ViPR Load Testing?

Hi,

Yes, ViPR engineering has put in a big effort to ensure performance is maintained & improved for 2.0.  The following is an overview of the efforts made to ensure this.

1. ViPR  has been subjected with an automated workload from UI with increasing number of concurrent user sessions to measure response time, throughout statistics and resource utilization at system and process level.  The following has been carried out:

1. Start with 10 concurrent users and go up to 100, pushing this up to 150-200 concurrent users.
2. Workload is sustain for 8 to 16 hours a day and repeated for few days to observe system health under load and response time degradation.

2. Subject ViPR with increasing workload using direct API method & execute a mixed workload (various north and south-bound operation on block and file) by increasing number of threads. Monitor system health and measure throughout and response time of various operations

1. Increase threads (concurrency of operations) from 20 to 300 in steps
2. Workload is sustained for 4 to 16 hours and repeated for few days

3.  Increase the size of the inventory

1. Create 0.5M, 1M, 1.5M devices (file and block) using various array simulator
2. At every ‘size’, subject ViPR with UI and API workload verifying that ViPR is able deliver decent response time and throughout at various scale level.

4. Longevity under load test by subjecting ViPR with concurrent load from UI and API.

1. Setup a large ViPR with 1M+ devices
2. Subject ViPR with concurrent workload from UI and API exercising most common north and south-bound operations. Monitor ViPR for 10-14 days at system and process level. Measure response time and throughout of operations.

Thanks,

Dee