Comparing, Tracing and Evaluating Caching Servers

1
Comparing, Tracing and Evaluating Caching Servers

• Presenter Yi-Hsiang Lin
• Advisor Dr. Ying-Dar Lin
• High-Speed Network Lab. of CIS, NCTUOct 25, 2001

2
OUTLINE

• Introduction
• Market review
• Squid The open source solution
• Benchmarking squid
• Conclusions
• Reference

3
1. Introduction

• Why caching? Where to cache?
• Major issues in caching (research papers)
• Products/packages survey
• squid (brief review)

4
Why caching? Where to cache?

• Why caching?
• Reduced network bandwidth or offload web server
• Faster Response to End User Increased Network
  Scalability
• Where to cache?
• Client side ? forward caching
• Server side ? reverse caching
• Network ? transparent caching

5
Forward caching
6
Reverse caching
7
Transparent caching
8
Major issues in caching

• About traffic and performanceOrigin Server
  Performance and CharacterizationProxy Cache
  Traffic CharacterizationUser Perception of Web
  PerformanceWeb Client Traffic Characterization
• About cache contentWeb Cache Consistency Web
  Cache Replacement HTTP Inter-cache
  Communication Prefetching
• Other issues and politics

9
Products/packages survey(1/2)
10
Products/packages survey(2/2)
11
Squid is

• a full-featured Web proxy cache
• designed to run on Unix systems
• free, open-source software
• the result of many contributions by unpaid (and
  paid) volunteers
• licensed under the terms of the GNU General
  Public License

12
Squid supports

• Cache content
• 1. proxying and caching of HTTP, FTP, and
  other URL's
• 2. proxying for SSL
• 3. caching of DNS lookups
• Cache mechanism
• 1. cache hierarchies, Cache Digests
• 2. ICP, HTCP, CARP, WCCP
• 3. transparent caching
• 4. HTTP server acceleration
• Extra services extensive access controls, SNMP

13
2. Market review

• Products/packages comparison
• Benchmark survey1. Benchmark items2. Benchmark
  results3. Observations/conclusions

14
Products/packages comparison
15
Benchmark items

• Total price
• Peak Throughput
• Cache Age(hour)
• Response times(Hit/All/Miss)
• Saving(Doc/Time)
• 1000 can buy(hit/sec, req/sec)
• Minutes Till First(Miss/Hit)

16
Benchmark results(1/2)
17
Benchmark results(2/2)
18
Observations/conclusions

• IBM-230Best hit response time ? use gigabit
  NICworst average response time, Saving(DOC),
  cache age ? Peak throughput, less disk
  space
• Lucent-50Best average response time,
  Saving(Time), minutes till First ? low
  throughput, big cacheworst Throughput, 1000
  can buy? high price
• Bigger Memory and High-speed network card are
  important for caching server

19
3. squid The open source solution

• Overview
• Configration example
• Protocols (ICP, HTTP, etc.)
• Program architecture
• Data structures
• Flowcharts

20
Overview

• Squid is derived from the ARPA-funded Harvest
  project
• Squid is a high-performance proxy caching server
  for web clients
• Squid handles all requests in a single,
  non-blocking, I/O-driven process

21
Example(1) TANet proxy
22
Example(2) Transparent
23
Protocols

• Supporting FTP, gopher, and HTTP data objects.
• caches DNS lookups, supports non-blocking DNS
  lookups, and implements negative caching of
  failed requests.
• Squid supports SSL, extensive access controls,
  and full request logging.
• using the lightweight Internet Cache Protocol

24
Data structures
25
Flowcharts
26
Flowcharts
27
4. Benchmarking squid

• Previous test results on squid
• Benchmark methodologies test items, tools, etc.
• Benchmark results

28
Previous test results(1/2)About cache size
29
Previous test results(2/2)About library,
algorithm
30
Benchmark methodologies

• Use polymix-1 of polygraph 2.5.5
• Change request rate(30/60/90/180)
• Install FreeBSD/Linux/Solaris on three disk
• Two cache diskcache size 500x2MB

31
Benchmark flowchart
32
Benchmark results(1/5)Hit response times(msec)
33
Benchmark results(2/5)Miss response times(msec)
34
Benchmark results(3/5)Avg. response times(msec)
35
Benchmark results(4/5)Hit ratio()
36
Benchmark results(5/5)Comparison of three OS
37
5. Conclusions

• Good for caching1. Big memory cache space and
  high-speed network2. Use the best disk access
  methods for the OS
• Best choice of squidFreeBSDDISKD (570 req/sec,
  avg. response time 6.4 sec)
• Squid is good enough for TANet traffic using.

38
Reference(1/4)

• 1 Squid Internet Object Cache
  lthttp//www.squid-cache.orggt
• 2 The FreeBSD Project lthttp//www.freebsd.orggt
• 3 Linuxlthttp//www.linux.orggt, Linux.com A
  Means to World Liberationlthttp//www.linux.comgt
• 4 Solaris product line lthttp//www.sun.com/soft
  ware/solaris/gt
• 5 Web Caching-related Papers and Articles
  lthttp//www.web-cache.com/Writings/papers.htmlgt
• 6 Web Caching Products and Software
  lthttp//www.web-cache.com/products.htmlgt
• 7 D. Wessels and K. Claffy National Laboratory
  for Applied Network Research/UCSD, Internet
  Cache Protocol version 2, RFC 2186, September
  1997
• 8 P. Vixie(ISC), D. Wessels(NLANR), Hyper Text
  Caching Protocol, RFC 2756, January 2000

39
Reference(2/4)

• 9 CARP, Cache Array Routing Protocol. Internet
  Draft draft-vinod-carp-v1-03.txt(expired)
• 10 WCCP, Web Cache Coordination Protocol.
  draft-wilson-wrec-wccp-v2-00.txt(expired)
• 11 Ara Network Technologies lthttp//www.aranetw
  ork.com/eng/products/web.htmgt
• 12 CacheFlow cIQ Content Delivery
  lthttp//www.cacheflow.com/products/index.cfm/gt
• 13 Cisco Cache Engine Series
  lthttp//www.cisco.com/warp/public/cc/pd/cxsr/500/gt
  
• 14 Compaq TaskSmart C-series
  lthttp//www.compaq.com/tasksmart/c4000/index.htmlgt
  
• 15 TMF(The Measurement Factory INC.)
  lthttp//www.measurement-factory.comgt

40
Reference(3/4)

• 16 A. Rousskov and D. Wessels, The Third
  Cache-off. Raw data and independent analysis at
  lthttp//www.measurement-factory.com/results/gt.
• 17 ARPA(Advanced Research Projects Agency) Now
  called Defense Advanced Research Projects Agency
  (DARPA), the U.S. government agency that funded
  the ARPANET
• 18 Harvest Project lthttp//harvest.cs.colorado.
  edu/gt
• 19 ????????????????lthttp//procy.ntu.edu.twgt???
  ?????????lthttp//proxy.ncu.edu.twgt?????????lthttp/
  /proxy.nctu.edu.twgt??????????lthttp//proxy.nhctc.e
  du.twgt???????????lthttp//proxy.chu.edu.twgt????????
  ???lthttp//proxy.ccu.edu.twgt???????????lt
  http//proxy.ncku.edu.tw/gt??????????lt
  http//proxy.nsysu.edu.twgt????

41
Reference(4/4)

• 20 The CodeRed Worm CERTlthttp//www.cert.orggt
  Incident Note IN-2001-09,13,19,23
• 21 Squid Programmers Guide, Chapter 2 Overview
  of Squid Components, lthttp//www.squid-cache.org/D
  oc/Prog-Guide/prog-guide-2.htmlgt
• 22 Squid Programmers Guide, Chapter 4 Flow of
  a Typical Request, lthttp//www.squid-cache.org/Doc
  /Prog-Guide/prog-guide-4.htmlgt
• 23 Web PolyGraph lthttp//www.web-polygraph.org/
  gt
• 24 ?????????? ????????? lthttp//proxy.ntu.edu.t
  w/ntu-proxy/stats.htmlgt
• 25 ???????? ??????? lthttp//proxy.nctu.edu.tw/p
  roxy/log.htmlgt
• 26 ????????????? lthttp//proxy.nsysu.edu.twgt