Anyone do this to level load the users?
We typically have 116 users at peak time, about 50/50 on Default/MES.
We currently have multiple appservers in Progress and I point different groups to certain Appservers.
We are starting a new implementation of EPICOR 10.1.5XX and wanted to see if anyone else does this?
I’m guessing most people are goign to want to know more about your hardware - servers, storage, network, etc - before they commit to any real suggestion. But my initial gut reaction is that even with 55 MES users and 55 regular users, having multiple app servers might be overkill (assuming there aren’t issues in the aforementioned hardware). Isn’t there a graph somewhere they like to show at Insights and other Epicor events where they supposedly stress test with hundreds - maybe thousands? - of users on one app server in E10 and it still performs better than a load balanced E9 setup? I might be remembering the wrong thing.
Sorry - meant to add that we currently run 10.0, and plan to run 10.1 with only one application server (VM). But we have about 60-80 Epicor users at peak usage, same split on MES vs. Regular client. But we also have our SQL DB Server storing the DB on a Intel P3700 PCIe SSD card.
For whatever that is worth.
Do you happen to know your CPU and Memory Allocation?
I understand that every network and config is different, but I rather than over commit resources than under commit.
Currently have: ESXi-5.5
Poweredge R620
20 CPUS x 3.0GHz
196GB RAM
Nimble Storage
AND
Currently have: ESXi-5.5
Poweredge R720
12 CPUS x 2.5GHz
196GB RAM
Nimble Storage
PLan is likely the following:
1 VM - 2012 R2 Std with SQL 2014
8 vCPU and 96 GB
1 VM - 2012 R2 Std
8 VCPU and 64GB
I ran a couple of dotbomb sites and have baby sat the SaaS implementation enough times to have an opinion or two. 
First, separate fail over needs from performance needs. Are you worried about hardware failure and need to fall over to other hardware? Do you want to ensure an app server crash does not take out all users?
Next, make sure you are on 10.1. There was a ton of farm improvements in 10.1 in the framework.
Third, measure. What reports / processes are typical for you? Does the processing of reports or MRP or whatever suck up all the available ram on the app pool / app server? This will likely have you isolate client app server(s) vs task servers.
Then you get into other considerations -
How are you going to load level the traffic to each ‘farm cluster’ - client vs task is a common split.
An F5 in front of each? Windows ARR?
Do you need to consider external access over https? Then the network binding of HttpsOffload might be of interest to let the F5 handle the ssl and free your app server to just handle client requests.
Once that is all on the whiteboard then you can determine what kind of hardware is needed for each node. Is the node hosting a VM? What’s in the VM? Just a single app server? Some combination of app servers functions?
We are actively monitoring the docker ecosystem and exploring possibilities there - it’s obviously the vNext of VMs at this point and we want to play happy with customers data centers as much as we want to be your data center in the cloud.
btw - I completely ignored identifying the needs for dev / test / live. Put that on the drawing board as well