Hi Sunil, Answering your questions in order:
1. We don’t have fixed performance figures for Project Clearwater (as opposed to our Clearwater Core product) – we just provide the tools to let the community do their own performance testing and analysis. This is partly because hardware varies so much (CPU architecture, speed, etc.) that it’s difficult to give precise figures. 2. When using the run_stress script, each emulated subscriber generates 1.3 calls/second (50% incoming, 50% outgoing). This can be increased with the ‘multiplier’ option (e.g. --multiplier 10 means 13 calls/second), but this also increases the registration rate – to increase the calls per second independently of registration rate, you’d need to edit the script. 3. Call success depends on the stress node and all of Sprout, Dime and Vellum. http://www.projectclearwater.org/technical/call-flows/ has more information on how they all get involved in call flows. 4. It’s difficult to predict exactly where the bottlenecks would be, but in general, the Sprout node does the most work (and so either it will be the bottleneck or you’ll use more Sprout nodes than other nodes). Different configurations will affect this a lot, though – if using homestead-prov rather than a HSS, the Cassandra database on Vellum will be doing more work, and if using Rf billing, the Dime node will be doing more work. 5. I don’t think we’ve tested a maximum value for --multiplier. My guess would be that it starts to break down if calls/registrations are happening faster than once per second per subscriber – so 100 might be OK but 2000 probably wouldn’t be. 6. Going up to 2 vCPUs and 4GB RAM on each node would be sensible, but past that point we’d recommend increasing the number of nodes instead – Clearwater is primarily architected to scale that way. 7. Our default throttling options should be well-tuned for most workloads – you shouldn’t need to change them. Hope that helps! Best, Rob From: Clearwater [mailto:[email protected]] On Behalf Of Sunil Kumar Sent: 29 March 2018 15:06 To: [email protected] Subject: [Project Clearwater] Stress Testing Queries Hi all, I want to increase the # of calls per seconds. Please help me out. I have few doubt Please reply: 1) How many max calls per sec can be make using 1 vCPU and 2 GB RAM, 10 GB HDD, manual setup with stress test. 2) How can i increase the no. of calls per sec, Is there any specific parameter for that or Do I need to change in script (run_stress) or anywhere else please let me know. 3) I am using the stress node for making calls, I want to ask, On which nodes the calls or stress test is depends other than sprout and stress node. I have created the subscribers on vellum node. 4) # of calls is limited by stress node or sprout node or any other node? 5) What is the maximum limitation of --multiplier, i have checked up to 20 only? 6) Should I increase the CPUs, RAM and HDD on single node or Do I create new node for larger deployments (multiple nodes), which is better way? 7) I have not use any throttling options yet. Do I need to use them, if yes please let me know which for increasing the CPS, how can i use (just mentioned in shared_config?) Please reply, i am waiting for it eagerly. Thanks in advance. Regards, suni
_______________________________________________ Clearwater mailing list [email protected] http://lists.projectclearwater.org/mailman/listinfo/clearwater_lists.projectclearwater.org
