Instance Performance

[orlando_f]

Hi Guru's,

Does anyone know what the performance capabilities are of a ServiceNow instance? i.e. How many concurrent users can login, log tickets at any given time? Is there any information out there regarding this

[Valor1 · Accepted solution]

What every ServiceNow production instance gets:

• Two UI Nodes (all figures below are per node)
• 2GB RAM for each node
• Sixteen User semaphores available on each node (total of 32 concurrent transactions, or 32 people hitting the "Submit" button all at the exact same moment)
• Eight additional integration / SOAP semaphores (per node, total of 16), to ensure that user-land transactions are not impacted by integration and ODBC traffic.
• Eight background schedulers for (16 total) email processing, scheduled jobs, workflow timers, Discovery, etc.
• Thirty-two DB connections (64 total)

I believe I worked on the same project as mguy and that customer had double the above stats.

In other words, you are fine with the default settings unless you are in the top 5% of ServiceNow customers, size-wise. The default setup will handle probably 8k+ process (IT) users and 200,000+ end users.

[divvi_vamsi]

I see three UI nodes for my instance, so is this something unusual or Servicenow had customized as such to distribute the load on the instance due to higher transaction rate?

What is the queue depth for each semaphore?

Thanks
Divya

[Valor1]

Given that SN is a cloud service, they manage the # of nodes, db sizing and capacity, etc.

I'd say that a typical   node could support roughly 500-700 concurrent sessions. You should (hopefully) never see very high numbers in the "queue depth", and should usually be 0 when you look at it. If you see the queue back up, it means that there are likely a number of long-running transactions that are backing things up.

Queue Depth really means that there are N transactions waiting for a semaphore. You can think of a Semaphore as a lane on a highway, and the typical node has 16 lanes on it (supporting sixteen cars / requests at a single moment). Most transactions are very short lived (300ms or less).