Server Statistics between Shared and Dedicated Instance

[yockee]

I have ArcGIS Enterprise 11.0 with components consisting of Portal (hosted on AWS), Enterprise Geodatabase (hosted locally on Postgres) . 

I have just a handfull of services that are being used. Some of them are used on daily basis and heavy edits too.

Here are the few services that I have :

1. Map Services. The biggest one contains building layer in polygon. There are 50K objects here. Some other 7 map services contains relatively really small records (around 100 records in total). These 8 map services are served using visibility range (visible on particular zoom scale). 

2. Feature Service. containing point that their attributes will be updated. There are 50K objects. This is also served using visibility range.   

But, when I look to its statistics, I am a little bit confused because of how the graph of Service Maximum Running Instances looks like.

Here is the image of server statistics:

The "Building - Polygon" is acting as background. It is a Map Service which I (accidentally) set the instance to Dedicate (min 3, max 7).

The "data entry - point" is where the actual transaction happenning.    It is a Feature Service where users accessing data via Field Map and Arcgis Pro.

The questions are:

1. Why the Service Maximum Running Instances of "data entry - point" is flat 0, although its Total Request and Maximum Response Time are always higher ? Is it normal ? 

2. In my case, what is the implication of having Shared Instance for the WFS service to the data being collected ?

3. Shouldn't it be the WFS hosted using Dedicated Instance ? I feel weird now.. we have been thought to use Dedicated Instance for WFS services. The system feels more stable now, only crash once for these 4 weeks running. When I had WFS hosted in a Dedicated Instance, I experienced more crashes...

In the previous experience, in another service in the very same server, I used to set the WFS layer to a Dedicated Instance. I tried some combinations between Max Number of instances and Heap Memory. The system *eventually crash* every some days  when I have 8 - 9 Dedicated Instances and 256 - 1024 MB of Heap Memory. After a few combinations, I have settled with Max Dedicated Instance of 7 and 128 MB of Heap Memory.

Need any advice.

Thanks

[TravisSaladino]

ok first off don't split up ArcGIS Enterprise like you you have. It either all should be in the cloud or it all should be on your local network. It's kind of like keeping all your cash money at your home. And then when you need to use it at the store to buy something, you have to run home to get your money once you've been asked for it by the cashier, and then running back to the store to pay. 

Ok with that out of the way 🙂

Your feature service is using the shared instance pool. There are 0 dedicated instances for that service. That is why the Max number of instances for that service is reported as 0. The requests to that service are being handled by the DynamicMappingHost map service ArcSOCs. By the way at 11.5 you can create a statistics report for the DynamicMappingHost service now... it's under the System folder. 

At https://enterprise.arcgis.com/en/server/latest/administer/windows/configure-service-instance-setting... we provide some guidance on when to use the shared instance pool. Namely, we suggest to use shared instance for "services that are infrequently used." Looking at the statistics report you shared, I would suggest this is a very frequently used service that should have dedicated instances. 

Also from the reports you shared, it seems at high load on the feature service there are some users that are seeing very poor performance with max response times between 5 to 25 seconds. That could be due to the network latency between ArcGIS Server (although you only said the portal in in the cloud, I assume ArcGIS Server is there too) and the RDBMS and the high load on the shared instance pool. 

As for the crashing with dedicated instances for the feature service. That's a different matter.

I hope that helps.