Download presentation
Presentation is loading. Please wait.
1
Measure Web Performance
The performance of a web system perceived by end users is decided by Browser performance Network path performance Server performance Measuring a web system with multiple geographically separated servers can be quite different from measuring a web system with a single server. Performance of a web server observed from clients at different network can be quite different. 11/6/2018 C. Edward Chow
2
Web Performance Metrics
Server connection rate: connections/sec (x3.15) SPECweb96 use peak “operations” rate instead of avg. rate Server error rate: err/sec Server thruput: Mbit/sec (x2.9) Little's Load Factor: (x1.66) Average response time: sec (x0.529) Error Level: % Average client thruput: Mbit/sec (x1.789) Average Std Dev Minimum Maximum Connect time (sec) Response time (sec) 11/6/2018 C. Edward Chow
3
Web Benchmarks Patrick Kilelea’s web page on WebReview, WebStone, original developed by SGI, obtained by Mindcraft. 3/95, First widely used web benchmarks. Freeware 2.5 release include plain HTML, CGI, NSAPI, ISPAI SPECweb96 Not free. Client software in C and Perl. detailed run rules: warmup_time(300s), run time (600s). with required mix HTTP request & sizes. The load is modeled "typical" ISP. not originate with a vendor of web server hardware or software less bias. Submit results 11/6/2018 C. Edward Chow
4
WebStone 2.5 Webmaster creates a load on a Web server by running one or more webclient processes on one or more client computers 11/6/2018 C. Edward Chow
5
11/6/2018 WebStone Setup In conf directory of webmaster, specify the Fileset: the composition of web documents In conf directory, edit the testbed file: specify the Server name, port #, and document directory List of client machines, # of processes in each machines Login and password for accessing Client machines (RCP for copy web document, RSH for start and coordinate the web clients, and receive performance results). RCP and RSH requires the destination machine to be setup for remote access. The .rhosts file of the corresponding accounts needs to contain webmaster machine name and its login name. Use RCP for copy web document, RSH for start and coordinate the web client requests 11/6/2018 C. Edward Chow
6
WebStone Script After installing WebStone2.5, run the following scripts ./webstone –setup This configures scripts and compiles programs. ./webstone –genfiles This generates web documents in Datafiles directory ./webstone This script runs the benchmarks, rcp documents to the webclients and webserver. (You may have to do it manually.) prints results on the console, and saves in bin/runs directory. Each iteration result is saved with the timestamp as filename 11/6/2018 C. Edward Chow
7
Filelist # @(#)filelist.standard 1.3
# Filelist for WebStone 2.5 Standard Run Rules, same as filelist.sample /file500.html #500 /file5k.html #5125 /file50k.html #51250 /file500k.html #512500 /file5m.html # This filelist consists of 5 different files. The number following the filename is the weight of this file in the distribution. All the weights are summed together and the frequency of each file is the weight of that file over the total weights. For example, in this fileset the weights add up to So the the file500k.html page will occur 350 out of 1000 times, and the file5m.html will occur once every 1000 pages. 11/6/2018 C. Edward Chow
8
WebStone Benchmark Results
WebStone2.5]$ ./webstone Clients: gandalf Estimated run time: 1 hours 39 minutes ***** Iteration 1, Total clients 20 ********** Sun Feb 28 04:46:23 MST 1999 Sun Feb 28 04:46:26 MST 1999 /home/guest/webstone/WebStone2.5/bin/webmaster -v -W -C /tmp/webclient -f /home/guest/webstone/WebStone2.5/bin/runs/990228_0446/config -t 10 -u /home/guest/webstone/WebStone2.5/conf/filelist -p S WebMaster name = bilbo.uccs.edu Client 0: gandalf # Processes: 20 Webserver: frodo Webmaster: bilbo.uccs.edu:1026 Waiting for READY from 20 clients All READYs received Sending GO to all clients All clients started at Sun Feb 28 04:46: 11/6/2018 C. Edward Chow
9
When server is not started
No connects from client 0. Is the Web server running? No connects from client 1. Is the Web server running? No connects from client 2. Is the Web server running? No connects from client 3. Is the Web server running? No connects from client 4. Is the Web server running? No connects from client 5. Is the Web server running? No connects from client 6. Is the Web server running? No connects from client 7. Is the Web server running? No connects from client 8. Is the Web server running? No connects from client 9. Is the Web server running? No connects from client 10. Is the Web server running? 11/6/2018 C. Edward Chow
10
Iteration with 40 clients, Page 0 Single Web Server Process (-X)
Waiting for clients completion... Reading results All clients ended at Sun Feb 28 05:17: ======================================================== Page # 0 Total number of times page was hit 6368 6368 connection(s) to server, 0 errors Average Std Dev Minimum Maximum Connect time (sec) Response time (sec) Response size (bytes) Body size (bytes) body bytes moved header bytes moved = total Page size 500 11/6/2018 C. Edward Chow
11
Iteration with 40 clients, Page 1
Total number of times page was hit 8401 8401 connection(s) to server, 0 errors Average Std Dev Minimum Maximum Connect time (sec) Response time (sec) Response size (bytes) Body size (bytes) body bytes moved header bytes moved = total Page size 5120 11/6/2018 C. Edward Chow
12
Iteration with 40 clients, Page 2
Total number of times page was hit 8401 8401 connection(s) to server, 0 errors Average Std Dev Minimum Maximum Connect time (sec) Response time (sec) Response size (bytes) Body size (bytes) body bytes moved header bytes moved = total Page size 5120 11/6/2018 C. Edward Chow
13
Iteration with 40 clients, Page 3
Total number of times page was hit 154 154 connection(s) to server, 0 errors Average Std Dev Minimum Maximum Connect time (sec) Response time (sec) Response size (bytes) Body size (bytes) body bytes moved header bytes moved = total Page size 11/6/2018 C. Edward Chow
14
Iteration with 40 clients, Page 4
Total number of times page was hit 16 16 connection(s) to server, 0 errors Average Std Dev Minimum Maximum Connect time (sec) Response time (sec) Response size (bytes) Body size (bytes) body bytes moved header bytes moved = total Page size 11/6/2018 C. Edward Chow
15
WEBSTONE 2.5b3 results: Single Server Process (-X option)
Total number of clients: Test time: minutes Server connection rate: connections/sec Server error rate: err/sec Server thruput: Mbit/sec Little's Load Factor: Average response time: sec Error Level: % Average client thruput: Mbit/sec Sum of client response times: sec Total number of pages read: 17271 connection(s) to server, errors 11/6/2018 C. Edward Chow
16
WEBSTONE 2.5b3 results with Dynamic Spare Child Servers
Total number of clients: Test time: minutes Server connection rate: connections/sec (x3.15) Server error rate: err/sec Server thruput: Mbit/sec (x2.9) Little's Load Factor: (x1.66) Average response time: sec (x0.529) Error Level: % Average client thruput: Mbit/sec (x1.789) Sum of client response times: sec (x1.667) Total number of pages read: (x3.147) 54360 connection(s) to server, 0 errors 11/6/2018 C. Edward Chow
17
WEBSTONE 2.5b3 results: Single Server Process (-X option)
Average Std Dev Minimum Maximum Connect time (sec) Response time (sec) Response size (bytes) Body size (bytes) body bytes moved header bytes moved = total 11/6/2018 C. Edward Chow
18
WEBSTONE 2.5b3 results with Dynamic Spare Child Servers
Average Std Dev Minimum Maximum Connect time (sec) Response time (sec) Response size (bytes) Body size (bytes) body bytes moved header bytes moved = total 11/6/2018 C. Edward Chow
19
Apache Serve Load Using /status with Dynamic Spare Child Servers
Current Time: Sunday, 28-Feb :55:03 MST Restart Time: Sunday, 28-Feb :07:33 MST Server uptime: 47 minutes 30 seconds Total accesses: Total Traffic: 4.1 GB CPU Usage: u s cu0 cs % CPU load 92.5 requests/sec MB/second kB/request 167 requests currently being processed, 0 idle servers WWWWWWWWWWWWWWWWWWLWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWL WWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWW WWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWWW.WS Scoreboard Key: "_" Waiting for Connection, "S" Starting up, "R" Reading Request, "W" Sending Reply, "K" Keepalive (read), "D" DNS Lookup, "L" Logging, "G" Gracefully finishing, "." Open slot with no current process 11/6/2018 C. Edward Chow
20
Impact of Increasing # of Web Client Processes with single server process
11/6/2018 C. Edward Chow
21
Impact of Increasing # of Web Client Processes with dynamic spare server processes
11/6/2018 C. Edward Chow
22
Compare Web Sites Performance
11/6/2018 C. Edward Chow
23
Homework #6 WebStone Benchmark
cp –r /home/guest/webstone/WebStone2.5 /home/<login> cp –r /home/guest/sites/site.webstone /home/<login>/sites Edit the WebStone2.5/conf/testbed file and replacing MAXCLIENTS=“60” original value is “100” TIMEPERRUN=“3” original value is “10” PORTNO=“<your port#>” Use bilbo as webmaster, gandalf as webclient. Start your web server without –X. Run ./webstone (no need to do setup, genfiles) Plot the server connection rate, server throughput, and average response time , vs the clinet numbers (from 20 to 60). Test apache web servers on frodo, jen, redcloud, walrus. On jen and walrus, use port 80. 11/6/2018 C. Edward Chow
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.