Amazon EC2 with Simple Storage Service S3

Copyright © 2014 Open Geospatial Consortium. 15 Figure 2 - Diagram illustrating concurrent user access to OGC-compliant map servers using an Amazon EC2 with an S3 network attached storage configuration

5.1.7 Performance results

The section below presents results from performance and scalability tests using the following two system configurations: a Amazon EC2 with ABS, and b Amazon EC2 with S3. All tests were conducted using a single Amazon computer instance c3.8xlarge for serving maps and all map requests were generated by the stress tester tool deployed on multiple computer instances m3.2xlarge. Each map request was randomly generated and no map results were cached. Amazon EC2 with EBS The performance tests conducted with this configuration are listed below: ฀ Comparative average response time for serving maps from WMS and WMTS using a single “c3.8xlarge” computer instance. ฀ Number of maps per seconds satisfied by the WMS and WMTS servers. ฀ Average response time for delivering maps from a WMTS service to a large number of concurrent users. Amazon EC2 with S3 The performance tests conducted with this configuration are listed below: 16 Copyright © 2014 Open Geospatial Consortium. ฀ Average response time for serving maps from a WMTS server using an EBS direct attached storage VS an S3 network attached storage service. ฀ Number of satisfied requests per second from a WMTS server using an EBS direct attached storage VS an S3 network attached storage service. ฀ Results from the largest test we performed using EC2 and S3.

5.1.7.1 Results from Amazon EC2 with Elastic Block Storage EBS

The first observation that can be made from the results in Figure 3 is that the WMTS service scales a lot better than the WMS does, as was expected. If we establish an acceptable response time for returning a map at 2 seconds, results in Figure 3 below indicate that a single “c3.8xlarge” virtual computer can easily serve 450 concurrent users with a good response time while a WMTS service can serve maps to 1,700 concurrent users. This makes the WMTS service almost 4 times more efficient at returning maps from imagery than the WMS. Figure 3 – Comparative average response time for serving maps from WMS and WMTS using a single “c3.8xlarge” computer instance