Traffic Engineering for CDNs by Matt Jansen [APRICOT 2015]
1 like716 views
The document discusses how content delivery networks (CDNs) use DNS-based traffic engineering and BGP routing interactions between CDNs and internet service providers (ISPs). It provides examples of how ISPs attempting to influence CDN traffic routing through BGP techniques like more specific route announcements or filtering can backfire and result in traffic blackholing or long term revenue loss for the ISP or upstream transit providers. The document recommends communication between CDNs and ISPs to achieve traffic engineering goals through techniques like embedded clusters instead of assumptions about routing behavior.
![©2012 AKAMAI | FASTER FORWARDTM
Example of Akamai mapping
• Notice the different A records for different locations:
[NYC]% host www.symantec.com
www.symantec.com CNAME e5211.b.akamaiedge.net.
e5211.b.akamaiedge.net. A 207.40.194.46
e5211.b.akamaiedge.net. A 207.40.194.49
[Boston]% host www.symantec.com
www.symantec.com CNAME e5211.b.akamaiedge.net.
e5211.b.akamaiedge.net. A 81.23.243.152
e5211.b.akamaiedge.net. A 81.23.243.145
DNS based Mapping Example](https://image.slidesharecdn.com/04-akamai-mj-traffic-engineering-apricot1425633293-150308182843-conversion-gate01/85/Traffic-Engineering-for-CDNs-by-Matt-Jansen-APRICOT-2015-5-320.jpg)
Traffic Engineering for CDNs by Matt Jansen [APRICOT 2015]
- 1. Traffic Engineering for CDNs Matt Jansen Akamai Technologies APRICOT 2015
- 2. ©2012 AKAMAI | FASTER FORWARDTM The world’s largest on-demand, distributed computing platform delivers all forms of web content and applications The Akamai Intelligent Platform Typical daily traffic: • More than 2 trillion requests served • Delivering over 25 Terabits/second • 15-30% of all daily web traffic The Akamai Intelligent Platform: 170,000+ Servers 2,000+ Locations 102+ Countries 1,300+ Networks 700+ Cities
- 3. ©2012 AKAMAI | FASTER FORWARDTM When content is requested from CDNs, the user is directed to the optimal server to serve this user There’s 2 common ways to do that: • anycast: the content is served from the location the request is received (easy to build, requires symmetric routing to work well) • DNS based: the CDN decides where to best serve the content from based on the resolver it receives the request from, and replies with the optimal server How CDNs Work
- 4. ©2012 AKAMAI | FASTER FORWARDTM Users querying a DNS-based CDNs will be returned different A (and AAAA) records for the same hostname depending on the resolver the request comes from This is called “mapping” The better the mapping, the better the CDN How DNS based CDNs Work
- 5. ©2012 AKAMAI | FASTER FORWARDTM Example of Akamai mapping • Notice the different A records for different locations: [NYC]% host www.symantec.com www.symantec.com CNAME e5211.b.akamaiedge.net. e5211.b.akamaiedge.net. A 207.40.194.46 e5211.b.akamaiedge.net. A 207.40.194.49 [Boston]% host www.symantec.com www.symantec.com CNAME e5211.b.akamaiedge.net. e5211.b.akamaiedge.net. A 81.23.243.152 e5211.b.akamaiedge.net. A 81.23.243.145 DNS based Mapping Example
- 6. ©2012 AKAMAI | FASTER FORWARDTM DNS based Mapping 1) end-user requests www.example.com from ISP NS 2) ISP NS recursively (multiple iterations) looks up www.example.com being referred to authoritative CDN NS (by CNAME) 3) ISP NS asks authoritative CDN NS 4) CDN NS looks up IP of requestor (ISP NS) and replies with IP of optimal cluster to serve content (closest cluster for that ISP) 5) ISP NS replies to end-user who 6) requests content from local Cluster end-user ISP NS 1.2.3.4 root/tld/intermediate NS (recursive lookup until reaching authoritative NS) CDN NS Closest cluster 5.6.7.8 1 3 6 2 NS 1.2.3.4? best cluster = 5.6.7.8 4 5
- 7. ©2012 AKAMAI | FASTER FORWARDTM Types of CDN Deployments (simplified) end-user ISP Internet IX Transit Embedded cluster in ISP Colocated Cluster with IX Infrastructure Cluster with Transit Backend Mapping system
- 8. ©2012 AKAMAI | FASTER FORWARDTM DNS based CDNs don’t necessarily have a backbone Those clusters can be standalone ‘Islands’ with no connectivity between them! DNS based CDNs usually don’t announce many prefixes as they will only include the local servers • it is not uncommon to see a single /24 from Akamai at an IX • you will receive a different set of prefixes on each peering This does not mean you will not see a lot of traffic • how many servers do you need to serve 10g nowadays? Typical DNS based Topology
- 9. ©2012 AKAMAI | FASTER FORWARDTM Transit Peer • Akamai does not have a backbone, each IX instance is independent • Cluster uses transit to fetch content origin • Content is served to peers over the IX • BGP session serves 2 purposes: • Route traffic strictly within the local instance • Tell our system which prefixes this cluster is allowed to serve • New prefixes being picked up by the system can take up to 24hrs Origin Server IX Content Typical large IX deployment Origin
- 10. Scenario 1: prefix withdrawal
- 11. ©2012 AKAMAI | FASTER FORWARDTM Attempting to shift traffic to another path ISP peers with CDN over 2 IXs • wants to shift traffic from A to B due to reduce backbone traffic • Typical BGP based techniques: 1) AS-Path prepending 2) MEDs 3) more/less specific announcements None of the above have the desired effect! • no network between clusters, any BGP parameters are only locally relevant • Might have undesired effect of shifting traffic to a upstream/competitor on the same IX end-user ISP A IX A IX B
- 12. ©2012 AKAMAI | FASTER FORWARDTM Problem solved… end-user IX B ISP withdraws some of their prefixes over IX A • Traffic falls back to transit on the same cluster (provided that still has the best performance) • this might result in them receiving the traffic in another location in their network so they’re happy Transit ISP A IX A
- 13. ©2012 AKAMAI | FASTER FORWARDTM …but not for long within 24hrs • The CDN backend system processes the fact that they don’t receive their prefixes at the IX A cluster • Traffic switches to another cluster where they do receive them (this might be one of their upstreams they have a PNI with in the same city) • Traffic comes back into their network again at the same router! Upstream of ISP A end-user Transit ISP A IX A
- 14. ©2012 AKAMAI | FASTER FORWARDTM Issues • BGP parameters only locally relevant • Delayed effect of BGP announcements on Mapping System • CDNs typically see your prefixes in many different locations
- 15. ©2012 AKAMAI | FASTER FORWARDTM Better solution • Talk to the CDN if you have issues with their traffic in a specific location • You can work together to achieve the result you’re looking for • Get a local embedded cluster
- 16. Scenario 2: more specific Route Announcement
- 17. ©2012 AKAMAI | FASTER FORWARDTM • ISP A is multi-homed to Transit Providers AS2002 and AS3003 • Transit Provider AS2002 peers with CDN • Transit Provider AS3003 does not peer with CDN • CDN sends traffic to ISP A via Transit Provider AS2002 ISP A AS1001 CDN AS20940 Transit Provider AS2002 Transit Provider AS3003 100.100.96.0/20 100.100.96.0/20 0.0.0.0/0 Transit Provider AS4003 Consistent Announcements IX
- 18. ©2012 AKAMAI | FASTER FORWARDTM • ISP A would like to balance traffic between the two upstream providers • ISP A prepends, then applies MED to Transit Provider AS2002. This has no effect on CDN traffic. • Eventually, ISP A de-aggregates the /20 and advertises more specific routes to Transit Provider AS3003 • What will happen? Loadbalancing
- 19. ©2012 AKAMAI | FASTER FORWARDTM • ISP A announces more specific routes to Transit Provider AS3003 • Transit Provider AS3003 announces new /24 to AS2002 • CDN IX router does not have a full-table, so traffic continue route to the /20 of AS2002 • ISP A is happy with the balanced traffic on dual Transit Providers ISP A AS1001 CDN AS20940 Transit Provider AS2002 Transit Provider AS3003 100.100.96.0/20 AS2002 Routing Table 100.100.100.0/24 AS3003 AS1001 100.100.99.0/24 AS3003 AS1001 100.100.96.0/20 AS1001 CDN AS20940 Routing Table 100.100.96.0/20 AS2002 AS1001 0.0.0.0/0 AS4003 100.100.96.0/20 0.0.0.0/0 Transit Provider AS4003 Loadbalancing works... IX Peering
- 20. ©2012 AKAMAI | FASTER FORWARDTM • Lost of revenue for Transit Provider AS2002 even though their peering/backbone is utilized • What happens if AS2002 does not like the traffic from one peer to the other? …but
- 21. ©2012 AKAMAI | FASTER FORWARDTM Transit provider filters traffic • In order to get rid of traffic between peers, Transit Provider AS2002 implements an ACL on IX port facing AS3003 • Traffic gets blackholed, ISP A’s eyeballs don’t receive traffic anymore! ISP A AS1001 CDN AS20940 Transit Provider AS2002 Transit Provider AS3003 100.100.96.0/20 100.100.96.0/20 ACL 0.0.0.0/0 Transit Provider AS4003 hostname AS2002-R1 ! interface TenGigabitEthernet1/1 ip access-group 101 out ! access-list 101 deny ip any 100.100.100.0 0.0.0.255 access-list 101 deny ip any 100.100.99.0 0.0.0.255 access-list 101 permit ip any any IX Peering
- 22. ©2012 AKAMAI | FASTER FORWARDTM • CDN observes ISP A end-users are unable to access some websites • CDN stops serving unreliable prefixes received from Transit Provider AS2002, traffic shifts from IX to Transit Provider AS4003 • ISP A can access all websites happily • Transit Provider AS2002 loses revenue ISP A AS1001 CDN AS20940 Transit Provider AS2002 Transit Provider AS3003 100.100.96.0/20 100.100.96.0/20 ACL 0.0.0.0/0 Transit Provider AS4003 Unintended Result IX Peering
- 23. ©2012 AKAMAI | FASTER FORWARDTM Issues • Don’t assume a full-table on any device on the internet • Filtering traffic results in: • short term traffic blackholing! • long term widthdrawal of traffic resulting in revenue loss
- 24. ©2012 AKAMAI | FASTER FORWARDTM • AS2002 filters the specific prefixes instead of the actual traffic • work with upstreams and/or CDN for finetuning • Get Transit Provider AS3003 to peer with CDN directly ;) • Get a local embedded cluster ISP A AS1001 CDN AS20940 Transit Provider AS2002 Transit Provider AS3003 100.100.96.0/20 100.100.96.0/20 Filter Specific route 100.100.99.0/24 100.100.100.0/24 0.0.0.0/0 Transit Provider AS4003 neighbor PEER-GROUP prefix-list DENY-SPECIFIC in ! ip prefix-list DENY-SPECIFIC seq 5 deny 100.100.100.0/24 ip prefix-list DENY-SPECIFIC seq 10 deny 100.100.99.0/24 ip prefix-list DENY-SPECIFIC seq 100 permit 0.0.0.0/0 le 32 Better solutions IX Peering
- 25. ©2012 AKAMAI | FASTER FORWARDTM • ISP A is single homed to Transit Provider AS2002 • ISP A obtains a /24 from Transit Provider AS2002’s address space • all works well ISP A AS1001 CDN AS20940 Transit Provider AS2002 100.100.97.0/24100.100.96.0/20 100.100.96.0/20 0.0.0.0/0 Transit Provider AS4003 100.100.97.0/24 CDN AS20940 Routing Table 100.100.96.0/20 AS2002 100.100.97.0/24 AS2002 AS1001 0.0.0.0/0 AS4003 100.100.97.0/24 Another variation of this Scenario IX
- 26. ©2012 AKAMAI | FASTER FORWARDTM • ISP A moves to new Transit Provider AS3003, but keeps using his previously assigned prefix 100.100.96.0/24 • CDN keeps serving traffic to ISP A via Transit Provider AS2002 due to the /20 being received there ISP A AS1001 CDN AS20940 Transit Provider AS2002 100.100.97.0/24 100.100.96.0/20 100.100.96.0/20 0.0.0.0/0 Transit Provider AS4003 100.100.97.0/24 CDN AS20940 Routing Table 100.100.96.0/20 AS2002 0.0.0.0/0 AS4003 Transit Provider AS3003 Provider Change Peering IX
- 27. ©2012 AKAMAI | FASTER FORWARDTM • Lost of revenue for Transit Provider AS2002 even though their peering/backbone is utilized • What happens if AS2002 does not like the traffic from one peer to the other? …but
- 28. ©2012 AKAMAI | FASTER FORWARDTM • In order to get rid of traffic between peers, Transit Provider AS2002 implements an ACL on IX port facing AS3003 • Traffic gets blackholed, ISP A’s eyeballs don’t receive traffic anymore! ISP A AS1001 CDN AS20940 Transit Provider AS2002 100.100.97.0/24 100.100.96.0/20 100.100.96.0/20 0.0.0.0/0 Transit Provider AS4003 100.100.97.0/24 CDN AS20940 Routing Table 100.100.96.0/20 AS2002 0.0.0.0/0 AS4003 Transit Provider AS3003 ACL hostname AS2002-R1 ! interface TenGigabitEthernet1/1 ip access-group 101 out ! access-list 101 deny ip any 100.100.97.0 0.0.0.255 access-list 101 permit ip any any Transit provider filters traffic IX Peering
- 29. ©2012 AKAMAI | FASTER FORWARDTM • CDN observes ISP A end-users are unable to access some websites • CDN stops serving unreliable prefixes received from Transit Provider AS2002, traffic shifts from IX to Transit Provider AS4003 • ISP A can access all websites happily • Transit Provider AS2002 loses revenue ISP A AS1001 CDN AS20940 Transit Provider AS2002 100.100.97.0/24 100.100.96.0/20 100.100.96.0/20 0.0.0.0/0 Transit Provider AS4003 100.100.97.0/24 CDN AS20940 Routing Table 100.100.96.0/20 AS2002 0.0.0.0/0 AS4003 Transit Provider AS3003 ACL hostname AS2002-R1 ! interface TenGigabitEthernet1/1 ip access-group 101 out ! access-list 101 deny ip any 100.100.97.0 0.0.0.255 access-list 101 permit ip any any Unintended Result Peering IX
- 30. ©2012 AKAMAI | FASTER FORWARDTM Issues • Don’t assume a full-table on any device on the internet • If you do announce a prefix others expect you to be able to serve traffic to all of it • Don’t allow customers to use your PA space as PI • Filtering traffic results in: • short term traffic blackholing! • long term widthdrawal of traffic resulting in revenue loss
- 31. ©2012 AKAMAI | FASTER FORWARDTM • Deaggregate the /20 if you can’t/won’t serve all of it • Only announce address space you will serve traffic to • Get a local embedded cluster ISP A AS1001 CDN AS20940 Transit Provider AS2002 100.100.97.0/24 100.100.96.0/24 0.0.0.0/0 Transit Provider AS4003 100.100.97.0/24 CDN AS20940 Routing Table 100.100.96.0/24 AS2002 100.100.98.0/23 AS2002 100.100.100.0/22 AS2002 100.100.104.0/21 AS2002 0.0.0.0/0 AS4003 Transit Provider AS3003 100.100.98.0/23 100.100.100.0/22 100.100.104.0/21 Better solutions IX Peering
- 32. Scenario 3: Split Route Announcement
- 33. ©2012 AKAMAI | FASTER FORWARDTM • ISP A is multi-homed to Transit Providers AS2002 and AS3003 • Transit Provider AS2002 peers with CDN • Transit Provider AS3003 does not peer with CDN • ISP A announces different prefix to different ISP • ISP A can access full internet ISP A AS1001 CDN AS20940 Transit Provider AS2002 Transit Provider AS3003 100.100.96.0/20 100.100.96.0/22 100.100.100.0/22 0.0.0.0/0 Transit Provider AS4003 CDN AS20940 Routing Table 100.100.96.0/22 AS2002 AS1001 100.100.100.0/22 AS2002 AS1001 0.0.0.0/0 AS4003 Split announcement IX
- 34. ©2012 AKAMAI | FASTER FORWARDTM • End Users are using IP Addresses 100.100.96.0/22, 100.100.100.0/22, 100.100.104.0/22, 100.100.108.0/22 • End Users are using ISP A NS 100.100.100.100 • CDN receives the NS Prefix 100.100.100.0/22 from AS2002 and maps the traffic for ISP A to this cluster • 100.100.96.0/22 100.100.100.0/22 traffic is routed via AS2002 while 100.100.104.0/22 100.100.108.0/22 traffic falls back to default route via AS4003, AS3003 ISP A AS1001 CDN AS20940 Transit Provider AS2002 Transit Provider AS3003 100.100.96.0/20 0.0.0.0/0 Transit Provider AS4003 ISP A AS1001 End User IP: 100.100.96.0/24 End User IP: 100.100.108.0/24 DNS: 100.100.100.100 100.100.96.0/22 100.100.100.0/22 Split announcement IX
- 35. ©2012 AKAMAI | FASTER FORWARDTM • This can work perfectly fine • But the path via the transit providers AS4003 & AS3003 might not be as good as the direct peering, 100.100.100.108.0/22 end users could have significantly worse performance • What will ISP A do if the user complain? Differing performance
- 36. ©2012 AKAMAI | FASTER FORWARDTM • ISP A swaps the route announcements • Both 100.100.96.0/22 and 100.100.108.0/22 are routed via AS2002 and end-users have the same performance • The end-user is happy and closes the ticket ISP A AS1001 CDN AS20940 Transit Provider AS2002 Transit Provider AS3003 100.100.96.0/20 0.0.0.0/0 Transit Provider AS4003 ISP A AS1001 End User IP: 100.100.96.0/24 End User IP: 100.100.108.0/24 DNS: 100.100.100.100 100.100.96.0/22 100.100.108.0/22 Problem solved… IX
- 37. ©2012 AKAMAI | FASTER FORWARDTM 24hrs later: • CDN no longer receives NS prefix 100.100.100.0/22 from AS2002 • CDN maps the traffic of ISP A to Cluster B (where they see AS3003’s prefixes) instead of Cluster A (which only peers with AS2002) • ISP A will receive the traffic from a completely different source potentially all via AS3003 now negating all the TE efforts DO NOT split nameserver and end-user prefixes when traffic engineering …but
- 38. Scenario 4: (Attempted) Content Filtering
- 39. ©2012 AKAMAI | FASTER FORWARDTM ISPs do receive requests from Government organizations to filter specific content The default action is often to block a specific source IP If this content is hosted by a CDN this will not do what you expect! Content filtering what it WILL DO: what it will NOT: blackhole random content to your end-users filter any specific content get your cluster suspended because of observed loss and all traffic served from upstream
- 40. ©2012 AKAMAI | FASTER FORWARDTM • Standard BGP traffic engineering will not have the expected results • Changes in announcements may have a delayed effect • Where mapping is based on NS, splitting nameserver and end-user prefixes over different providers will have unexpected effects • Not all clusters have a full table • splitting more specific announcements over different links can cause unintended behavior • Announcing prefixes with holes results in blackholing traffic • Talk to your CDN partners for finetuning traffic • DO NOT filter traffic by IP Summary
- 41. ©2012 AKAMAI | FASTER FORWARDTM Matt Jansen mj@akamai.com as20940.peeringdb.com Questions?