1 / 57

ElasticSwitch : Practical Work-Conserving Bandwidth Guarantees for Cloud Computing

ElasticSwitch : Practical Work-Conserving Bandwidth Guarantees for Cloud Computing. Lucian Popa Praveen Yalagandula * Sujata Banerjee Jeffrey C. Mogul + Yoshio Turner Jose Renato Santos. HP Labs * Avi Networks + Google. Goals. Provide Minimum Bandwidth Guarantees in Clouds

neva
Download Presentation

ElasticSwitch : Practical Work-Conserving Bandwidth Guarantees for Cloud Computing

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. ElasticSwitch: Practical Work-Conserving Bandwidth Guarantees for Cloud Computing Lucian Popa Praveen Yalagandula*Sujata Banerjee Jeffrey C. Mogul+Yoshio Turner Jose Renato Santos HP Labs *Avi Networks +Google

  2. Goals • Provide Minimum Bandwidth Guarantees in Clouds • Tenants can affect each other’s traffic • MapReducejobs can affect performance of user-facing applications • Large MapReduce jobs can delay the completion of small jobs • Bandwidth guarantees offer predictable performance

  3. Goals • Provide Minimum Bandwidth Guarantees in Clouds VS BX BZ BY Hose model Z X Y Virtual (imaginary) Switch Bandwidth Guarantees VMs of one tenant Other models based on hose model such as TAG [HotCloud’13]

  4. Goals • Provide Minimum Bandwidth Guarantees in Clouds • Work-Conserving Allocation • Tenants can use spare bandwidth from unallocated or underutilized guarantees

  5. Goals • Provide Minimum Bandwidth Guarantees in Clouds • Work-Conserving Allocation • Tenants can use spare bandwidth from unallocated or underutilized guarantees • Significantly increases performance • Average traffic is low [IMC09,IMC10] • Traffic is bursty

  6. Goals • Provide Minimum Bandwidth Guarantees in Clouds • Work-Conserving Allocation Everything reserved & used ElasticSwitch Free capacity XY bandwidth Bmin Bmin Bmin X Y Time

  7. Goals • Provide Minimum Bandwidth Guarantees in Clouds • Work-Conserving Allocation • Be Practical • Topology independent: work with oversubscribed topologies • Inexpensive: per VM/per tenant queues are expensive  work with commodity switches • Scalable: centralized controller can be bottleneck  distributed solution • Hard to partition: VMs can cause bottlenecks anywhere in the network

  8. Goals • Provide Minimum Bandwidth Guarantees in Clouds • Work-Conserving Allocation • Be Practical

  9. Prior Work

  10. Outline • Motivation and Goals • Overview • More Details • Guarantee Partitioning • Rate Allocation • Evaluation

  11. ElasticSwitch Overview: Operates At Runtime Tenant selects bandwidth guarantees. Models: Hose, TAG, etc. VMs placed, Admission Control ensures all guarantees can be met Oktopus[SIGCOMM’10] Hadrian[NSDI’10] CloudMirror[HotCLoud’13] VM setup Enforce bandwidth guarantees & Provide work-conservation Runtime ElasticSwitch

  12. ElasticSwitch Overview: Runs In Hypervisors • Resides in the hypervisor of each host • Distributed: Communicates pairwise following data flows VM VM VM ElasticSwitch Hypervisor Network VM VM VM ElasticSwitch ElasticSwitch Hypervisor Hypervisor

  13. ElasticSwitch Overview: Two Layers Guarantee Partitioning Provides Guarantees Rate Allocation Provides Work-conservation Hypervisor

  14. ElasticSwitch Overview: Guarantee Partitioning 1. Guarantee Partitioning: turns hose model into VM-to-VM pipe guarantees VM-to-VM control is necessary, coarser granularity is not enough

  15. ElasticSwitch Overview: Guarantee Partitioning 1. Guarantee Partitioning: turns hose model into VM-to-VM pipe guarantees VS Intra-tenant BX BZ BY Z X Y BXY BXZ VM-to-VM guarantees  bandwidths as if tenant communicates on a physical hose network

  16. ElasticSwitch Overview: Rate Allocation 1. Guarantee Partitioning: turns hose model into VM-to-VM pipe guarantees 2. Rate Allocation: uses rate limiters, increases rate between X-Y above BXY when there is no congestion between X and Y Work-conserving allocation VS Inter-tenant BX BZ BY Z X Y BXY RateXY ≥ X Y Limiter Hypervisor Hypervisor Unreserved/Unused Capacity

  17. ElasticSwitch Overview: Periodic Application Guarantee Partitioning Applied periodically and on new VM-to-VM pairs VM-to-VM guarantees Demand estimates Rate Allocation Applied periodically, more often Hypervisor

  18. Outline • Motivation and Goals • Overview • More Details • Guarantee Partitioning • Rate Allocation • Evaluation

  19. Guarantee Partitioning – Overview VS1 BX BQ Z X Y T Q Z T BXZ BTY BXY X Y Max-min allocation BQY • Goals: • Safety – don’t violate hose model • Efficiency – don’t waste guarantee • No Starvation – don’t block traffic Q

  20. Guarantee Partitioning – Overview VS1 BX BQ Z X Y T Q Z T BX = … = BQ= 100Mbps 33Mbps 66Mbps 33Mbps X Y Max-min allocation 33Mbps • Goals: • Safety – don’t violate hose model • Efficiency – don’t waste guarantee • No Starvation – don’t block traffic Q

  21. Guarantee Partitioning – Operation VS1 BX BQ Z X Y T Q Z T TY BY XZ XY XY BX BXY = min(BX, BY ) X Y X Y XY XY BX BY QY BY Hypervisor divides guarantee of each hosted VM between VM-to-VM pairs in each direction Q Source hypervisor uses the minimum between the source and destination guarantees

  22. Guarantee Partitioning – Safety VS1 BX BQ Z X Y T Q Z T TY BY XZ XY XY BX BXY = min(BX, BY ) X Y X Y XY XY BX BY QY BY Q Safety: hose-model guarantees are not exceeded

  23. Guarantee Partitioning – Operation VS1 BX BQ Z X Y T Q Z T BX = … = BQ= 100Mbps XY XY BXY = min(BX, BY ) X Y Q

  24. Guarantee Partitioning – Operation VS1 BX BQ Z X Y T Q Z T BX = … = BQ= 100Mbps TY XY XY XZ BY = 33 BXY = min(BX, BY) BX = 50 BXY = 33 X Y X Y XY XY BY = 33 BX = 50 QY BY = 33 Q

  25. Guarantee Partitioning – Efficiency VS1 BX BQ Z X Y T Q Z T BX = … = BQ= 100Mbps TY XZ BY = 33 BX = 50 BXY = 33 X Y XY XY BY = 33 BX = 50 QY BY = 33 1 Q What happens when flows have low demands? Hypervisor divides guarantees max-min based on demands (future demands estimated based on history)

  26. Guarantee Partitioning – Efficiency VS1 BX BQ Z X Y T Q Z T BX = … = BQ= 100Mbps 66 TY XZ BY = 33 BX = 50 BXY = 33 X Y XY XY BY = 33 BX = 50 33 QY BY = 33 1 Q What happens when flows have low demands? 2 How to avoid unallocated guarantees?

  27. Guarantee Partitioning – Efficiency VS1 BX BQ Z X Y T Q Z T BX = … = BQ= 100Mbps 66 TY XZ BY = 33 BX = 50 BXY = 33 X Y XY XY BY = 33 BX = 50 33 QY BY = 33 Source considers destination’s allocation when destination is bottleneck Q Guarantee Partitioning converges

  28. Outline • Motivation and Goals • Overview • More Details • Guarantee Partitioning • Rate Allocation • Evaluation

  29. Rate Allocation RXY Spare bandwidth Fully used X BXY Guarantee Partitioning BXY Time Congestion data Rate Allocation Rate RXY Limiter Y

  30. Rate Allocation RXY= max(BXY, RTCP-like) X Guarantee Partitioning BXY Congestion data Rate Allocation Rate RXY Limiter Y

  31. Rate Allocation RXY= max(BXY, RTCP-like) Another Tenant Guarantee

  32. Rate Allocation RXY= max(BXY, Rweighted-TCP-like) Weight is the BXY guarantee L = 1Gbps RXY = 333Mbps X Y BXY = 100Mbps RXT = 666Mbps Z T BZT = 200Mbps

  33. Rate Allocation – Congestion Detection • Detect congestion through dropped packets • Hypervisors add/monitor sequence numbers in packet headers • Use ECN, if available

  34. Rate Allocation – Adaptive Algorithm • Use Seawall [NSDI’11] as rate-allocation algorithm • TCP-Cubic like • Essential improvements (for when using dropped packets) Many flows probing for spare bandwidth affect guarantees of others

  35. Rate Allocation – Adaptive Algorithm • Use Seawall [NSDI’11] as rate-allocation algorithm • TCP-Cubic like • Essential improvements (for when using dropped packets) • Hold-increase: hold probing for free bandwidth after a congestion event. Holding time is inversely proportional to guarantee. Rate increasing Guarantee Holding time

  36. Outline • Motivation and Goals • Overview • More Details • Guarantee Partitioning • Rate Allocation • Evaluation

  37. Evaluation Setup • Implementation in Linux • Logic in user-space: controls rate limiters, sends control packets • Modified kernel OVS • Testbed • ~100 servers • 1Gbps tree network

  38. Evaluation – Many-to-one VS1 TCP 450Mbps L = 1Gbps X X VS2 450Mbps Z Z … UDP Edge or core

  39. Evaluation – Many-to-one X Z Throughput (Mbps) Senders to Z

  40. Evaluation – Many-to-one X Z No Protection VM Z takes all the bandwidth Throughput (Mbps) … Senders to Z

  41. Evaluation – Many-to-one X Z No Protection Static Reservation (e.g., Oktopus) Wasted bandwidth Throughput (Mbps) Senders to Z

  42. Evaluation – Many-to-one X Z No Protection Static Reservation (e.g., Oktopus) ElasticSwitch Work-conserving Throughput (Mbps) Senders to Z

  43. Evaluation – Many-to-one X Z No Protection Provides guarantees Static Reservation (e.g., Oktopus) ElasticSwitch Throughput (Mbps) Senders to Z

  44. Evaluation – Many-to-one X Z ElasticSwitch Ideal behavior Throughput (Mbps) Senders to Z

  45. Evaluation – MapReduce • Setup • 44 servers, 4x oversubscribed topology, 4 VMs/server • Each tenant runs one job, all VMs of all tenants same guarantee • Two scenarios: • Light • 10% of VM slots are either a mapper or a reducer • Randomly placed • Heavy • 100% of VM slots are either a mapper or a reducer • Mappers are placed in one half of the datacenter

  46. Evaluation – MapReduce CDF Worst case shuffle completion time / static reservation

  47. Evaluation – MapReduce No Protection ElasticSwitch Longest completion reduced from No Protection Work-conserving pays off: finish faster than static reservation CDF Light Setup Worst case shuffle completion time / static reservation

  48. Evaluation – MapReduce ElasticSwitch enforces guarantees in worst case ElasticSwitch No Protection up to 160X CDF Guarantees are useful in reducing worst-case shuffle completion Heavy Setup Worst case shuffle completion time / static reservation

  49. ElasticSwitch Summary • Properties • Bandwidth Guarantees: hose model or derivatives • Work-conserving • Practical: oversubscribed topologies, commodity switches, decentralized • Design: two layers • Guarantee Partitioning: provides guarantees by transforming hose-model guarantees into VM-to-VM guarantees • Rate Allocation: enables work conservation by increasing rate limits above guarantees when no congestion HP Labs is hiring!

  50. Backup Slides

More Related