1 / 53

Dynamic Regulation of Mobile 3G/HSPA Uplink Buffer with Receiver-Side Flow Control

Yin Xu, Wai Kay Leong, Ben Leong National University of Singapore. Dynamic Regulation of Mobile 3G/HSPA Uplink Buffer with Receiver-Side Flow Control. Ali Razeen Duke University. US smartphone penetration exceeded 50% in Q2, 2012 Mobile data traffic growing rapidly as well.

audra
Download Presentation

Dynamic Regulation of Mobile 3G/HSPA Uplink Buffer with Receiver-Side Flow Control

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. Yin Xu, Wai Kay Leong, Ben Leong National University of Singapore Dynamic Regulation of Mobile 3G/HSPA Uplink Buffer with Receiver-Side Flow Control Ali Razeen Duke University

  2. US smartphone penetration exceeded 50% in Q2, 2012 Mobile data traffic growing rapidly as well Smartphones are everywhere Source: http://www.chetansharma.com/USmarketupdateQ22012.htm

  3. Users uploading significant amounts of data in the form of photos and videos e.g. AT&T observed 40% more data uploaded than downloaded during a football match (7 Feb 2012) • Uploads often conducted in the background Not just for surfing the web…

  4. What are the users doing? Hmm… I got the new HTC one X…lala Hmm…I still have 10GB of data….lolo Hmm…I got 25GB extra DropBox space!! Hmm…Upload all my photo to DropBox!!! Hmm…Show offmy new phone!! …Facebook……blabla…. ah!! I cannot refresh my facebook wall! !!Stupid phone!! Stupid network!! ….....I should complain!! What happened?

  5. Background uploads can degrade downloads significantly! The Problem

  6. Android Phones • 3 Local Telcos • 7.2 Mbps downlink • 2.0 Mbps uplink Experiment Setup Client Upload until download completion Server Download 1MB Upload 1MB Download 1MBwithout upload

  7. Download Roundtrip Time With Upload (s) Without Upload (s)

  8. Download Throughput With Upload (kbps) Without Upload (kbps)

  9. Problem is exacerbated by low uplink throughput Download Performance Uplink Bandwidth (kbps)

  10. Experiment Setup (Loopback ) Download 1MB Server Continuous background upload Client

  11. RTT dominated by uplink delay Delay (s) Packet arrival time (s)

  12. NOT caused by ACK Compression • Data Pendulum Problem [Heusse et al. 2011] • Sized properly, buffers take turns to fill up • Sized improperly, low-speed link with large buffer becomes the sole bottleneck • Uplink is the bottleneck in a 3G/HSPA mobile network Understanding the Problem

  13. Can we just size the uplink buffer correctly?

  14. Understanding the Problem Upload Throughput (kbps) NOPE! CANNOT USE A FIXED SIZE BUFFER! Time of Day (24 hours)

  15. Optimizing how the ACKs are sent [Balakrishnan et al. 1999/2002] Using different queues for dataand ACK packets[Podlesny et al. 2012] TCP Vegas[Brakmo et al. 1995] All Sender-Side Solutions Previous Solutions

  16. Not General • Only works for the devices already deployed with the solution • Devices may use network interfaces other than 3G/HSPA (e.g. Wi-Fi) • “Implement complexity at the server, not the client” • It may take years to update client-side software [Adya et al. 2011] Why not Sender-Side Solutions?

  17. Receiver-Side Flow Control (RSFC) Our Solution

  18. Our Approach Can implement at ISP network proxies Works transparently for any device using the 3G/HSPA mobile network Changes immediately deployable

  19. Practical Deployment Easily deployed at ISP proxy Base station A good place for RSFC Proxy Internet 19

  20. Freeze-TCP[Goff et al. 2000] • Reducing delay for interactive applications while maintaining throughput for bulk transfers [Spring et al. 2000] • Improving fairness [Kalampoukas et al. 2002, Andrew et al. 2008] Used for Other Purposes Receiver-Side Solutions

  21. Key Idea Reduce # of packets in the uplink buffer by adjusting the TCP receiver window (rwnd) What is the right rwnd?

  22. rwnd Value Set to bandwidth-delayproduct (BDP)? Not so simple… How do we estimate BDP? Network fluctuations

  23. Approach: Negative Feedback Estimate time packet spends in buffer tbuff using TCP Timestamp Set a threshold T tbuff > T, clamp rwnd tbuff < T, increase rwnd

  24. Estimating tbuff Receiver Sender Packets in buffer Timestamp: TSval = ts Time x ts x tbuff tr – ts = RD Relative Delay x x x tu time = tr

  25. Estimating tbuff Receiver Sender Packets in buffer Timestamp: TSval = ts Time x x ts tu time = tr • tr – ts = RDmin • Minimal Relative Delay tbuff = RD – RDmin No need to synchronize sender and receiver!

  26. Measure receive rate ρat receiver • Minimal RTT (RTTmin) • Ideal window is the bandwidth-delay product: • rwnd = ρ× RTTmin Estimating BDP

  27. tbuff > T, rwnd = ρ× RTTmin (fast state) • tbuff < T, rwnd++(slow state) • In our implementation • T is set to RTTmin Summary

  28. Changes in bandwidth • Decrease in the delay • Increase in the delay • Slight increase:detect increased receive rate ρ • Large increase: monitor state Handling changes in the network ? See details in paper!

  29. Reduces RTT Improves download throughput Reduces webpage loading time Fair and efficient Adapts to changes in network conditions Compatible with sender-side algorithms Evaluation

  30. Evaluation • Reduces RTT • Improves download throughput • Reduces webpage loading time • Fair and efficient • Adapts to changes in network conditions • Compatible with sender-side algorithms 30

  31. Reduce RTT Client Server Upload 1MB with RSFC Upload 1MB with Cubic

  32. Reduction in RTT

  33. Reduction in RTT

  34. No Reduction in Throughput

  35. Improve Download Throughput Upload (u1)with Cubic until download completes Upload (u2)with RSFC until download completes Client Server Download 1MB (d1) Download 1MB (d2) Download (d0) 1MBwithout upload

  36. Better Downstream Performance

  37. Better Downstream Performance

  38. Better Downstream Performance

  39. Little Impact on Upstream

  40. Improve Web Surfing • Alexa top 100 sites Client Server Upload with RSFC until website is loaded Upload with Cubic until website is loaded Web surfing without upload Web surfing Web surfing

  41. Webpages Load Faster

  42. Webpages Load Faster

  43. Webpages Load Faster

  44. Conclusion Saturated uplink can cause serious performance degradation Receiver-Side Flow Control Reduces queuing delay significantly Improves downstream performance Reduces loading time of webpages Compatible with existing TCP variants Easily deployed at ISP proxies

  45. THANKYOU

  46. Run two RSFC uploads concurrently Calculate Jain fairness index: Fairness of Competing RSFC Uploads

  47. Fairness of Competing RSFC Uploads

  48. Run two RSFC uploads concurrently • Compare the aggregate throughput to a single TCP: Efficiency of Competing RSFC Uploads

  49. Efficiency of Competing RSFC Uploads

  50. Efficiency of Competing RSFC Uploads

More Related