. 7
2. Conventions used in this document......................... 8
3. Usage of AppMetaData for 5G Edge Computing................ 9
3.1. Assumptions.......................................... 9
3.2. IP Layer Metrics to Gauge Application Behavior...... 10
3.3. AppMetaData Constrained Optimal Path Selection...... 11
4. Soft Anchoring of an ANYCAST Flow........................ 11
5. Manageability Considerations............................. 13
6. Security Considerations.................................. 13
7. IANA Considerations...................................... 14
8. References............................................... 14
8.1. Normative References................................ 14
8.2. Informative References.............................. 15
Dunbar, et al. Expires January 9, 2023 [Page 2]
Internet-Draft AppMetaData for 5G EC Service
9. Acknowledgments.......................................... 15
1. Introduction
This document describes the problems in the 5G Edge computing
environment and how BGP can be used to propagating additional
information, so that the ingress routers in the 5G Local Data
Network can make path selection not only based on the routing
distance but also the running environment of the destinations.
The goal is to improve latency and performance for 5G Edge
Computing services.
1.1. 5G Edge Computing Background
In 5G Edge Computing (EC), one Application can be hosted on
multiple Servers in different EC data centers that are close
in proximity. The 5G Local Data Networks (LDN)that connect the
EC data centers with the 5G Base stations consist of a small
number of dedicated routers.
When a User Equipment (UE) initiates application packets using
the destination address from a DNS reply or its cache, the
packets from the UE are carried in a PDU session through 5G
Core [5GC] to the 5G UPF-PSA (User Plan Function - PDU Session
Anchor). The UPF-PSA decapsulates the 5G GTP outer header and
forwards the packets from the UEs to its directly connected
Ingress router of the 5G LDN. The LDN for 5G EC is responsible
for forwarding the packets to the intended destinations.
When the UE moves out of coverage of its current gNB (next-
generation Node B) and anchors to a new gNB, the 5G SMF
(Session Management Function) could select the same UPF or a
new UPF for the UE per standard handover procedures described
in 3GPP TS 23.501 and TS 23.502. If the UE is anchored to a
new UPF-PSA when the handover process is complete, the packets
to/from the UE is carried by a GTP tunnel to the new UPF-PSA.
Per TS 23.501-h20 Section 5.8.2, the UE may maintain its IP
address when anchored to a new UPF-PSA unless the new UFP-PSA
belongs to different mobile operators. 5GC may maintain a path
from the old UPF to the new UPF for a short time for the SSC
[Session and Service Continuity] mode 3 to make the handover
process more seamless.
Dunbar, et al. Expires January 9, 2023 [Page 3]
Internet-Draft AppMetaData for 5G EC Service
1.2. 5G Edge Computing Network Properties
In this document, 5G Edge Computing Network refers to multiple
Local IP Data Networks (LDN) in one region that interconnect
the Edge Computing data centers. Those IP LDN networks are the
N6 interfaces from 3GPP 5G perspective.
The ingress routers to the 5G Edge Computing Network are the
routers directly connected to 5G UPFs. The egress routers to
the 5G Edge Computing [EC] Network are the routers that have a
direct link to the EC servers. The EC servers and the egress
routers are co-located. Some of those Edge Computing Data
centers may have virtual switches or Top of Rack [ToR]
switches between the egress routers and the servers. But
transmission delay between the egress routers and the EC
servers is negligible, which is too small to be considered in
this document.
When multiple EC servers are attached to one App Layer Load
Balancer, only the IP addresses of the App Layer Load Balancer
are visible to the 5G LDNs. How an App Layer Load balancer
manages the individual servers is out of the scope of the
network layer.
The 5G EC Services are registered premium services that
require super-low latency and very high SLA. Most services by
the UEs are not part of the registered 5G EC Services.
Dunbar, et al. Expires January 9, 2023 [Page 4]
Internet-Draft AppMetaData for 5G EC Service
+--+
|UE|---\+---------+ +------------------+
+--+ | 5G | +--------+ | S1: aa08::4450 |
+--+ | Site +--+-+---+ +----+ |
|UE|----| A |PSA1| Ra| | R1 | S2: aa08::4460 |
+--+ | +----+---+ +----+ |
+---+ | | | | | S3: aa08::4470 |
|UE1|---/+---------+ | | +------------------+
+---+ |IP Network | L-DN1
|(3GPP N6) |
| | | +------------------+
| UE1 | | | S1: aa08::4450 |
| moves to | +----+ |
| Site B | | R3 | S2: aa08::4460 |
v | +----+ |
| | | S3: aa08::4470 |
| | +------------------+
| | L-DN3
+--+ | |
|UE|---\+---------+ | | +------------------+
+--+ | 5G | | | | S1: aa08::4450 |
+--+ | Site +--+--+---+ +----+ |
|UE|----| B |PSA2| Rb | | R2 | S2: aa08::4460 |
+--+ | +--+-+----+ +----+ |
+--+ | | +-----------+ | S3: aa08::4470 |
|UE|---/+---------+ +------------------+
+--+ L-DN2
Figure 1: App Servers in different edge DCs
1.3. Problem#1: ANYCAST in 5G EC Environment
Increasingly, Anycast is used by various application providers
and CDNs because Anycast provides better and faster resiliency
to failover events than GEO database DNS-based load balancing,
which relies on DNS to provide a different IP based on source
address.
Anycast address leverages the proximity information present in
the network (routing) layer. It eliminates the single point of
failure and bottleneck at the DNS resolvers. Anycast address
can be assigned to multiple app layer load balancers to
Dunbar, et al. Expires January 9, 2023 [Page 5]
Internet-Draft AppMetaData for 5G EC Service
leverage network condition for balanced forwarding. Another
benefit of using the ANYCAST address is removing the
dependency on UEs. Some UEs (or clients) might use their
cached IP addresses for an extended period instead of querying
DNS.
Client using Virtual IP address is a common practice in Cloud
Native networking, e.g., Kubernetes, to scale dynamic changes
of app servers' instantiations. Virtual IP requires the
destination gateway node to perform address translation for
return traffic, which is unsuitable for underlay network nodes
with millions of flows passing by. The Cloud Native network
can also leverage network condition to balance forwarding
among multiple Cloud Gateway nodes by assigning the same
virtual IP address (ANYCAST).
Having multiple locations of the same IP address in the 5G EC
LDN can be problematic if path selection is solely based on
routing cost as the routing cost differences to reach
different egress routers can be very small. This list
elaborates the issues in detail:
a) Path Selection: When a new flow comes to an ingress node
(Ra), how to avoid instability with Anycast flipping
between paths to the same address. The problem also
exists in the BGP multipath environment, with the optimal
path selected based on routing cost metrics.
b) Ingress node forwards the packets from one flow to the
same ANYCAST server.
a.k.a. Flow Affinity, or Flow-based load balancing.
Almost all vendors have supported flow or session based
ECMP load balancing and not per packet to avoid out of
order packets for decades. When a flow is hashed to an
ECMP path, the flow remains on that path for the life of
the flow until the flow ends.
The ingress node, (Ra/Rb), can use Flow ID (in IPv6
header) or UDP/TCP port number combined with the source
address to enforce packets in one flow being placed in
one tunnel to one Egress router. No new features are
needed.
Dunbar, et al. Expires January 9, 2023 [Page 6]
Internet-Draft AppMetaData for 5G EC Service
c) When a UE moves to a new 5G site in the middle of a
communication session with an EC server, a method is
needed to stick the flow to the same EC server, which is
required by 5G Edge Computing: 3GPP TR 23.748. [5g-edge-
compute-sticky-service] describes several approaches to
achieve stickiness in the IPv6 domain.
Note: most EC services have shorter sessions, e.g.,
shorter TCP sessions. Most likely, when a UE is moving to
a new 5G site, the TCP session via the old UPF to an EC
server is already finished. Only a very small percentage
of registered EC services need to stick to the original
server when handover to a new cell tower.
From BGP perspective, the multiple servers with the same IP
address (ANYCAST)attached to different egress routers is the
same as multiple next hops for the IP address.
This draft describes the BGP UPDATE to enable ingress routers
to take the App Server load, the capacity index, and the
location preference into consideration when computing the
optimal path to the egress routers.
1.4. Problem #2: Unbalanced Anycast Distribution due to UE
Mobility
Usually, higher capacity EC servers are placed in a metro data
center to accommodate more UEs in the proximity needing the
services, and fewer are placed in remote sites. When there is
a special event occurring at a remote site for a short period,
e.g., 1~2 days, the EC servers in the remote site might be
heavily utilized. In contrast, the EC servers of the same app
in the metro DC can be very underutilized. Since the condition
can be short-lived, it might not make business sense to adjust
EC capacity among DCs. Sometimes, UEs swarming to a specific
site are not anticipated.
1.5. Problem 3: Application Server Relocation
When an EC server is added to, moved, or deleted from a 5G EC
Data Center, the routing protocol needs to propagate the
Dunbar, et al. Expires January 9, 2023 [Page 7]
Internet-Draft AppMetaData for 5G EC Service
changes to 5G PSA or the PSA adjacent routers. After the
change, the cost associated with the site might change as
well.
Note: for ease of description, the Edge Application Server and
Application Server are used interchangeably throughout this
document.
2. Conventions used in this document
A-ER: Egress Router to an Application Server, [A-ER] is
used to describe the last router that the
Application Server is attached. For a 5G EC
environment, the A-ER can be the gateway router to
a (mini) Edge Computing Data Center.
Application Server: An application server is a physical or
virtual server that hosts the software system for
the application.
Application Server Location: Represent a cluster of servers at
one location serving the same Application. One
application may have a Layer 7 Load balancer,
whose address(es) are reachable from an external
IP network, in front of a set of application
servers. From an IP network perspective, this
whole group of servers is considered as the
Application server at the location.
Edge Application Server: used interchangeably with Application
Server throughout this document.
EC: Edge Computing
Edge Hosting Environment: An environment providing the support
required for Edge Application Server's execution.
NOTE: The above terminologies are the same as
those used in 3GPP TR 23.758
Edge DC: Edge Data Center, which provides the Edge
Computing Hosting Environment. An Edge DC might
Dunbar, et al. Expires January 9, 2023 [Page 8]
Internet-Draft AppMetaData for 5G EC Service
host 5G core functions in addition to the
frequently used application servers.
gNB next generation Node B
L-DN: Local Data Network
PSA: PDU Session Anchor (UPF)
SSC: Session and Service Continuity
UE: User Equipment
UPF: User Plane Function
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL
NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT
RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be
interpreted as described in BCP 14 [RFC2119] [RFC8174] when,
and only when, they appear in all capitals, as shown here.
3. Usage of AppMetaData for 5G Edge Computing
AppMetaData consists of metrics about the running environment
at the egress routers to which EC servers are directly
attached.
3.1. Assumptions
From the IP Layer, the EC servers or their respective load
balancers are identified by their IP addresses. Those IP
addresses are the identifiers to the EC servers throughout
this document. Here are some assumptions about the 5G EC
services:
- Only the registered EC services, which are only a small
portion of the services, need to incorporate the
destination capacity metrics for optimal forwarding.
- The 5G EC controller or management system can send those
EC service identifiers to relevant routers.
- The ingress routers' local BGP path compute algorithm
includes a special plugin that can compute the path to
Dunbar, et al. Expires January 9, 2023 [Page 9]
Internet-Draft AppMetaData for 5G EC Service
the optimal Next Hop (egress router) based on the BGP
AppMetaData TLV received for the registered EC services.
The proposed solution is for the egress routers, a.k.a. A-ERs
in this document, that have direct links to the EC Servers to
collect various measurements about the Servers' running status
[5G-EC-Metrics] and advertise the metrics to other routers in
5G EC LDN (Local Data Network).
3.2. IP Layer Metrics to Gauge Application Behavior
[5G-EC-Metrics] describes the IP Layer Metrics that can gauge
the application servers running status and environment:
- IP-Layer Metric for App Server Load Measurement:
The Load Measurement to an App Server is a weighted
combination of the number of packets/bytes to the App Server
and the number of packets/bytes from the App Server which
are collected by the A-ER to which the App Server is
directly attached.
The A-ER is configured with an ACL that can filter out the
packets for the Application Server.
- Capacity Index
a numeric number, configured on all A-ERs in the domain
consistently, is used to represent the capacity of the
application server attached to an A-ER. At some sites, the
IP address exposed to the A-ER is the App Layer Load
balancer that have many instances attached. At other sites,
the IP address exposed is the server instance itself.
- Site preference index:
is used to describe some sites are more preferred than
others. For example, a site with higher bandwidth has a
higher preference number than other.
In this document, the term "Application Server Egress Router"
[A-ER] is used to describe the last router that an Application
Server is attached. For the 5G EC environment, the A-ER can be
the gateway router to the EC DC where multiple Application
servers are hosted.
Dunbar, et al. Expires January 9, 2023 [Page 10]
Internet-Draft AppMetaData for 5G EC Service
3.3. AppMetaData Constrained Optimal Path Selection
The main benefit of using ANYCAST is to leverage the network
layer conditions to select an optimal path to the application
instantiated in multiple locations.
When the ingress routers to the 5G LDN are informed of the
Load and Capacity Index of the App Servers at different EC
data centers, they can incorporate those metrics with the
network path conditions for path selection.
Here is an algorithm that computes the cost to reach the App
Servers attached to Site-i relative to another site, say Site-
b. When the reference site, Site-b, is plugged in the formula,
the cost is 1. So, if the formula returns a value less than 1,
the cost to reach Site-i is less than reaching Site-b.
CP-b * Load-i Pref-b * Network-Delay-i
Cost-i= (w *(----------------) + (1-w) *(-------------------------))
CP-i * Load-b Pref-i * Network-Delay-b
Load-i: Load Index at Site-i, it is the weighted
combination of the total packets or/and bytes sent to and
received from the Application Server at Site-i during a
fixed time period.
CP-i: capacity index at Site-i, a higher value means higher
capacity.
Delay-i: Network latency measurement (RTT) to the A-ER that
has the Application Server attached at the site-i.
Pref-i: Preference index for the Site-i, a higher value
means higher preference.
w: Weight for load and site information, which is a value
between 0 and 1. If smaller than 0.5, Network latency and
the site Preference have more influence; otherwise, Server
load and its capacity have more influence.
4. Soft Anchoring of an ANYCAST Flow
"Sticky Service" in the 3GPP Edge Computing specification
(3GPP TR 23.748) is about flows from a UE sticking to a
specific location when the UE moves from one 5G Site to
another.
Dunbar, et al. Expires January 9, 2023 [Page 11]
Internet-Draft AppMetaData for 5G EC Service
"Soft Anchoring" is a mechanism for ingress routers to apply
preference to the path towards the previous server location
when the UE is anchored to a new UPF and continue using its
cached IP for the EC server.
Let's assume one application "App.net" is instantiated on
four servers that are attached to four different routers R1,
R2, R3, and R4 respectively. It is desired for packets to the
"App.net" from UE-1 to stick with one server, say the App
Server attached to R1, even when the UE moves from one 5G
site to another. However, when there is a failure reaching R1
or the Application Server attached to R1, the packets of the
flow "App.net" from UE-1 need to be forwarded to the
Application Server attached to R2, R3, or R4.
We call this kind of sticky service "Soft Anchoring", meaning
that anchoring to the site of R1 is preferred, but other
sites can be chosen when the preferred site encounters a
failure.
Here is a mechanism to achieve Soft Anchoring:
- Assign a group of ANYCAST addresses to one application.
For example, "App.net" is assigned with 4 ANYCAST
addresses, L1, L2, L3, and L4. L1/L2/L3/L4 represents
the location preferred ANYCAST addresses.
- For the App.net Server attached to a router, the router
has four Stub links to the same Server, L1, L2, L3, and
L4 respectively. The cost to L1, L2, L3, and L4 is
assigned differently for different egress routers. For
example,
o When attached to R1, the L1 has the lowest cost,
say 10, when attached to R2, R3, and R4, the L1 can
have a higher cost, say 30.
o ANYCAST L2 has the lowest cost when attached to R2,
higher cost when attached to R1, R3, R4
respectively.
o ANYCAST L3 has the lowest cost when attached to R3,
higher cost when attached to R1, R2, R4
respectively, and
Dunbar, et al. Expires January 9, 2023 [Page 12]
Internet-Draft AppMetaData for 5G EC Service
o ANYCAST L4 has the lowest cost when attached to R4,
higher cost when attached to R1, R2, R3
respectively
- When a UE queries for the "App.net" for the first time,
the DNS reply has the location preferred ANYCAST
address, say L1, based on where the query is initiated.
- When the UE moves from one 5G site-A to Site-B, UE
continues sending packets of the "App.net" to ANYCAST
address L1. The routers will continue sending packets to
R1 because the total cost for the App.net instance for
ANYCAST L1 is lowest at R1. If any failure occurs making
R1 not reachable, the packets of the "App.net" from UE-1
will be sent to R2, R3, or R4 (depending on the total
cost to reach L1 attached to R2/R3/R4).
If the Application Server supports the HTTP redirect, more
optimal forwarding can be achieved.
- When a UE queries for the "App.net" for the first time,
the global DNS reply has the ANYCAST address G1, which
has the same cost regardless of where the Application
servers are attached.
- When the UE initiates the communication to G1, the
packets from the UE will be sent to the Application
Server that has the lowest cost, say the Server attached
to R1. The Application server is instructed with HTTPs
Redirect to reply with a location-specific URL, say
App.net-Loc1. The client on the UE will query the DNS
for App.net-Loc1 and get the response of ANYCAST L1. The
subsequent packets from the UE-1 for App.net are sent to
L1.
5. Manageability Considerations
To be added.
6. Security Considerations
To be added.
Dunbar, et al. Expires January 9, 2023 [Page 13]
Internet-Draft AppMetaData for 5G EC Service
7. IANA Considerations
Here are new Sub-TLV types requiring IANA registration:
Type = TBD1: Aggregated Load Measurement Index derived from
the Weighted combination of bytes/packets sent to/received
from the App server.
Type = TBD2: Raw measurements of packets/bytes to/from the
App Server address.
Type = TBD3: Capacity value sub-TLV
Type = TBD4: Site preference value sub-TLV
8. References
8.1. Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997.
[RFC4364] E. rosen, Y. Rekhter, "BGP/MPLS IP Virtual Private
networks (VPNs)", Feb 2006.
[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>.
[RFC8200] s. Deering R. Hinden, "Internet Protocol, Version 6
(IPv6) Specification", July 2017
Dunbar, et al. Expires January 9, 2023 [Page 14]
Internet-Draft AppMetaData for 5G EC Service
8.2. Informative References
[3GPP-EdgeComputing] 3GPP TR 23.748, "3rd Generation
Partnership Project; Technical Specification Group
Services and System Aspects; Study on enhancement of
support for Edge Computing in 5G Core network
(5GC)", Release 17 work in progress, Aug 2020.
[5G-EC-Metrics] L. Dunbar, H. Song, J. Kaippallimalil, "IP
Layer Metrics for 5G Edge Computing Service", draft-
dunbar-ippm-5g-edge-compute-ip-layer-metrics-00,
work-in-progress, Oct 2020.
[5G-Edge-Sticky] L. Dunbar, J. Kaippallimalil, "IPv6 Solution
for 5G Edge Computing Sticky Service", draft-dunbar-
6man-5g-ec-sticky-service-00, work-in-progress, Oct
2020.
[RFC5521] P. Mohapatra, E. Rosen, "The BGP Encapsulation
Subsequent Address Family Identifier (SAFI) and the
BGP Tunnel Encapsulation Attribute", April 2009.
[BGP-SDWAN-Port] L. Dunbar, H. Wang, W. Hao, "BGP Extension
for SDWAN Overlay Networks", draft-dunbar-idr-bgp-
sdwan-overlay-ext-03, work-in-progress, Nov 2018.
[SDWAN-EDGE-Discovery] L. Dunbar, S. Hares, R. Raszuk, K.
Majumdar, "BGP UPDATE for SDWAN Edge Discovery",
draft-dunbar-idr-sdwan-edge-discovery-00, work-in-
progress, July 2020.
[Tunnel-Encap] E. Rosen, et al "The BGP Tunnel Encapsulation
Attribute", draft-ietf-idr-tunnel-encaps-10, Aug
2018.
9. Acknowledgments
Acknowledgements to Sue Hares and Donald Eastlake for their
review and contributions.
Dunbar, et al. Expires January 9, 2023 [Page 15]
Internet-Draft AppMetaData for 5G EC Service
This document was prepared using 2-Word-v2.0.template.dot.
Authors' Addresses
Linda Dunbar
Futurewei
Email: ldunbar@futurewei.com
Kausik Majumdar
Microsoft
Email: kmajumdar@microsoft.com
Haibo Wang
Huawei
Email: rainsword.wang@huawei.com
Gyan Mishra
Verizon
Email: gyan.s.mishra@verizon.com
Dunbar, et al. Expires January 9, 2023 [Page 16]
