Microsoft Dynamics AX 2012 Benchmarks: Hyper-V, EP, and High-volume Inventory

We are happy to announce that the following benchmarks will be available, just in time for Convergence 2012. We have
published both summary and detail documents for each benchmark.

  • Enterprise Portal benchmark

         In January 2012, Microsoft conducted a benchmark of Enterprise Portal for
         Microsoft Dynamics AX 2012 to measure its performance and scalability
         characteristics.The benchmark runs a variety of functional scenarios that
         use Enterprise Portal, thereby providing a view of ERP workload
         performance on Microsoft Dynamics AX 2012. This benchmark includes
         creating timesheets, and entering expenses and purchase requisitions. The benchmark
         architecture consisted of two load-balanced Application Object Server
         (AOS) instances that served web users from ten Enterprise Portal servers,
         and three workflow servers to process approvals. The benchmark simulated
         10,000 concurrent users divided among three roles.

The benchmark demonstrated a sustained rate of more than 741,587 lines per hour for
these concurrent scenarios. With this concurrent workload, a 48-core,
256-gigabyte (GB) database server sustained 55-percent CPU utilization; two
12-core, 16-GB AOS servers sustained 57-percent CPU utilization; ten 8–12 core,
16-GB Enterprise Portal servers sustained 47-percent to 85-percent CPU
utilization; and an 8-core, 16-GB Microsoft Windows Server AppFabric cache
server sustained 7-percent CPU utilization and 73-percent memory utilization.

The benchmark results demonstrate the ability of Enterprise Portal to handle
various loads concurrently, without compromising performance and scalability
for critical business processes.

  • Hyper-V benchmark

         In December 2011, Microsoft conducted a benchmark to measure the
         performance and scalability characteristics of Microsoft
         Dynamics AX 2012 on physical servers, Windows Server 2008 R2
         Hyper-V virtual servers, and a mixture of physical servers and Hyper-V
         virtual servers.

The benchmark results showed 20.68-percent performance degradation in response time
when all servers are virtualized, 13.79-percent degradation when only the AOS and
Enterprise Portal servers are virtualized, and 3.44-percent degradation when
only the Enterprise Portal server is virtualized. Similar performance
degradations in throughput were seen in the virtualized servers.

  • High volume inventory benchmark for retail environments

         Microsoft Corporation conducted a benchmark to simulate an inventory
         workload with more than 4.96 million transaction lines. The goal of the
         benchmark was to simulate three types of core inventory transaction and
         demonstrate that Microsoft Dynamics AX 2012 can effectively handle very
         high transactional volumes. The inventory transactions simulated the
         import of point of sale (POS) transactions for a large retail chain into
         Microsoft Dynamics AX.

         The benchmark results demonstrate that Microsoft Dynamics AX 2012 can
         successfully process more than 55 million transaction lines aggregated to
         3.68 million inventory adjustment journal lines. The benchmark also
         creates and posts more than 1 million transfer order lines and more than
         90 thousand item arrival lines.

Comments (5)

  1. Mark Veurink says:

    Nice benchmark! Do you have any idea which component had the biggest impact in the degradition of the response time (CPU/ Memory / Disk I/O)?

    In AX environments it is sometimes hard to find the "bottleneck", no main component (CPU/ Memory / Disk I/O) is running at full capacity and still diferences exist between several server setups.

    Personal i found that CPU architecture can make a big difference in virtualized environments. Pre XEON 5500 series processors for example perform much worse. Any idea wich processor architecture is used in these tests?

  2. Eugene Shamshurin says:

    I concur with Mark: Which CPUs/servers were used for the benchmark, and also which disks? And has anybody tried to go ssd on SQL?

  3. Mark is right.  Giving no memory pressure for the SQL/AOS/EP,  CPU will make a big difference in virtualized environments.  The hardware details are listed in the benchmark detailed documents.  For Virtualized environment it has 4 cores and 12G rams,  Host has 8 cores and 16G rams HP Mach1 R2 box with Intel Xeon CPU X5650@2.67GHz.

  4. Sebastian Widz says:

    Is there a benchmark toolkit for AX 2012 R2. What tool do you use to generate the workload?

  5. Johnson Liu says:

    Can I think all these benchmark is tested in SINGLE instance with TWO AOS servers?