Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Congestion Control for Spraying with Congested Paths

Barak Gerstein, Mark Silberstein, Isaac Keslassy

Abstract

Packet spraying approaches are increasingly deployed in datacenter networks. However, their combination with existing congestion control algorithms (CCAs) may lead to poor QoS, especially when some of the paths are congested. In this paper, we first model the throughput collapse of a wide array of CCAs when some of the paths are congested. We explain that since CCAs are typically designed for single-path routing, their estimation function focuses on the latest feedback and mishandles feedback that reflects multiple paths. We propose using a median feedback that is more robust to the varying signals that come with multiple paths. We introduce MSwift and MNSCC, which apply this median principle to Google's Swift and Ultra Ethernet's NSCC. We demonstrate that they can improve both CCAs, reaching better QoS both under congested paths and in uncongested networks.

Congestion Control for Spraying with Congested Paths

Abstract

Packet spraying approaches are increasingly deployed in datacenter networks. However, their combination with existing congestion control algorithms (CCAs) may lead to poor QoS, especially when some of the paths are congested. In this paper, we first model the throughput collapse of a wide array of CCAs when some of the paths are congested. We explain that since CCAs are typically designed for single-path routing, their estimation function focuses on the latest feedback and mishandles feedback that reflects multiple paths. We propose using a median feedback that is more robust to the varying signals that come with multiple paths. We introduce MSwift and MNSCC, which apply this median principle to Google's Swift and Ultra Ethernet's NSCC. We demonstrate that they can improve both CCAs, reaching better QoS both under congested paths and in uncongested networks.

Paper Structure

This paper contains 18 sections, 6 theorems, 11 equations, 14 figures, 1 table.

Table of Contents

  1. Introduction
  2. The $\Theta(1/\sqrt{q})$ CCA Multi-Path Rule
  3. CCAs
  4. Model
  5. Results
  6. MSwift
  7. Median framework for CCAs
  8. MSwift and MNSCC
  9. Nyquist rate
  10. Evaluations
  11. Settings
  12. Baseline workload
  13. Permutation workloads
  14. Sensitivity to parameters
  15. Additional scenarios
  16. ...and 3 more sections

Key Result

Theorem 1

The throughput of a single TCP flow is

Figures (14)

  • Figure 1: Scenario with single congested path.
  • Figure 2: Model comparison using the theoretical results in \ref{['tab:throughput-models']}.
  • Figure 3: AIMD dynamics for Swift, providing some intuition for why the grey area of $1/q$ is roughly proportional to $W \times W$.
  • Figure 4: Baseline workload: sprayed permutation flows with 4 ECMP flows.
  • Figure 5: CDF of FCT with baseline.
  • ...and 9 more figures

Theorems & Definitions (6)

  • Theorem 1: TCP
  • Theorem 2: DCTCP
  • Theorem 3: Swift
  • Theorem 4: SMaRTT
  • Theorem 5: Reordering-resilient Swift
  • Theorem 6: Median-based CCA