The user asked about ControlUp’s Protocol Bandwidth Limit measurement and how it is calculated. It was suggested that this measurement may change due to a CPU Readiness problem in VMware, as the guest OS has CPU demand but the host cannot satisfy the demand. It was also suggested to check the CO-stop (%CSTP), which is shown as CPU Excessive SMP Use in ControlUp on the VM level, as well as check for any snapshots present on the virtual machine that could increase %CSTP values and reduce virtual-machine performance.
Read the entire ‘Measuring and Explaining ControlUp’s Protocol Bandwidth Limit’ thread below:
How does Control Up calculate the current "Protocol Bandwidth Limit". It is a changing number. We are having an issue with CPU Readiness which means out vCPU/pCPU ratio is way too high. I believe that my changing or dropping "Protocol Bandwidth Limit" is really only a symptom of my vCPU Scheduliing problem in VMware. The time slicing is making it appear that "Protocol Bandwidth Limit" is dropping. I do not believe that avilable bandwidth is changing at least not as dramatically as it appears in the Control Up monitoring numbers. Can anyone please explain how that number is caluclated or monitored and if my theory explains this behavior?
I believe we get it from VMware and it’s not a calculation of ourselves But I already moved to the couch so can’t look it up.
Hey Richard, it is taken from the remoting protocol. I don’t know specifics but I believe @member mentioned that typically it is done by sending 2 packets of fixed size and calculating how long it takes to transfer the packets
You can see the same bandwidth numbers in perfmon counters. Which ones are used depends on the protocol thoughDo you see a Processor Queue on the affected boxes?
Does anyone know if a CPU Readiness problem will cause the Protocol Bandwidth Limit to "seem" like it is dropping?
I know I have a CPU Ready problem for sure, but I have other engineers that believe it might be networking or something else.Take this with a grain of salt, because I am guessing, I don’t actually know.
But lets take Citrix for example. As far as I’m aware, the virtual desktop agent running on the VM is part of calculating the bandwidth limit.
If we can assume that, then the machine suffering from CPU ready would affect that. As the guest OS has CPU demand but the host cannot satisfy the demand. Effectively not completing work in a timely manner.
I’d also recommend looking at CO-stop (%CSTP).
> Co-Stop – Amount of time a SMP virtual machine was ready to run, but incurred delay due to co-vCPU scheduling contention
Taken from https://vmware2112.wordpress.com/2012/07/05/what-does-the-cpu-co-stop-counter-measure-cstp/CO-stop (%CSTP) is shown as CPU Excessive SMP use in ControlUp by the way on the VM levelThanks Dennis. I appreciate your efforts!
No problem 🙂
One more thing and I’ll stop rambling.
> One or more snapshots are present on the virtual machine: As storage I/O for snapshots grows, Co-Stop (%CSTP) values for a VM with multiple vCPUs can increase as the vCPUs wait for I/O completion. To reduce the high %CSTP values and improve virtual-machine performance, consolidate any snapshots onto the main virtual disk.
Continue reading and comment on the thread ‘How does ControlUp calculate the current “Protocol Bandwidth Limit”?’. Not a member? Join Here!
Categories: All Archives