Home > Blogs > VMware VROOM! Blog


Virtual SAN 6.2 Performance with OLTP and VDI Workloads

Virtual SAN is a VMware storage solution that is tightly integrated with vSphere—making storage setup and maintenance in a vSphere virtualized environment fast and flexible. Virtual SAN 6.2 adds several features and improvements, including additional data integrity with software checksum, space efficiency features of RAID-5 and RAID-6, deduplication and compression, and an in-memory client read cache.

We ran several tests to compare the performance of Virtual SAN 6.1 and 6.2 to make sure they were on par with each other.

In addition, we wanted to know how new feature performance compared to a 6.2 baseline with no new features enabled. The tests used benchmark workloads that simulate real-world activities in online stores and brokerage firms (online transaction processing, or OLTP) and in a virtual desktop infrastructure (VDI) environment. We published the test results in the following papers:

One such test used a workload that simulated typical user actions in an online brokerage application. We used a 4-host cluster with 3 VMs on each host, for a total of 12 VMs. The following graphs show the virtual machines’ IOPs per host and disk space usage saving for the Brokerage workload.

vm-iops-disk-saving

6.2 R5 (RAID-5 configured) maintains almost the same IOPs per host as 6.2 but brings down the space usage in the cluster from about 3200GiB to 2280GiB—that is 29% space saving. 6.2 D, which has deduplication and compression enabled, brings IOPs down by 25% to about 30,000, but meanwhile saves 88% disk space, taking only 373GiB on the disks. In the 6.2 R5+D case, where RAID-5 is used together with deduplication and compression, IOPs are down further by 11% to 25,000. However, the disk space saving is 92%, using only 261GiB space in the cluster.

Please note that the substantial space saving is observed because each Brokerage workload virtual machine contains duplicable and compressible data that can be reduced significantly by the deduplication and compression feature in Virtual SAN 6.2. (The actual space saving in your production environment will depend on the workload.)

Read more test results and find more information about test configuration in the published papers for OLTP and VDI workloads.

3 thoughts on “Virtual SAN 6.2 Performance with OLTP and VDI Workloads

  1. Vaughn Stewart

    === Disclaimer: Pure Storage Employee ===

    Julie,

    Thank you for the post. Per the document “VMware Virtual SANTM 6.2 Performance with Online Transaction Processing Workloads” these results were obtained via a single VM running on 4 and 8-node all-flash VSAN configurations.

    Historically, testing the performance of a single VM produces results that are unobtainable in the real world where the hardware is supporting multiple VMs. The inclusion of additional VMs introduces I/O contentions that reveal themselves in increased storage latency and increased ESXi CPU & memory utilization.

    I would like to ask VMware to run this same test with 1 VM per VSAN node in order to compare results from shared and dedicated HW infrastructures.

    Cheers,
    -vaughn

    Reply
    1. Jase McCarty

      **VMware Employee**
      While I haven’t dug into the details of the paper, I will mention the following:

      For the DVD Store workload:
      On Page 4, it states: “Unless otherwise mentioned in the text, 4 DVD Store virtual machines per node are used in the experiments.”

      For the Brokerage workload:
      On Page 5, it states: “Unless otherwise mentioned in the text, 3 virtual machines per node are used in the experiments

      I’ll make the following suggestions that we update the following items in the paper for better clarity:
      Appendix B is updated to include “Each Brokerage workload virtual machine…” rather than “The Brokerage workload virtual machine…”
      Appendix C is updated to include “Each DVD Store workload virtual machine…” rather than “The DVD Store workload virtual machine…”

      Thank you for bringing this to our attention.

      V/R,
      Jase

      Reply
    2. Zach Shen

      Hello Vaughn,

      I’m the author of the performance whitepapers referenced in this blog. As Jase already pointing out in his comment, the results here are indeed running *multiple* VMs on *every* host in the cluster as indicated by the exact text from the paper. My apologies if you were confused by looking at the results in the blog alone.

      Regards,
      Zach

      Reply

Leave a Reply

Your email address will not be published. Required fields are marked *

*