Candidate to be tested under Virtualizing Business Critical Enterprise Applications

[AX2009Tech]

Hi,

One suggestion for future performance tests under the category Virtualizing Business Critical Enterprise Applications - Microsoft Dynamics AX 2012 (R2) running SQL Server 2012 SP1 with Analysis Services, Reporting Services and SharePoint Foundation 2010 with special focus on AOS (application server) and SQL Server performance characteristics.

Why? Since Microsoft has it's own hypervisor in Hyper-V, they will never invest in such performance tests (I bet), but a lot of customers are in fact running AX on vmWare. And AX is a mission critical application since it's an Enterprise Resource Planning application with a typical OLTP load (also some OLAP load).

I think a lot of people would appreciate such an initiative to get some facts on the table.

Microsoft has done a benchmark comparing the performance on the same hardware for 3 implementation scenarios: physical, virtual and mixed. The test showed a two digits (percentage) performance loss when running all servers as virtual and a little less when the SQL Server was running physical and all others virtual. The SharePoint server (AX Web application called Enterprise Portal) was not supprisingly the best candidate to virtualize in this benchmark. Since the document is partner restriced (Partner Confidensial), I can't publish the report here or give the real numbers. I can ask for permission to share it if this is of interest.

I don't know how to vote for applications to test, but it's worth a shot. AND I'm aware of the benchmarks for SQL Server and SharePoint, but it's still the application that makes the difference with it's unique characteristics.

[mcowger]

You'll never see this, because Microsoft won't allow anyone to publish benchmarks of their apps running on VMware.

[Josh26]

It would also be hard to get an analysis that means a lot.

It may be "critical" and people may discuss at length "performance" but I've run AX successfully on very lowly specced virtual machines for a significant number of users - usually based on what the support agent recommends as those specs.

AX isn't an app you can easily benchmark. It's not like you can measure "throughput". At least with Exchange, you can test "latency with x number of users", but AX doesn't have a "standard query" like "read email", nearly everything you do takes a non-deterministic amount of time.

Then there's the fact that, because the support agents respond to anything and everything by deleting a cache and letting it rebuild over several hours, AX is rarely working to its potential :smileysilly:

[AX2009Tech]

I see your point, but at the same time I don't see the really big difference between other MS products and AX except that SQL Server is a generic service used by many products and applications. But what about SharePoint? Quite similar I think - You don't find many standard, out of the box deployments of SharePoint around...

Anyway - point taken and the only possibility is to run a private benchmark.

[AX2009Tech]

Agreed, but my point was that many customers (and their AX partners) running AX on vmWare would consider such a test as very valuable.

I know AX quite well from the technical perspective and agree with you to some extent that it's difficult to benchmark AX. Anyway MS has done this on Hyper-V and I findt this quite valuable when discussing virtualization with our customers.

I'm positive to virtualization, but when the customer bring up predictable performance, I get a little bit sceptic about virtualization. My experience shows that all the x-factors contributing to the dynamic nature of virtualization, adds a lot of complexity when performance issues arise and the customer is experiencing the performance issue in AX. It depends heavily in the type of cloud - private (dedicated platform for the customer) gives more room for optimalizing compared to public clouds (several customers sharing the same virtual platform).

I haven't checked, but if such a benchmark has been performed for i.e. SAP, I don't see the big difference since they are competing ERP systems.

[AX2009Tech]

To give you some details without compromising the confidentiality, the MS AX Performance Team has revealed some details on their blog.

I also find some best practises for SAP on the vmWare site, but it seems to be a SAP initiative. Maybe that's the key take away from this discussion and then it's up to "someone" with the competence, access to a suitable vmWare platform and the necessary set of tools, to do "something" in this area.

Case closed