Performance Difference Distributed / Standalone Systems

(0) Share

Report

Posted on by Community Member

Hi,

we currently made some performance testing on different environments. One of our main insights was a difference between standalone setups with one AOS and the database on the same server and distributed setups with a multi AOS environment and an own database Server. For most of our long running queries, the runtime on the standalone setup was faster by factor 3.

Obviously, time gets lost in some communication processes / network etc - wasnt really successfull figuring out the exact reasons.

Does anyone have reference values, can tell me what would be a normal performance difference between such Systems and what the exact reasons for the differences are? Feel free to ask, if you need further information - and thanks for any help in advance!

*This post is locked for comments

I have the same question (0)

All responses (7)

Answers (0)

Suggested answer

nmaenpaa 101,160 Moderator on at

Like (0)

Report

An SQL query should not be slower if the database runs on a different server. But if there is latency between the servers, every roundtrip will take some time.

So if you really see that some queries are 3x slower, the problem is somewhere else than in the communication between the AOS and the database.

On the other hand, if your code makes 100 000 calls (queries) to the database, you might see a difference. In general you should try to minimize the number of database roundtrips in your code.

Maybe your standalone SQL server has poor performance?

What is the latency between your AOS server and database server? You can test it with a simple ping command.

Also, Microsoft suggests that you use a dedicated server to run the SQL Server.

https://docs.microsoft.com/en-us/dynamicsax-2012/appuser-itpro/configure-sql-server-and-storage-settings

Was this reply helpful? Yes No
Suggested answer

Tai Phan 55 on at

Like (0)

Report

Distributed: Resource assigned only for DB Server, TemDB hosted on this server, not share hardware with the other application, must consider network card (>1GB). Then this scenario will have good Performance than standalone.

Standalone: All resources using in same Physical SVR. Sometime it conflict, DB file size increase too fast... this is not recommend for performance in Production environment,...

This is SQL Server topology recommendations for availability and performance from Microsoft

docs.microsoft.com/.../sql-server-topology-recommendations-for-availability-and-performance

Was this reply helpful? Yes No
Community Member on at

Like (0)

Report

Thanks for your answer. The productive environment has a distributed setup, the SQL Server does not need to share any ressources, neither do the AOS Servers. Also, the network workload is somewhere around 1%. Yet, is has the worse performance compared to the standalone (Test-) environment.

Was this reply helpful? Yes No
Community Member on at

Like (0)

Report

Thanks for your answer. The raw performance of the standalone SQL Server seems not to be the problem. The latency between the AOS Server and the Database Cluster Server (same for the instance Servers) is at under 1ms.

Was this reply helpful? Yes No
Suggested answer

nmaenpaa 101,160 Moderator on at

Like (0)

Report

If the latency is under 1 ms, then it's definetely not an issue. What do you mean by "raw performance of the standalone SQL Server"?

The most usual bottleneck in AX database performance is disk speed.

You can use DynamicsPerf tool to analyze the performance of your system: blogs.msdn.microsoft.com/.../dynamicsperf-2-0-deployment-guide

The first thing to check here would be the Windows Performance Monitor counters which help you identify any issues with disk speed or other server resources: blogs.msdn.microsoft.com/.../setting-up-windows-performance-monitor-templates

Was this reply helpful? Yes No
Community Member on at

Like (0)

Report

We currently changed the disks, resulting in an average Disk IO Speed of around 2ms/Read and Write. For the CPU, the usage averages at 30%, peaking at 80% under heavy load. The database Servers run on 2.80GHz Prozessors (2 Cores) with 192GB RAM.

Was this reply helpful? Yes No
Suggested answer

nmaenpaa 101,160 Moderator on at

Like (1)

Report

It's not so much about the average but the peak values in read/write performance. Anyway we can't see into your system, but you can use the DynamicsPerf tools to analyze it.

But back to your original question - no, distributed system is not supposed to be slower, and most likely your problem is somewhere else than in the distribution of the system.

Was this reply helpful? Yes No