The magic is quite simple: RAM, RAM, RAM.
We are running the productive application with roughly the same No of user:
2 Application Server (one for PAD and WEB, the second one for Exchange connector and interface, Geo Location and ERP interfaces (Flatfile import with mmim module) via scheduled tasks at night.
We are currently migrating to 2k16 R2 Win Server (4 CPU, 16 GB RAM for Application and 4CPU 24GB RAM for DB)
Though our DB Server is only dedicated to CRM. so only 1 DB instance, but Dev, PreProd, UAT and Prod DB and all Designer DB in the same instance - so different from a DB cluster....
A significant impact to the system performance is also giving via configuration of the Virus scanner installed on the server! (Especially the Appl. Server)
We have been thinking about a load balancer, but skipped it due to the fact that the system was stable and fast enough.
Regarding Enterprise - We are, but I am not able to tell the differences as we started with Enterprise right from the beginning!
Can I ask you an another question.
What is de performance of you application and DB server doing when multiple iPad are running the full sync at the same time?
What we are seeing is that our DB server (dedicated CRM.prod and running 8core and 16Mb) is taking 100% load and that web performance is heavily impacted.Wile our app server (also dedicated 16Mb 8core) is hardly doing anything?
Thanks in advance,
Well CPU is going up of course during the sync and is taking up to 100% on the Application server but the load on the DB server is only giving short peaks and is not significantly harming the performance for WEB user.
What kind of DB are you using Windows SQL (Version and OS) or is it Oracle?
In the past we were facing some issues as well due to the configuration of the infrastructure and the Discs we had put in place.
It might be worth to increase the memory to at least 24 GB in case it is a SQL DB... We have a DB size of approx. 45 GB.
Also the configuration of the sync rules for the PAD is very important. For us, we have an average sync time of 1.5 min (max 2 min) for an initial full sync.
When you talk about multiple full syncs, about how many clients di you speak (we try to make the user not to run more than 7 - 10 full syncs in parallel during training and roll out) and in normal usage the no of parallel syncs is not so high anyway).
So we never had a stress test with more than 10 sessions in parallel...
I am not an expert on DB administration, but if the DB is having this amount of regression/high-load while the application server is hardly affected it might be a problem with the indexes of the DB ?!
Or maybe the Virus Scanner is not sufficiently configured on the DB server?
2 people found this helpful
some days ago we worked on a case that sounds similar to your issue.
During the synchronization invalidate cache is executed by CRM.pad. This is by design to always work with the up2date data model. Depending on the data in the designer database, this process may take a couple of seconds. During this time, the cache is also not accessible. This means that all actions are blocked until the cache is available again.
To solve this we just analyzed the server side log in debug mode and found some SQL statements accessing the designer database and lasting a couple of seconds. We then added a few indexes (more or less based on the hints we got from the MSSQL tuning advisor) and it got sorted out.
We can also take this ahead in a support case if you like.
Thanks Andre, this is helpful and we follow your steps.