The CreatorCon Call for Content is officially open! Get started here.

doug_schulze
ServiceNow Employee
ServiceNow Employee

Had a request from a great customer in regards to the resources and impact Discovery will have within their environment. Thought it would be great to share because I know they are probably not the only ones thinking about this as well....


Q: How much RAM per instance when multiple instances are on the same hardware?

A: We define how much ram each MidServer instance will use in our wrapper.conf file where it is default to be 512mb. With standard application overhead you can expect to need that much memory available for each instance you want to run on one piece of hardware.

Q: How many items can safely be discovered at a time, per instance, on the same hardware before it's too much?

A: The number of items we can 'safely' discovery is unlimited. What is the variable is the time it takes to discovery the 'X' amount of targets. This is dictated by the number of threads available for the MidServer application to run at one time. We currently set the default number of threads to 25 and is able to be modified up or down to suit your needs. This along with being able to modify the memory available to the application can help tune how fast you want to discover the 'X' number of devices in your environment. So if you think 3 hours to discovery 2000 devices is too slow then you can increase the amount of threads (and memory available to the application) to get the task done quicker.

Q: Will the application actually use multiple processor cores?

A: Yes it will... The beauty of Java

Q: What kind of memory and processor spikes will occur as the application is running?
(Speaks to needing it's own server or putting it on existing servers alongside other applications.)

A: The only memory spike we have come across has been with discovering the big edge routers and the spike was less then 50Mb. For CPU spikes we have identified that gathering the installed software on windows machines can take 1/2 the CPU on the (mid-server host) for a 4-5 second period. We have proven this to be completely windows related and validated to happen whether our application asks the question or any other remote query of the same request.

1 Comment