AnsweredAssumed Answered

Understanding Hosted Service Performance

Question asked by donovanartz on Aug 16, 2020
Latest reply on Sep 16, 2020 by donovanartz

We are observing *errors* or a *delay* intermittently when queries are made using the feature layer API through our ArcGIS server.




It appears that the ArcGIS server intermittently needs to "cold start" our hosted services -- as if it was a traditional service that wasn't running.



How are hosted services implemented?  Are they also served from arcsoc processes?  If we add cores to our hosting server, will more arcsoc processes automatically be generated?  Is there any setting we can use to spawn more arcsoc processes (and take more memory) to eliminate "cold starts" of hosted services?


Currently, we've got a single ArcGIS hosting server (4 cores, 16GB) federated with our Portal.


We have about 30 active feature services backed by the Spatiotemporal Big Data Store (registered with the hosting server).  Each of these feature services receives *live* data from our GeoEvent server (writing via the API), and may be powering any number of operations dashboards (reading via the API).


When we load an operations dashboard, we can watch the API calls to the feature services in our browser developer console -- it is here we can see delays and errors (e.g., "500").


Thanks for any insights!