AI helps you reading Science
AI generates interpretation videos
AI extracts and analyses the key points of the paper to generate videos automatically
AI parses the academic lineage of this thesis
AI extracts a summary of this paper
Our work uses the TCP-level statistics in the server network stack to help Content distribution networks detect and diagnose performance problems
Identifying performance bottlenecks in CDNs through TCP-level monitoring
W-MUST@SIGCOMM, pp.49-54, (2011)
Content distribution networks (CDNs) need to make decisions, such as server selection and routing, to improve performance for their clients. The performance may be limited by various factors such as packet loss in the network, a small receive buffer at the client, or constrained server CPU and disk resources. Conventional measurement tech...More
PPT (Upload PPT)
- Content distribution networks (CDNs) run replicated servers to deliver content to a large number of clients.
- To optimize data-transfer performance, CDNs need to pick the right server for each group of clients, and select routing paths that traverse the Internet quickly .
- CDNs can optimize their software or upgrade their machines if the servers are limiting the performance, work with their ISPs if network performance is spotty, or notify clients if a small receive buffer is the bottleneck.
- CDNs need to collect and analyze measurement data.
- Conventional measurement techniques are either too coarse-grained for diagnosing problems, or too expensive for achieving good coverage
- Content distribution networks (CDNs) run replicated servers to deliver content to a large number of clients
- This paper presents a tool for monitoring and analyzing TCP statistics, and an analysis of a CoralCDN node in PlanetLab for six weeks
- Our analysis shows that more than 10% of connections are server-limited at least 40% of the time, and many connections are limited by the congestion window despite no packet loss
- We show that network performance problems persist over time, allowing measurements from one time period to guide a Content distribution networks’s future decisions about server and path selection
- Our work uses the TCP-level statistics in the server network stack to help Content distribution networks detect and diagnose performance problems
- We plan to explore better ways to pinpoint network performance bottlenecks, as well as techniques for diagnosing performance problems for virtual machines running on shared computing platforms in the cloud
- The authors' analysis shows that more than 10% of connections are server-limited at least 40% of the time, and many connections are limited by the congestion window despite no packet loss.
- More than 10% of connections are application-limited for more than 40% of their lifetime.
- In CoralCDN, more than 80% of the connections last less than 1 second.
- As shown in Table 4, 1.26% of the connections experienced a packet loss rate higher than 20%
- The authors' work uses the TCP-level statistics in the server network stack to help CDNs detect and diagnose performance problems.
- The authors plan to explore better ways to pinpoint network performance bottlenecks, as well as techniques for diagnosing performance problems for virtual machines running on shared computing platforms in the cloud.
- The authors are exploring how to combine the TCP-level statistics with application logs in order to pinpoint the performance bugs in online services
- Table1: Key TCP Statistics in our Tool can easily distinguish these performance limitations using the TCP-level statistics as summarized in Table 2
- Table2: Performance Limitation Classifier the server network stack, the clients, and the server applications. We also accumulate the number of lost packets
- Table3: Percentage of Connections that have Different TCP Performance Problems the server. As such, directing clients to the closest Coral proxy (in terms of RTT) is critically important for improving performance. In addition, the network stack can use a larger initial congestion window, as some commercial content providers do [<a class="ref-link" id="c4" href="#r4">4</a>]
- Table4: Percentage of Connections that have Different Levels of Packet Loss
- Route Views Project. http://www.routeviews.org/.
- Web100 Project. http://www.web100.org/.
- Web10G Project. http://web10g.org/.
- N. Dukkipati, T. Refice, Y. Cheng, J. Chu, T. Herbert, A. Agarwal, A. Jain, and N. Sutin. An Argument for Increasing TCP’s Initial Congestion Window. ACM SIGCOMM Computer Communication Review, 40, 2010.
- R. Fonseca, M. J. Freedman, and G. Porter. Experiences with Tracing Causality in Networked Services. In USENIX INM/WREN, San Jose, CA USA, April 2010.
- R. Fonseca, G. Porter, R. H. Katz, S. Shenker, and I. Stoica. X-Trace: A Pervasive Network Tracing Framework. In USENIX NSDI, Cambridge, MA USA, April 2007.
- M. J. Freedman. Experiences with CoralCDN: A Five-Year Operational View. In USENIX NSDI, San Jose, CA USA, April 2010.
- M. J. Freedman, E. Freudenthal, and D. Mazieres. Democratizing Content Publication with Coral. In USENIX NSDI, San Francisco, CA USA, March 2004.
- A. Gerber, J. Pang, O. Spatscheck, and S. Venkataraman. Speed Testing without Speed Tests: Estimating Achievable Download Speed from Passive Measurements. In ACM IMC, Melbourne, Australia, November 2010.
- R. Krishnan, H. V. Madhyastha, S. Srinivasan, S. Jain, A. Krishnamurthy, T. Anderson, and J. Gao. Moving Beyond End-to-End Path Information to Optimize CDN Performance. In ACM IMC, Chicago, IL USA, November 2009.
- H. V. Madhyastha, T. Isdal, M. Piatek, C. Dixon, T. Anderson, A. Krishnamurthy, and A. Venkataramani. iPlane: An information plane for distributed services. In USENIX OSDI, Seattle, WA USA, November 2006.
- M. Mathis, J. Heffner, and R. Raghunarayan. RFC 4898: TCP Extended Statistics MIB. http://www.ietf.org/rfc/rfc4898.txt, May 2007.
- V. N. Padmanabhan, L. Qiu, and H. J. Wang. Server-based inference of Internet link lossiness. In IEEE INFOCOM, San Francisco, CA USA, March 2003.
- R. S. Prasad, M. Murray, C. Dovrolis, and k. claffy. Bandwidth estimation: Metrics, measurement techniques, and tools. In IEEE Network, November/December 2003.
- A.-J. Su, D. R. Choffnes, A. Kuzmanovic, and F. E. Bustamante. Drafting Behind Akamai. In ACM SIGCOMM, Pisa, Italy, September 2006.
- M. Yu, A. Greenberg, D. Maltz, J. Rexford, L. Yuan, S. Kandula, and C. Kim. Profiling Network Performance for Multi-tier Data Center Applications. In USENIX NSDI, Boston, MA USA, March 2011.
- M. Zhang, C. Zhang, V. Pai, L. Peterson, and R. Wang. PlanetSeer: Internet Path Failure Monitoring and Characterization in Wide-Area Services. In USENIX OSDI, San Francisco, CA USA, December 2004.
- Y. Zhang, L. Breslau, V. Paxson, and S. Shenker. On the Characteristics and Origins of Internet Flow Rates. In ACM SIGCOMM, Pittsburgh, PA USA, August 2002.