1 / 32

Microsoft Research Stanford University

Data Center TCP (DCTCP). Mohammad Alizadeh , Albert Greenberg, David A. Maltz , Jitendra Padhye Parveen Patel, Balaji Prabhakar , Sudipta Sengupta , Murari Sridharan Presented by Shaddi Hasan. Microsoft Research Stanford University. Case Study: Microsoft Bing.

brier
Download Presentation

Microsoft Research Stanford University

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. Data Center TCP (DCTCP) Mohammad Alizadeh, Albert Greenberg, David A. Maltz, JitendraPadhyeParveen Patel, BalajiPrabhakar, SudiptaSengupta, MurariSridharan Presented by ShaddiHasan Microsoft Research Stanford University

  2. Case Study: Microsoft Bing • Measurements from 6000 server production cluster • Instrumentation passively collects logs • Application-level • Socket-level • Selected packet-level • More than 150TB of compressed data over a month

  3. Partition/Aggregate Application Structure Deadline = 250ms MLA MLA TLA Picasso • Time is money • Strict deadlines (SLAs) • Missed deadline • Lower quality result ……… 1. Art is a lie… 1. 1. Deadline = 50ms 2. The chief… 2. Art is a lie… 2. Art is… ….. 3. ….. ….. 3. 3. Picasso “I'd like to live as a poor man with lots of money.“ “The chief enemy of creativity is good sense.“ “Computers are useless. They can only give you answers.” “Bad artists copy. Good artists steal.” “Art is a lie that makes us realize the truth. “It is your work in life that is the ultimate seduction.“ “Inspiration does exist, but it must find you working.” “Everything you can imagine is real.” Deadline = 10ms Worker Nodes

  4. Workloads • Partition/Aggregate (Query) • Short messages [50KB-1MB] (Coordination, Control state) • Large flows [1MB-50MB] (Data update) Delay-sensitive Delay-sensitive Throughput-sensitive

  5. Impairments • Incast • Queue Buildup • Buffer Pressure

  6. Incast Worker 1 • Synchronized mice collide. • Caused by Partition/Aggregate. Aggregator Worker 2 Worker 3 RTOmin = 300 ms Worker 4 TCP timeout

  7. Incast Really Happens • Requests are jittered over 10ms window. • Jittering switched off around 8:30 am. MLA Query Completion Time (ms) 99.9th percentile is being tracked. Jittering trades off median against high percentiles.

  8. Queue Buildup Sender 1 • Big flows buildup queues. • Increased latency for short flows. Receiver Sender 2 • Measurements in Bing cluster • For 90% packets: RTT < 1ms • For 10% packets: 1ms < RTT < 15ms

  9. Data Center Transport Requirements • High Burst Tolerance • Incast due to Partition/Aggregate is common. • Low Latency • Short flows, queries • 3. High Throughput • Continuous data updates, large file transfers The challenge is to achieve these three together.

  10. Tension Between Requirements High Throughput Low Latency High Burst Tolerance • Deep Buffers: • Queuing Delays • Increase Latency • Shallow Buffers: • Bad for Bursts & • Throughput Objective: Low Queue Occupancy & High Throughput DCTCP • AQM – RED: • Avg Queue Not Fast • Enough for Incast • Reduced RTOmin (SIGCOMM ‘09) • Doesn’t Help Latency

  11. The DCTCP Algorithm

  12. Nugget time • TCP causes problems because it tries to fill buffers • Congestion is signaled by loss  buffer overflow. • Use ECN to signal congestion, not loss • Keeps buffers short, reduces retransmissions

  13. Review: The TCP/ECN Control Loop Sender 1 ECN = Explicit Congestion Notification ECN Mark (1 bit) Receiver Sender 2

  14. Small Queues & TCP Throughput:The Buffer Sizing Story • Bandwidth-delay product rule of thumb: • A single flow needs buffers for 100% Throughput. Cwnd Buffer Size B Throughput 100%

  15. Small Queues & TCP Throughput:The Buffer Sizing Story • Bandwidth-delay product rule of thumb: • A single flow needs buffers for 100% Throughput. • Appenzellerrule of thumb (SIGCOMM ‘04): • Large # of flows: is enough. Cwnd Buffer Size B Throughput 100%

  16. Small Queues & TCP Throughput:The Buffer Sizing Story • Bandwidth-delay product rule of thumb: • A single flow needs buffers for 100% Throughput. • Appenzellerrule of thumb (SIGCOMM ‘04): • Large # of flows: is enough. • Can’t rely on stat-mux benefit in the DC. • Measurements show typically 1-2 big flows at each server, at most 4.

  17. Small Queues & TCP Throughput:The Buffer Sizing Story • Bandwidth-delay product rule of thumb: • A single flow needs buffers for 100% Throughput. • Appenzellerrule of thumb (SIGCOMM ‘04): • Large # of flows: is enough. • Can’t rely on stat-mux benefit in the DC. • Measurements show typically 1-2 big flows at each server, at most 4. Real Rule of Thumb: Low Variance in Sending Rate → Small Buffers Suffice B

  18. Two Key Ideas • React in proportion to the extent of congestion, not its presence. • Reduces variancein sending rates, lowering queuing requirements. • Mark based on instantaneous queue length. • Fast feedback to better deal with bursts.

  19. Data Center TCP Algorithm B K Don’t Mark Mark Switch side: • Mark packets when Queue Length > K. • Sender side: • Maintain running average of fractionof packets marked (α). • In each RTT: • Adaptive window decreases: • Note: decrease factor between 1 and 2.

  20. DCTCP in Action (Kbytes) Setup: Win 7, Broadcom 1Gbps Switch Scenario: 2 long-lived flows, K = 30KB

  21. Why it Works • High Burst Tolerance • Large buffer headroom → bursts fit. • Aggressive marking→sources react before packets are dropped. • Low Latency • Small buffer occupancies → low queuing delay. 3. High Throughput • ECN averaging →smooth rate adjustments, low variance.

  22. Baseline Background Flows Query Flows

  23. Baseline Background Flows Query Flows • Low latency for short flows.

  24. Baseline Background Flows Query Flows • Low latency for short flows. • High throughput for long flows.

  25. Baseline Background Flows Query Flows • Low latency for short flows. • High throughput for long flows. • High burst tolerance for query flows.

  26. Conclusions • DCTCP satisfies all our requirements for Data Center packet transport. • Handles bursts well • Keeps queuing delays low • Achieves high throughput • Features: • Very simple change to TCP and a single switch parameter. • Based on mechanisms already available in Silicon.

  27. Thoughts • Convergence times? • Actually requires persistent connections • Section 4.2.2: “All communication is over long-lived connections, so there is no three-way handshake for each request.” • TCP will blow DCTCP away if sharing a buffer

  28. Thoughts • This is basically the right answer for handling incast and meeting deadlines • Use ECN to keep queues short, rather than filling up your buffers and waiting for loss. • D3 incorporates explicit knowledge of deadlines, but this complicates API and necessitates more changes. • Complementary to MPTCP (or any other TCP for that matter)

  29. Scaled Background & Query10x Background, 10x Query

  30. Thoughts • “The key contribution here is […] the act of deriving multi-bit feedback from the information present in the single-bit sequence of marks.” • Calculating a moving average is not a contribution. • Using ECN + clever scheme for adjusting CWIN is!

  31. Analysis • How low can DCTCP maintain queues without loss of throughput? • How do we set the DCTCP parameters? • Need to quantify queue size oscillations (Stability). 85% Less Buffer than TCP

More Related