Web Performance 성민영 SNU Computer Systems lab.. 2 차례 4 Modeling the Performance of HTTP Over Several Transport Protocols. 4 Summary Cache : A Scaleable.

Slides:



Advertisements
Similar presentations
Summary Cache: A Scalable Wide-Area Web Cache Sharing Protocol Li Fan, Pei Cao and Jussara Almeida University of Wisconsin-Madison Andrei Broder Compaq/DEC.
Advertisements

1 Transport Protocols & TCP CSE 3213 Fall April 2015.
Chapter 7 – Transport Layer Protocols
Pervasive Web Content Delivery with Efficient Data Reuse Chi-Hung Chi and Cao Yang School of Computing National University of Singapore
Cooperative Caching of Dynamic Content on a Distributed Web Server Vegard Holmedahl, Ben Smith, Tao Yang Speaker: SeungLak Choi, DB Lab., CS Dept.
School of Information Technologies TCP Congestion Control NETS3303/3603 Week 9.
© 2007 Cisco Systems, Inc. All rights reserved.Cisco Public 1 Version 4.0 OSI Transport Layer Network Fundamentals – Chapter 4.
Web Caching Schemes1 A Survey of Web Caching Schemes for the Internet Jia Wang.
1 Prefetching the Means for Document Transfer: A New Approach for Reducing Web Latency 1. Introduction 2. Data Analysis 3. Pre-transfer Solutions 4. Performance.
Transport Layer 3-1 outline r TCP m segment structure m reliable data transfer m flow control m congestion control.
Transport Layer 3-1 Fast Retransmit r time-out period often relatively long: m long delay before resending lost packet r detect lost segments via duplicate.
CSCE 515: Computer Network Programming Chin-Tser Huang University of South Carolina.
Internet Networking Spring 2006 Tutorial 12 Web Caching Protocols ICP, CARP.
Summary Cache: A Scalable Wide-Area Web Cache Sharing Protocol By Abuzafor Rasal and Vinoth Rayappan.
Internet Networking Spring 2002 Tutorial 11 T/TCP (TCP extension for Transactions)
Hypertext Transfer Protocol Kyle Roth Mark Hoover.
1 Web Performance Modeling Chapter New Phenomena in the Internet and WWW Self-similarity - a self-similar process looks bursty across several time.
Network Traffic Measurement and Modeling CSCI 780, Fall 2005.
1 Spring Semester 2007, Dept. of Computer Science, Technion Internet Networking recitation #13 Web Caching Protocols ICP, CARP.
CSCE 515: Computer Network Programming Chin-Tser Huang University of South Carolina.
Internet and Intranet Protocols and Applications Section V: Network Application Performance Lecture 11: Why the World Wide Wait? 4/11/2000 Arthur P. Goldberg.
Internet Networking Spring 2002 Tutorial 13 Web Caching Protocols ICP, CARP.
1 K. Salah Module 6.1: TCP Flow and Congestion Control Connection establishment & Termination Flow Control Congestion Control QoS.
1Bloom Filters Lookup questions: Does item “ x ” exist in a set or multiset? Data set may be very big or expensive to access. Filter lookup questions with.
Web Caching Schemes For The Internet – cont. By Jia Wang.
1 The Mystery of Cooperative Web Caching 2 b b Web caching : is a process implemented by a caching proxy to improve the efficiency of the web. It reduces.
Copyright © 2005 Department of Computer Science CPSC 641 Winter Tutorial: TCP 101 The Transmission Control Protocol (TCP) is the protocol that sends.
Internet Basics.
Process-to-Process Delivery:
SMUCSE 4344 application layer. SMUCSE 4344 application vs. application-layer protocols application-layer protocol is just one piece –how the end hosts.
Networking Basics TCP/IP TRANSPORT and APPLICATION LAYER Version 3.0 Cisco Regional Networking Academy.
Advanced Network Architecture Research Group 2001/11/149 th International Conference on Network Protocols Scalable Socket Buffer Tuning for High-Performance.
Transport Layer 4 2: Transport Layer 4.
Lecture#2 on Internet and World Wide Web. Internet Applications Electronic Mail ( ) Electronic Mail ( ) Domain mail server collects incoming mail.
Protocol(TCP/IP, HTTP) 송준화 조경민 2001/03/13. Network Computing Lab.2 Layering of TCP/IP-based protocols.
Transport over Wireless Networks Myungchul Kim
CSE679: Computer Network Review r Review of the uncounted quiz r Computer network review.
Dr. Yingwu Zhu Summary Cache : A Scalable Wide- Area Web Cache Sharing Protocol.
HighSpeed TCP for High Bandwidth-Delay Product Networks Raj Kettimuthu.
CSE 461 HTTP and the Web. This Lecture  HTTP and the Web (but not HTML)  Focus  How do Web transfers work?  Topics  HTTP, HTTP1.1  Performance Improvements.
Communication Paradigm for Sensor Networks Sensor Networks Sensor Networks Directed Diffusion Directed Diffusion SPIN SPIN Ishan Banerjee
Networking Fundamentals. Basics Network – collection of nodes and links that cooperate for communication Nodes – computer systems –Internal (routers,
Transport Layer3-1 TCP throughput r What’s the average throughout of TCP as a function of window size and RTT? m Ignore slow start r Let W be the window.
Chapter 24 Transport Control Protocol (TCP) Layer 4 protocol Responsible for reliable end-to-end transmission Provides illusion of reliable network to.
ICP and the Squid Web Cache Duane Wessels and K. Claffy 산업공학과 조희권.
Doc.: IEEE /1317r0 Submission December 2009 Vinko Erceg, BroadcomSlide 1 Internet Traffic Modeling Date: Authors: NameAffiliationsAddressPhone .
Measuring the Capacity of a Web Server USENIX Sympo. on Internet Tech. and Sys. ‘ Koo-Min Ahn.
Defining High-Speed Protocols: Five Challenges and an Example that Survives the Challenges Joseph D. Touch IEEE Journal on Selected Areas in Communications.
Hint-based Acceleration of Web Proxy Cache Daniela Rosu Arun Iyengar Daniel Dias IBM T.J.Watson Research Center Unversity of Yuan Ze,Syslab Mike Tien
Renesas Electronics America Inc. © 2010 Renesas Electronics America Inc. All rights reserved. Overview of Ethernet Networking A Rev /31/2011.
79 Sidevõrgud IRT 4060/ IRT 0020 vooruloeng 8 / 3. nov 2004 Vooülekanne Avo Ots telekommunikatsiooni õppetool, TTÜ raadio- ja sidetehnika inst.
09/13/04 CDA 6506 Network Architecture and Client/Server Computing Peer-to-Peer Computing and Content Distribution Networks by Zornitza Genova Prodanoff.
IP Routing table compaction and sampling schemes to enhance TCAM cache performance Author: Ruirui Guo a, Jose G. Delgado-Frias Publisher: Journal of Systems.
TCP/IP1 Address Resolution Protocol Internet uses IP address to recognize a computer. But IP address needs to be translated to physical address (NIC).
Ó 1998 Menascé & Almeida. All Rights Reserved.1 Part VIII Web Performance Modeling (Book, Chapter 10)
WWW and HTTP King Fahd University of Petroleum & Minerals
COMP 431 Internet Services & Protocols
Web Caching? Web Caching:.
Exam 1 Review.
Internet Networking recitation #12
Transport Layer Unit 5.
TCP Sequence Number Plots
Edge computing (1) Content Distribution Networks
IS 4506 Server Configuration (HTTP Server)
CSE 461 HTTP and the Web.
Networking Theory (part 2)
Lecture 1: Bloom Filters
Computer Networks Protocols
Networking Theory (part 2)
Presentation transcript:

Web Performance 성민영 SNU Computer Systems lab.

2 차례 4 Modeling the Performance of HTTP Over Several Transport Protocols. 4 Summary Cache : A Scaleable Web Cache Sharing Protocol. 4 Web Server Workload Characterization.

Modeling the Performance of HTTP Over Several Transport Protocols.

4 Transport Protocols (1/2) 4 TCP 4 Persistent-Connection HTTP (P-HTTP) –Proposed by Padmanabhan and Mogul. –A variant of HTTP that uses one TCP connection to carry multiple HTTP requests. –Amortizes TCP’s connection overhead over multiple HTTP interactions. –A version of P-HTTP is part of HTTP 1.1 spec. –Pipelining can be used to get better performance.

5 Transport Protocols (2/2) 4 Transaction TCP –Caches per-host information sufficient to bypass the TCP’s three-way handshake and avoid slow start. –Also, shortens TCP’s TIME_WAIT period from 240 to 12 sec. 4 UDP-Based Request-Response Protocols –Reliable message passing protocol built atop UDP. –ex) Asynchronous Reliable Delivery Protocol(ARDP). –ARDP borrows TCP-style flow-control, congestion- avoidance, and retransmission algorithms. –Avoids TCP’s three-way handshake.

6 Network and Traffic Model (1/3) 4 Network Model –Network characteristics round-trip time (rtt) bandwidth (bw) maximum segment size (mss) segment-transmission time (stt) stt = mss / bw maximum useful window size (muws) muws =  rtt / stt 

7 Network and Traffic Model (2/3) –Network characteristics for several existing network.

8 Network and Traffic Model (3/3) 4 Traffic Model –small page : single 5 KB web page –medium page : single 25 KB web page –large page : single 100 KB web page –small cluster : single 3,220 KB page with three embedded images, sizes 57,613B, 2,344B, and 14,190B –large cluster : single 100 KB page with 10 embedded 25 KB images

9 Protocol Analysis (1/7) 4 Classes of protocols –TCP, Connection caching protocols(P-HTTP, T/TCP), and UDP-based request-response protocols. 4 Minimum Transmit Times –minimum possible transaction time

10 Protocol Analysis (2/7) –a series of n independent requests pipelined 4 Simple Model –one round-trip overhead per reply rtt / (reply size /bw) 4 HTTP over TCP –TCP slow-start

11 Protocol Analysis (3/7) Performance Overhead for Ethernet, modem, and ISDN networks is reasonable (muws at most 2) Networks such as Fast-Ethernet, Fast-Internet, ADSL have higher overheads (much higher muws)

12 Protocol Analysis (4/7) TCP congestion avoidance overhead.(S TCP /S min )

13 Protocol Analysis (5/7) 4 HTTP over TCP with connection caching –P-HTTP, T/TCP

14 Protocol Analysis (6/7) –Observations Caching TCP performance is somewhat better than standard TCP for the cluster cases. Overhead is still high for the Fast-Ethernet and Fast- Internet ( large bandwidth-delay product) 4 HTTP over UDP-Based Protocols –ARDP avoids TCP’s three-way handshake

15 Protocol Analysis (7/7) –Avoiding the 3-way handshake is especially helpful for single, brief request-response interactions.

Summary Cache : A Scaleable Web Cache Sharing Protocol.

17 Web Caching 4 Internet Cache Protocol (ICP) –A web cache sharing protocol by Harvest group –ICP discovers cache hits in other proxies by having the proxy multicast a query message to all other proxies whenever a cache miss occurs. –Not widely deployed because of the overhead. –One of the alternatives : cache array routing protocol that partitions the URL space among proxies.

18 Overhead of ICP –Not a scalable protocol As the number of proxies increases, the overhead quickly becomes prohibitive. Simulations result shows that ICP incurs considerable overhead even when # proxies is as low as four. –The effort spent on processing ICP is proportional to the total number of cache misses experience by other proxies, instead of proportional to the number of actual remote cache hits.

19 Summary Cache (1/7) 4 Summary Cache Scheme –Each proxy stores a summary of URLs of documents cached at every other proxy. –If the requested document might be stored in other proxies, the proxy sends out requests to the relevant proxies to fetch the document. –Scalable : summaries do not have to be up to date or accurate. Errors –false misses, false hits, remote stale hits

20 Summary Cache (2/7) 4 Impact of Update Delays –Delaying update Delaying update until the percentage of cached documents that are new reaches a threshold. Updating summaries upon regular time intervals. –Simulation results shows that the degradation in total cache hit ratio increases linearly with the update threshold. –False hit ratio is very small though it does increases linearly with the threshold.

21 Summary Cache (3/7) 4 Summary Representation –Summaries need to be stored in the main memory. –Tow naïve summary representation exact-directory –use 16-byte MD5 –consumes too much memory. server-name –generates too many false hits, significantly increases the network traffic.

22 Summary Cache (4/7)

23 Summary Cache (5/7) 4 Bloom Filters –Invented by Burton Bloom in m bits Bit Vector v Element a H 1 (a) =P 1 H 2 (a) =P 2 H 3 (a) =P 3 H 4 (a) =P 4

24 Summary Cache (6/7) –k independent hash functions with a range {1,…,m} –False positive after inserting n keys into a table of size m the probability of a false positive is the right side is minimized for k=ln2 * m / n,

25 Summary Cache (7/7) 4 Bloom Filters as Summaries –8, 16, 32 bits for each document. –4 hash functions. 4 Experiment Result –Bloom filter summaries has virtually the same cache hit ratio as the exact-directory approach. –In terms of total size of inter-proxy network msg, Bloom filter based summaries improve over ICP by 55% to 64%.

26

27

Web Server Workload Characterization : The Search for Invariants

29 Invariants Found in Web Server Workloads (1/3) 4 Success Rate –Success rate for lookups at server = 88% 4 File Types –HTML and image files account for % of requests 4 Mean Transfer Size –Mean transfer size <= 21 kilobytes 4 One Time Referencing –Approximately one-third of the files and bytes accessed in the log are accessed only once in the log.

30 Invariants Found in Web Server Workloads (2/3) 4 Size Distribution –File size distribution is Pareto with 0.40 <  < Concentration of References –10% of the files accessed account for 90% of server requests and 90% of the bytes transferred. 4 Inter-Reference Times –File inter-reference times are exponentially distributed and independent

31 Invariants Found in Web Server Workloads (3/3) 4 Remote Requests –Remote sites account for >= 70% of the accesses to the server, and >= 60% of the bytes transferred. 4 Wide Area Usage –Web servers are accessed by 1000’s of domains, with 10% of the domains accounting for >= 75% of usage.

32 Self-Similarity 4 Recent work has suggested that WWW traffic may be self-similar 4 Moving from the bottom plot to the top plot, bursti ness clearly exists across several different time scales