A proposed bench test for ArcGIS Server
ArcGIS Server can be demanding of resources <ul><li>Load test February, 2010 </li></ul><ul><li>11 simultaneous users </li>...
CPU use <ul><li>4 CPU cores </li></ul><ul><li>AMD Opteron 275 </li></ul>
Memory use <ul><li>5 Gb RAM </li></ul>
Network <ul><li>1 Gb/sec fibre channel </li></ul>
How to optimize What is the goal of the server?
Our business needs <ul><li>Replace ArcIMS applications with ArcGIS Server </li></ul><ul><li>Replace 16 existing W2K3/AMD 2...
We chose to focus on throughput and performance <ul><li>We have, by some definitions: </li></ul><ul><ul><li>39 application...
Define parameters
How to define “capacity” <ul><li>We defined capacity as the maximum number of map requests a host or system can serve per ...
How to define performance <ul><li>We defined “performance” as the average time it takes a server to respond to a map reque...
Set targets
Target parameters <ul><li>We estimated that our ArcIMS environment was capable of producing about 40 maps / second </li></...
Load testing methodology <ul><li>Client </li></ul><ul><li>Map service </li></ul>
Load testing client <ul><li>JavaScript application (use browser as client) </li></ul><ul><li>“ export_map” requests to RES...
Map service <ul><li>Main map service for one of our major, planned AGS applications </li></ul><ul><li>Large number of laye...
ArcGIS Server capacity defined by CPU
SPEC <ul><li>Standard Performance Evaluation Corporation </li></ul><ul><li>www.spec.org </li></ul><ul><li>Hardware provide...
Proposed standard bench test
Test map requests—why? <ul><li>Our tests have shown that for ArcGIS Server, at least, mapping remains the most resource in...
Proposed vector map service <ul><li>Multiple data layers </li></ul><ul><li>Nationwide coverage </li></ul><ul><li>Variety o...
Proposed image map service--why <ul><li>In ArcGIS Server, at least, rendering dynamic imagery remains most expensive opera...
Client factors <ul><li>Web client </li></ul><ul><li>Client can have significant effect on performance </li></ul><ul><li>Ra...
Ideally, a test could be standardized to apply to multiple platforms (not just AGS)
Your ideas? Erik Sandin Wisconsin Dept. Natural Resources [email_address] 608-264-8917
Upcoming SlideShare
Loading in …5
×

Proposed bench test for gis servers

792 views

Published on

ArcGIS Server, WiDNR, performance, capacity, benchmark

Published in: Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
792
On SlideShare
0
From Embeds
0
Number of Embeds
11
Actions
Shares
0
Downloads
18
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Proposed bench test for gis servers

  1. 1. A proposed bench test for ArcGIS Server
  2. 2. ArcGIS Server can be demanding of resources <ul><li>Load test February, 2010 </li></ul><ul><li>11 simultaneous users </li></ul><ul><li>1 hour </li></ul>
  3. 3. CPU use <ul><li>4 CPU cores </li></ul><ul><li>AMD Opteron 275 </li></ul>
  4. 4. Memory use <ul><li>5 Gb RAM </li></ul>
  5. 5. Network <ul><li>1 Gb/sec fibre channel </li></ul>
  6. 6. How to optimize What is the goal of the server?
  7. 7. Our business needs <ul><li>Replace ArcIMS applications with ArcGIS Server </li></ul><ul><li>Replace 16 existing W2K3/AMD 275 blades with ? </li></ul>
  8. 8. We chose to focus on throughput and performance <ul><li>We have, by some definitions: </li></ul><ul><ul><li>39 applications </li></ul></ul><ul><ul><li>14 million hits / month </li></ul></ul><ul><ul><li>40,000 users / month </li></ul></ul><ul><li>We knew ArcGIS Server was resource demanding </li></ul><ul><li>We knew ArcGIS Server could be slow compared to ArcIMS </li></ul>
  9. 9. Define parameters
  10. 10. How to define “capacity” <ul><li>We defined capacity as the maximum number of map requests a host or system can serve per second (“maps / second”) </li></ul>
  11. 11. How to define performance <ul><li>We defined “performance” as the average time it takes a server to respond to a map request (milliseconds or seconds) </li></ul>
  12. 12. Set targets
  13. 13. Target parameters <ul><li>We estimated that our ArcIMS environment was capable of producing about 40 maps / second </li></ul><ul><li>We estimated that our average ArcIMS response time was 1 second (1000 ms) </li></ul>
  14. 14. Load testing methodology <ul><li>Client </li></ul><ul><li>Map service </li></ul>
  15. 15. Load testing client <ul><li>JavaScript application (use browser as client) </li></ul><ul><li>“ export_map” requests to REST API </li></ul><ul><li>Configurable parameters </li></ul><ul><ul><li>Host </li></ul></ul><ul><ul><li>Map services </li></ul></ul><ul><ul><li>Number of asynchronous requests </li></ul></ul><ul><li>Randomizable parameters </li></ul><ul><ul><li>Map services </li></ul></ul><ul><ul><li>extent </li></ul></ul>
  16. 16. Map service <ul><li>Main map service for one of our major, planned AGS applications </li></ul><ul><li>Large number of layers </li></ul><ul><li>Geometrically complex layers </li></ul><ul><li>Contains imagery </li></ul>
  17. 17. ArcGIS Server capacity defined by CPU
  18. 18. SPEC <ul><li>Standard Performance Evaluation Corporation </li></ul><ul><li>www.spec.org </li></ul><ul><li>Hardware providers conduct standardized tests and report results </li></ul><ul><li>Spec_int_rate2006 </li></ul>
  19. 19. Proposed standard bench test
  20. 20. Test map requests—why? <ul><li>Our tests have shown that for ArcGIS Server, at least, mapping remains the most resource intensive operation </li></ul><ul><li>OPINION: I suspect it is why people come to your site and that almost every request of another type will involve mapping </li></ul>
  21. 21. Proposed vector map service <ul><li>Multiple data layers </li></ul><ul><li>Nationwide coverage </li></ul><ul><li>Variety of scales </li></ul><ul><li>Geometrically complex </li></ul><ul><li>Relatively uniform complexity across the country </li></ul>
  22. 22. Proposed image map service--why <ul><li>In ArcGIS Server, at least, rendering dynamic imagery remains most expensive operation. </li></ul><ul><li>Can choose to avoid this by using cached imagery </li></ul><ul><li>Make image test separate </li></ul>
  23. 23. Client factors <ul><li>Web client </li></ul><ul><li>Client can have significant effect on performance </li></ul><ul><li>Randomizable / separate tests? </li></ul>
  24. 24. Ideally, a test could be standardized to apply to multiple platforms (not just AGS)
  25. 25. Your ideas? Erik Sandin Wisconsin Dept. Natural Resources [email_address] 608-264-8917

×