Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
The BBC’sExperience ofpreparing for the2012 LondonOlympicsFor Velocity London2012            Andy “Bob” Brockhurst        ...
Introduction•   The Team•   LAMP (without the M)•   Tomcat Java Service Layer•   Custom apache modules•   Varnish with ext...
How the BBC works•     One domain[1]•     Two technology stacks[2]•     Cert’s and SSL•     ProxyPass’•     Apps are a TLD...
Network Topology• Dual DC[1]• No DC affinity[2][1] One more soon(ish)[2] Well a couple of apps do[3][3] We dont talk about...
Network Topology
Traffic Routing•   TM -> PAL•   TM -> Varnish -> TM -> PAL•   TM -> Service Layer•   TM -> Varnish -> TM -> Service Layer
Traffic Routing    Requestiplayer |  everything elsesport |      v     TM      .-> TM .--> TM .--> TM     | / | /         ...
Environments•   Integration•   Test•   Staging•   Live•   Journalism
Right let’s do some testing
Why?• Too much change  – Network Architecture  – Server Configurations  – Load balancers  – Peering points• High Profile• ...
Gaining Confidence• Load testing on Stage  – Tests individual applications  – Single endpoints only  – No concurrent load•...
Other objectives• Maintain BAU• Handle failure gracefully• Deliver Expectation
“What the Abdication did for Radioand the Coronation did for Television,London 2012 will do for Online.”
Current Volumetrics• Big numbers for sport  – 9M users/day  – 90M views/day• Punishing peaks  – Saturday football final sc...
Expected Volumetrics• Expected peaks  – 1.5M concurrent users  – 60k different sports pages    • 2,500 per minute  – 30% v...
Timeline• March 2012 (T minus 5 months)  – Team members assigned  – Resilience testing  – Performance testing• Testing wit...
Olympics Run-up•   Jubilee (2nd June)•   Euros (8th June)•   Wimbledon (25th June)•   Formula One
Cloud Testing• International testing• Detailed test results
Cloud Testing•   First performance test breaks live•   Exposed monitoring issues•   Couldn’t internally diagnose•   Lots o...
Early Findings•   Stop tests•   Monitoring•   UK Data centre capacity•   UK Data centre network segments
(Not) Caching kills• Conditional modules• Non-Olympics related modules  – Commenting / Favourites• Lowers cachability• Tes...
What is a failure?•   Error 500?•   Blank pages?•   Stale content?•   Slow pages?•   Burning data centres?
Resilience Testing• Kill backends• Traffic Manager  – Screw with headers  – Screw with status (418 anyone)  – Truncate bod...
Early findings• Failure mode testing  – Everything is a SPOF  – Performance sucks in a failure
Specific findings•   Monitoring Thresholds•   Verbose logging, everywhere•   Timeouts•   No data•   Volumetrics•   Unfair ...
Verbose Logging•   Wrong levels configured•   Diagnostic information•   Expected/Handled errors•   Too much detail•   Hurt...
Not enough logging•   Fatals with no logging•   Unhandled conditions•   Monitoring holes•   Operations staff blind
Platform Configuration• Unfair load-balancing  – Remove older commodity servers• Competitive service applications  – Re-ho...
“Timeouts at lower levels in the architecture MUST be set shorter than the timeouts configured at higher levels of the arc...
Timeouts• Frontend/Backend timeouts  – Frontends with lower timeouts  – Caches never populated• Alter backends to return e...
More timeouts• Unspecified timeouts• Wrongly specified timeout units  – ms/sec
Poor Application Performance• Multiple synchronous content requests• International cachability• Missing negative caching  ...
Testing frequency• Every two weeks• Every week• Every other day
One week before…  The opening ceremony…• 1st successful test on Live  – with no errors at all.
Performance Overview• Did find problems    – Weren’t found on stage•   In all architecture layers•   Components believed t...
Resilience Overview•   Teams never tested failure scenarios•   Assumed that services didn’t fail•   Inconsistent use of fl...
Other problems• Running a “fake” Olympics    – That is invisible to the public    – Did consider publicising a test•   No ...
Other problems• RCA complicated by shared platform• Testing stopped by BAU/TX• High reliance on key staff  – Some tests su...
Working with external tester• Workflow testing differed  – User journeys  – Direct linking to hotspots• Very responsive to...
Did it work• YES – Found and fixed issues – Before they bit us – On production – With little impact on BAU
Recommendations•   Increase stage capacity•   Intelligent load balancing•   Test NFRs in Development•   Caching, caching a...
Some Statistics
Daily Reach (M)
Streaming Views (M) Wed 1st Aug
Unique Browsers (M)
Thanks for listening•   Thanks to flickr users:     –   dgjones           •     Office Dalek, London, 14-10-06           •...
Special Thanksto:      – David Holroyd        • Technical Architect BBC Sport (Olympics)      – Matt Clark        • Senior...
Thanks for listening• This presentation:  – TBC• Me:  – Andy “Bob” Brockhurst  – Twitter: b3cft (and pretty much anywhere ...
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Velocity london 2012 bbc olympics
Upcoming SlideShare
Loading in …5
×

Velocity london 2012 bbc olympics

7,418 views

Published on

Talk for Velocity London 2012

Published in: Technology
  • Be the first to comment

Velocity london 2012 bbc olympics

  1. The BBC’sExperience ofpreparing for the2012 LondonOlympicsFor Velocity London2012 Andy “Bob” Brockhurst Principal Engineer BBC Platforms/Frameworks
  2. Introduction• The Team• LAMP (without the M)• Tomcat Java Service Layer• Custom apache modules• Varnish with extensions• ZendFramework.. – ...customised a.k.a PAL• Barlesque
  3. How the BBC works• One domain[1]• Two technology stacks[2]• Cert’s and SSL• ProxyPass’• Apps are a TLD• 360+ apps• Everyone shares everything[3][1] Okay there are several but they are all really the same one.[2] Okay, three if you are going to be picky.[3] Yes really, everything!
  4. Network Topology• Dual DC[1]• No DC affinity[2][1] One more soon(ish)[2] Well a couple of apps do[3][3] We dont talk about them
  5. Network Topology
  6. Traffic Routing• TM -> PAL• TM -> Varnish -> TM -> PAL• TM -> Service Layer• TM -> Varnish -> TM -> Service Layer
  7. Traffic Routing Requestiplayer | everything elsesport | v TM .-> TM .--> TM .--> TM | / | / | / | | / | / | / | v / v / api v / v Varnish PAL Varnish Dynamite
  8. Environments• Integration• Test• Staging• Live• Journalism
  9. Right let’s do some testing
  10. Why?• Too much change – Network Architecture – Server Configurations – Load balancers – Peering points• High Profile• Gain confidence
  11. Gaining Confidence• Load testing on Stage – Tests individual applications – Single endpoints only – No concurrent load• Real hardware• Real data – As much as possible• Real Journalism
  12. Other objectives• Maintain BAU• Handle failure gracefully• Deliver Expectation
  13. “What the Abdication did for Radioand the Coronation did for Television,London 2012 will do for Online.”
  14. Current Volumetrics• Big numbers for sport – 9M users/day – 90M views/day• Punishing peaks – Saturday football final scores 4000 pv/s – ~750k Concurrent users• Wimbledon – 1700 pv/s
  15. Expected Volumetrics• Expected peaks – 1.5M concurrent users – 60k different sports pages • 2,500 per minute – 30% video via iPlayer
  16. Timeline• March 2012 (T minus 5 months) – Team members assigned – Resilience testing – Performance testing• Testing with External Partner
  17. Olympics Run-up• Jubilee (2nd June)• Euros (8th June)• Wimbledon (25th June)• Formula One
  18. Cloud Testing• International testing• Detailed test results
  19. Cloud Testing• First performance test breaks live• Exposed monitoring issues• Couldn’t internally diagnose• Lots of tail, grep, awk, sed.
  20. Early Findings• Stop tests• Monitoring• UK Data centre capacity• UK Data centre network segments
  21. (Not) Caching kills• Conditional modules• Non-Olympics related modules – Commenting / Favourites• Lowers cachability• Testing an immature product• Subsequent testing exposed more
  22. What is a failure?• Error 500?• Blank pages?• Stale content?• Slow pages?• Burning data centres?
  23. Resilience Testing• Kill backends• Traffic Manager – Screw with headers – Screw with status (418 anyone) – Truncate body• Introduce waits• Limit cache sizes• Reduce network bandwidth
  24. Early findings• Failure mode testing – Everything is a SPOF – Performance sucks in a failure
  25. Specific findings• Monitoring Thresholds• Verbose logging, everywhere• Timeouts• No data• Volumetrics• Unfair load balancing
  26. Verbose Logging• Wrong levels configured• Diagnostic information• Expected/Handled errors• Too much detail• Hurts health/forensic reporting
  27. Not enough logging• Fatals with no logging• Unhandled conditions• Monitoring holes• Operations staff blind
  28. Platform Configuration• Unfair load-balancing – Remove older commodity servers• Competitive service applications – Re-home critical applications
  29. “Timeouts at lower levels in the architecture MUST be set shorter than the timeouts configured at higher levels of the architecture.”
  30. Timeouts• Frontend/Backend timeouts – Frontends with lower timeouts – Caches never populated• Alter backends to return early
  31. More timeouts• Unspecified timeouts• Wrongly specified timeout units – ms/sec
  32. Poor Application Performance• Multiple synchronous content requests• International cachability• Missing negative caching – Bypassed shared caches
  33. Testing frequency• Every two weeks• Every week• Every other day
  34. One week before… The opening ceremony…• 1st successful test on Live – with no errors at all.
  35. Performance Overview• Did find problems – Weren’t found on stage• In all architecture layers• Components believed to be “fine” were not• Stage is not suitable for this level of testing• Proposal for any future “high profile” event• CDNs didn’t really get tested
  36. Resilience Overview• Teams never tested failure scenarios• Assumed that services didn’t fail• Inconsistent use of flagpoles• Reliance on mod_cache stale-on-error
  37. Other problems• Running a “fake” Olympics – That is invisible to the public – Did consider publicising a test• No A/B (bucket) testing capability• Some tests affected BAU• No real test of the HLS HDS streaming• Platform monitoring cycle
  38. Other problems• RCA complicated by shared platform• Testing stopped by BAU/TX• High reliance on key staff – Some tests suffered• No CDN testing – At their request – Places unfair load on infrastructure• Unable to simulate network congestion
  39. Working with external tester• Workflow testing differed – User journeys – Direct linking to hotspots• Very responsive to altering tests• Did add extra complexity
  40. Did it work• YES – Found and fixed issues – Before they bit us – On production – With little impact on BAU
  41. Recommendations• Increase stage capacity• Intelligent load balancing• Test NFRs in Development• Caching, caching and more caching• Kill load tests quickly• Improve internal load testing• Profile frontends under load• Better post analysis tools
  42. Some Statistics
  43. Daily Reach (M)
  44. Streaming Views (M) Wed 1st Aug
  45. Unique Browsers (M)
  46. Thanks for listening• Thanks to flickr users: – dgjones • Office Dalek, London, 14-10-06 • http://www.flickr.com/photos/dgjones/284592369 – b3cft • Bombe rebuild detail • http://www.flickr.com/photos/b3cft/3797123899 – Karindalziel • Clouds • http://www.flickr.com/photos/nirak/644336486 – Enjoy Surveillance • What are you looking at? • http://www.flickr.com/photos/enjoy-surveillance/34795807/ – Solo • 45th Annual Watsonville Fly-in and Air Show • http://www.flickr.com/photos/donsolo/4959045491/in/photostream/ – SF Brit • Sunset over Iguazu • http://www.flickr.com/photos/cnbattson/4333692253/• Olympics Photos: www.london2012.com• Other Photos: EpicWin, FailBlog, Haha-Business
  47. Special Thanksto: – David Holroyd • Technical Architect BBC Sport (Olympics) – Matt Clark • Senior Technical Architect BBC Sport
  48. Thanks for listening• This presentation: – TBC• Me: – Andy “Bob” Brockhurst – Twitter: b3cft (and pretty much anywhere online) – www.kingkludge.net

×