BGPDNS: Using BGP Topology for Scalable Multi-homing in DNS

BGPDNS: Using BGP Topology for Scalable Multi-homing in DNS

BGPDNS is a protocol that uses BGP topology information for DNS resource record sorting, allowing for scalable multi-homing without the burden of traditional AS-based multi-homing.

  • Uploaded on | 6 Views
  • leopoldo leopoldo

About BGPDNS: Using BGP Topology for Scalable Multi-homing in DNS

PowerPoint presentation about 'BGPDNS: Using BGP Topology for Scalable Multi-homing in DNS'. This presentation describes the topic on BGPDNS is a protocol that uses BGP topology information for DNS resource record sorting, allowing for scalable multi-homing without the burden of traditional AS-based multi-homing.. The key topics included in this slideshow are BGPDNS, BGP topology, DNS RR sorting, scalable multi-homing, AS-less multi-homing,. Download this presentation absolutely free.

Presentation Transcript

1. BGPDNS Using BGP topology information for DNS RR sorting a scalable way of multi-homing Andr Oppermann Claudio Jeker SWINOG-4 Meeting Berne, 17. April 2002

2. What is BGPDNS about? BGPDNS is a concept and a protocol for doing AS- less and PI IP-less server multi-homing U s e the topology information contained in the global BGP routing table to sort the multiple DNS resource records because Traditional AS-based server multi-homing is a burden to the global Internet routing system E xcessive consumption of AS number and non- aggregat ed prefix es

3. The demand In many cases it is desired by the customers to connect (important) servers to more than one upstream ISP. Reasons include: Acquire redundancy in case (the link to) one upstream ISP fails Balance/share load over more than one upstream ISP Become independent from individual ISPs

4. Todays solution Today these objectives have to be satisfied by : requesting PI IP space obtaining an AS number participat e in the global BGP routing

5. Shortcomings of today s solution Whilst some advantages, this approach has several drawbacks to the Internet at large and to the newly multi-homed customer : Large fragmentation of IP address space (bad) E xcessive memory and computing power requirements in the Internet core routers (good for Cisco and Juniper) E xhaustion of current AS number space requires upgrade with transition to 32 bit AS numbers (very bad)

6. Shortcomings to customers Running and tuning BGP requires significant knowledge and experience as well as continued monitoring and adjustments BGP without knowledgeable tuning quickly leads to unintended asymmetric traffic patterns through the upstream ISPs Unqualified modification on BGP router quickly leads to disconnection from the global Internet because of missing route announcements or route flap dampening Misconfiguration of the routing table entries quickly lead to bogus route announcements ( like a full /8 or multiple /24 instead of an aggregate ) and can cause serious traffic interruptions ( hello Teleglobe Europe! ) ISPs have to employ very strict filters towards their multi-homed customers because of these frequent problems These filters in turn decrease flexibility and increase complexity while representing a significant source of errors in themselves.

7. Summary of shortcomings Many times the requestors of non-aggregated PI IP address space and AS numbers are not aware of these implications and lack sufficient technological background knowledge to qualify and quantify the impact on themselves and the Internet in general Many times only one or a subset of one of the reasons for AS based multi-homing is given by the requestor Unfortunatly in these cases the cure of AS based multi-homing is often worse than the disease of being single-homed

8. An alternative is needed

9. Overview of BGPDNS The BGPDNS approach combines the power of BGP with the ease of DNS BGPDNS does BGP but does not require an unique AS number on the customer side BGPDNS does not need PI IP address space and fully maintains aggregates

10. Details of BGPDNS In BGPDNS multi-homing , the server operator has upstream links to more than one ISP From each of these ISPs the operator also receives a reasonable IP prefix out of their aggregates O ne IP address of each prefix of the ISPs is assigned to the multi-homed server The router connecting to all these ISPs does policy routing to direct the outgoing packets to the ISP where the prefix belong s to .

11. Details of BGPDNS

12. Details of BGPDNS All of the IP address es of this server are put into DNS as multiple A records to the same name IN A IN A IN A Instead of round-robin were going to use BGP for ordering!

13. Details of BGPDNS The BGPDNS server establishes a BGP listening- only session with each of the ISPs route-servers to get a comprehensive view of the Internet topology from its own perspective It does normal best-path evaluation either subject to the default rules or custom crafted metrics as in normal AS-based multi-homing

14. Details of BGPDNS

15. Details of BGPDNS User types into her browser (her IP address is Authorative DNS server at BGPDNS site receives request for DNS server finds multiple IN A records for DNS server has to find out which IN A is best reachable for this user Ask BGPDNS server!

16. Details of BGPDNS

17. Details of BGPDNS BGPDNS server receives request from DNS server containing: source IP of DNS request ( list of possible answers (, and BGPDNS server looks up the best path to the requestor via the BGP topology information show ip bgp

18. Details of BGPDNS sh ow ip bgp 19 BGP routing table entry for 19 2 .0. 2 . 0 /2 1 Paths: (3 available, best #2 , table Default-IP-Routing-Table) 100 8235 1836 286 3333 10.1.1 .18 from Origin IGP, localpref 200, valid, external 300 1836 286 3333 from Origin IGP, metric 10, localpref 200, valid, external, best 200 9177 8210 3333 from Origin IGP, metric 10, localpref 100, valid, external

19. Details of BGPDNS BGPDNS takes leftmost AS number of best path (which is the active path from our point of view) BGPDNS takes a list of all prefixes of the upstreams show ip bgp regexp ^leftmost-as$ BGPDNS loops over the IN A records to find the local IP that is within one of these upstream prefixes and assigns the highest weight to it It has to be because all the IN A IPs are form our upstreams If not, normal round-robin applies as usual

20. Details of BGPDNS show ip bgp regexp ^ 300 $ BGP table version is 0, local router ID is Status codes: s suppressed, d damped, h history, * valid, > best, i - internal Origin codes: i - IGP, e - EGP, ? - incomplete Network Next Hop Metric LocPrf Weight Path *> /2 0 10 200 0 300 i 300 1836 286 3333 from Origin IGP, metric 10, localpref 200, valid, external, best IP address Preference 0 0 1000

21. Details of BGPDNS BGPDNS returns the packet to DNS server with best path IP set to highest preference DNS server sorts response with highest preference first and answers to the user User uses the first IP address of the DNS response to establish a connection with the target server

22. Details of BGPDNS

23. Details of BGPDNS

24. Compar i son to normal BGP Responsiveness to link state or topology changes is immediate for new requests If a link fails, the corresponding IP wont be chosen for new requests anymore because all BGP NLRI information is gone, hence no more best path for this upstream In the case of a upstream link failure , the maximum black hole time for a particular requestor is the configured DNS resource record expiration timeout Affects o nly requestors who had this particular upstream link as best path The DNS RR expiration timeout has to be chosen carefully !

25. Convergence timers BGP today has a global convergence and propagation time of approx. 3 minutes as shown by recent research Determining the optimal DNS RR expiration timeout is balancing between two opposite tradeoffs: If the expiry timeout is low this will add latency during a session leading to poor responsiveness experience by the user because of multiple successive DNS requests if the expiration timeout is high this may reuse cached values with now sub-optimal path information or, if the link of the preferred IP has gone down in the meantime, to partial unreachability until the cached DNS RR expires We recommend DNS RR expiry timeout s between 20 seconds and 2 hours

26. Advantages of BGPDNS Link and ISP redundancy Load balancing over more than one link and ISP Independence of a single ISP Connection path symmetry between the server and the client (because of prefix based routing) No impact on global BGP routing system

27. Disadvantages of BGPDNS Each server requires as many IP addresses as it is connected to ISPs Which is relative, AS-based multi-homing needs at least a /24, here in this example with three ISPs I can get away with 3x /27 or so For cached DNS resource records ; only timeout - based convergence Additional load on the DNS system Which doesnt seem to be a problem (Thanks Akamai!)

28. Rationale for BGPDNS IP Address space depletion is not as fast and does not have the same impact as AS number space depletion and Internet core router memory consumption With next generation IP numbering ( IPv6 ?) address space depletion is no longer a issue Content delivery networks like Akamai have proven that DNS RR based global load balancing is working on a large scale and does not have a negative impact on the Internet at large nor on the individual user

29. Rationale for BGPDNS For all services that are either stateless, have only short-lived or restartable sessions BGPDNS is well suited and provides equal results as true AS-based multi-homing applies to HTTP, HTTPS, SMTP, POP, IMAP, FTP (in part), NNTP (client sessions)... By keeping IP address space aggregation intact and the positive effects on AS numbers and router memory by far outweigh the negative effects of BGPDNS by requiring one IP address per server and upstream ISP

30. The BGPDNS protocol The BGPDNS protocol is spoken between the DNS server and the BGP listener The communication is stateless and uses the UDP datagrams for communication Because of its close relationship to the BGP the port number 179/UDP is chosen for the BGPDNS task on the BGP listener

31. The BGPDNS protocol

32. BGPDNS reference implementation DJBDNS tinydns as DNS server Has nice non-threaded internal design Its impossible to follow bind9 code without suffering serious brain damange Zebra bgpd as BGPDNS server More or less structured internal design Any other recent and stable OpenSource BGP daemon? Available on

33. Important in implementation Response time is critical The BGPDNS must do fast lookups to not delay the DNS RTT too much Lookup times must be far below 1 second Timeouts If the DNS server receives no response within 1 second from the BGPDNS server it will send out random order

34. BGPDNS performance cvs> sh dnssort statistics DNSsort statistics ------------------ Received packets: 139909174 sent packets: 138057953 dropped packets: 0 currently queued: 0 / 128 DNSsort cache table for view 0 ---------------------------------- AS 5378 : 697113 hits AS 15623: 377189 hits AS 8237 : 11598 hits AS 15517: 77078 hits AS 8220 : 4415231 hits AS 6667 : 0 hits AS 15600: 132782 hits AS 6893 : 475114 hits AS 13646: 53162 hits AS 1836 : 8518605 hits AS 9044 : 515931 hits AS 6776 : 66926 hits AS 12350: 394278 hits AS 12520: 148551 hits AS 8758 : 78749 hits AS 20940: 0 hits AS 9177 : 97422320 hits AS 8327 : 130593 hits AS 8235 : 5372 hits AS 12429: 607428 hits AS 8833 : 579531 hits AS 13250: 51576 hits AS 13030: 53294 hits AS 15667: 161435 hits AS 8271 : 39818 hits 145 hits/second average 2.2% CPU on a PIII-550/FreeBSD4.4

35. Failure cases BGPDNS server does not respond If the DNS server does not receive answers for some time it should mark the BGPDNS server as defektive and answer with random sorting Loss of all BGP sessions If the BGPDNS looses all BGP session it will simply answer all requests without any preference set Network not in table If the network prefix of the end-user is not in the BGP routing-table it will simply answer without any preference set

36. Security considerations Authorization of BGPDNS requestors MD5 shared secret (like in OSPF) D o S /O verlo ading attacks Cant be done much, BGPDNS should not fall over but provide some form of overload protection Spoofing of requests/answers MD5 shared secret Filter 179/UDP on border router / firewall Information leakage The DNS information is public anyway

37. Recommendation More scrunity for AS number requests In case of services that are stateless, have only short-lived or restartable sessions A BGPDNS policy like the RIPE HTTP policy or Static Dial Up policy should be applied And BGPDNS should be considered as an alternative then

38. Other Projects www.SuperSparrow.ORG We found out about it quite some time after the initial idea during the research phase for the BGPDNS project SuperSparrow has serious scalability issues; it uses t elnet s to the routers to access the BGP path information The last release is 0.0.0 and over a year old from 9th January 2001

39. The authors Andr Oppermann, Idea and concept Claudio Jeker, Implementation Other projects by the authors q mail-ldap

40. Questions and comments? Yes, you can grill us now! Youll find a RFC draft and patches on: