Our own experience is that there are big differences in how a given hardware configuration will perform, depending on whose hyperconverged software stack you’re using.
If performance is important to you, you should know what you’re getting before you buy.
With regards to VSAN, we’ve been continually publishing the results of our own internal testing, and done so with enough detail so that someone could reproduce the results if desired (scroll to the bottom of this page for a sampling). We’ve also supported independent reviewers such as StorageReview.com to share their own unbiased results.
That being said, we’d like to do more — much more.
Wouldn’t it be great if anyone could easily do their own head-to-head testing?
To help customers make better informed choices, we’re introducing a free new tool that makes storage performance testing on hyperconverged clusters much, much easier.
We call it HCIbench, as in “hyperconverged infrastructure benchmark”. It’s essentially an automation wrapper around the popular and proven Vdbench open source benchmark tool that makes it far easier to automate testing across a hyperconverged cluster.
The people who’ve tried it tell us that it’s a huge step forward in simplicity and repeatability. Easier testing = more testing + better testing.
Why Testing Hyperconverged Is Harder Than It Looks
Doing performance testing on familiar external storage arrays can be relatively straightforward. Set up a workload or two — and crank it up. Once you have all the hardware bits, these are not hard tests to configure and run.
Hyperconverged systems are different: each and every server is expected to run many dozens of VMs, and also contributes storage resources that are part of the whole. Rare would it be to see someone build a good-sized cluster, and only run a small number of VMs on it.
In the real world, most clusters are crammed full with hundreds of busy VMs, each driving their own storage traffic. And that’s exactly what you want to test, if possible.
But it’s not as simple as it sounds.
Doing this “busy cluster” testing using standard tools (e.g. Vdbench, FIO, IOmeter, etc.) usually means a lot of scripted setup, and a lot of data gathering.
It’s unnecessarily complex, it burns time, and mistakes inevitably happen.
HCIbench makes this task far simpler — it asks you to specify your desired testing parameters (size of working set, IO profile, number of VMs and VMDKs, etc.) and then spawns multiple instances of Vdbench on multiple servers. After the test run is done, it conveniently gathers all the results in one place for easy review — and resets itself for the next test run.
If you’re not familiar with Vdbench, you should be. It’s a great go-to tool for testing storage performance with many useful features. We’ve built HCIbench on top of it to make it especially easy to use when testing hyperconverged storage performance.
The people who’ve tried HCIbench so far really like it — they tell us it has dramatically sped up their testing efforts. That means more tests can be run in whatever time you have available — and ending up with more useful results. Vdbench is already open source, we intend to make HCIbench available via open source as soon as we can.
How To Get A Copy
Feel free to grab a free copy here.
Installation and setup is relatively straightforward. In a matter of minutes, you’ll be running your own cluster-scale storage performance tests.
And, if you’re somewhat new to storage performance testing, I’ve written a quick do-it-yourself primer here.
Have Questions? Need Help? Want To Publish Your Findings?
If you have questions or need help with the tool, please leave a note in the VSAN community forum here. Or email us at VSANperformance@vmware.com
If you want to publish your VSAN findings, we certainly encourage that — but please do respect our EULA and submit your testing configuration and methodology to VSANperformance@vmware.com prior to publishing. We’ll do our best to get back to you quickly.
Note: we won’t approve publishing results on unsupported hardware or software configurations for obvious reasons.
For your convenience, a partial list of published VSAN performance results appears below:
VSAN 6 Performance White Paper
Synthetic Test Results
JetStress Test Results
Dell FX2 DVDstore Results
Oracle on VSAN 6
VMmark Results (internal testing)
LoginVSI (VDI) Results
Multiple Microsoft Workloads