posted

10 Comments

HCIBench stands for “Hyper-converged Infrastructure Benchmark”, which is an automation wrapper around the popular and proven VDBench open source benchmark tool that makes it far easier to automate testing across a hyper-converged cluster.

HCIBench was released last August, one year later, we published another major release and moved it from Dropbox to Fling. If you are new to HCIBench, you are more welcome to catch up by reading these posts written by Chuck Hollis, Cormac Hogan and Kevin O’Brien.

If you already have HCIBench installed, you may want to update to the latest version to take advantage of all the new features!

What’s new:

  • HCIBench service and the VDBench client have moved to Photon OS 1.0, which is an operating system developed and maintained by VMware.
  • HCIBench now has added another disk initialization option specifically for the storage that has deduplication enabled.

NOTE: The previous version doesn’t support disk initialization on a deduplication enabled storage. Updating the HCIBench is REQUIRED if you need to test storage that has deduplication enabled.

  • HCIBench now supports multiple datastores deployment. Multiple datastores in the same cluster could be tested at the same time for performance comparison.
  • HCIBench now is able to provide DHCP service if the test VLAN doesn’t have it.
  • HCIBench now creates a VM folder dynamically instead of creating a folder named “perf-test-vms”, this allows the user to have multiple HCIBench Appliances to test multiple clusters in the same vSphere data center at the same time.

Ok, now please allow me, as the writer of HCIBench, to briefly go through the installation and configuration process and provide some tips. For the complete set of installation instructions please refer to the  HCIBench user guide.

First, download the HCIBench OVA from https://labs.vmware.com/flings/hcibench and deploy it to vSphere 5.5 or later environment. Because the vApp option is used for deployment, HCIBench doesn’t support deployment on a standalone ESXi host, the ESXi host needs to be managed by a vCenter server.

When configuring the network, if you don’t have DHCP service on the VLAN that the VDBench client VMs will be deployed on, the “Private Network” needs to be mapped to the same VLAN because HCIBench will be able to provide the DHCP service.

hcibench_blog1

After configuring the public network and the root password, you can start to deploy the HCIBench OVA.

If the public network needs to be reconfigured after deployment, refer to the administration guide. NOTE: do NOT edit the file /etc/systemd/network/eth1.network or edit eth1 of HCIBench.

After HCIBench is booted up, open the browser and go to http://HCIBench_IP:8080, the page may take up-to 3 minutes to load after boot-up so be patient.

In the configuration page, the first section is the vSphere environment, the value of vCenter Username needs to be the UPN (User principle name) format e.g. “Username@Domain” rather than Down-Level Logon Name format e.g. “Domain\Username”.

hcibench_blog2

If the network (here we use Private_VLAN_NO_DHCP as example) specified does not have DHCP service, make sure:

  1. Map HCIBench eth1 to “Private_VLAN_NO_DHCP”
  2. Check Enable DHCP Service on the Network
hcibench_blog3

Then HCIBench will provide DHCP service on VLAN “Private_VLAN_NO_DHCP” during the testing.

In Cluster Hosts Information section, if the type of network specification above is using a Distributed Switch PortGroup, then Deploy on Hosts must be UNCHECKED. Having Directly Deploy on Hosts configured is able to speed up the deployment process but can be used in conjunction with a Virtual Standard Switch.

hcibench_blog4

In the section of Vdbench Guest VM Specification, the Number of VMs must be integral multiple of the number of datastores specified above. For example, if there are 3 datastores specified above, the number of VMs must be 3*N.

Now scroll down to the Vdbench Testing Configuration section, you can either Generate Vdbench Parameter File by Yourself or Upload a Vdbench parameter file. Either way, the parameter file is located in /opt/automation/vdbench-param-files. After generating or uploading the VDBench file, click Refresh then you will be able to see all the files generated or uploaded. Select one particular parameter file you want or just Use All, if you choose to run all the parameter files they will be executed sequentially.

hcibench_blog5

NOTE: When creating the VDBench parameter file by yourself, make sure the number of disks are not more than the Number of Data Disks you specified earlier, disks starts from sda, so if you specify 4 disks for example, you should specify /dev/sda through /dev/sdd.

Disk initialization is designed to avoid storage first-write penalty during the performance testing. For the Disk initialization, you have 3 options: NONE, ZERO and RANDOM. NONE means you choose not to do disk initialization before testing, this could be used if you just want to do a quick testing and have a brief idea how the storage performance looks like, however the first-write penalty may apply which might impact the results. ZERO could be used for initializing a storage without deduplication enabled while RANDOM is particularly designed for storage with deduplication enabled.

hcibench_blog6

We highly recommend to do disk initialization for a serious performance testing, even if this might take a couple of hours longer than not doing that.

If you are deploying the HCIBench appliance for the first time you will be asked to upload the VDBench binary file. We are not allowed to re-distribute VDBench. Just download the VDBench zip file from Oracle website, and upload it directly. This is a one-time effort, once uploaded, this section will become invisible. After uploading, you should be able to see your VDBench zip file in your browser: http://HCIBench_IP:80/vdbench-source/

In case you need to update the VDBench zip file, you can simply go to HCIBench console and delete the old zip file from /opt/output/vdbench-source and upload the new one through the page.

hcibench_blog7

Here are some tips that I think are really important to use HCIBench:

  1. Make sure the VDBench client VM network has enough IPs available if DHCP service is provided by your environment.
  2. Make sure the VDBench client VM network can have DHCP and SSH services to go through.
  3. If the test failed e.g. no results showed up or the results are showing all 0, check the logs in HCIBench:/opt/automation/logs and check the deployment log first.
  4. On the configuration page there are two places for uploading files, the first one is for VDBench parameter files and the second one is for VDBench binary zip file. The second one will be invisible after uploading.
  5. All the results are stored in /opt/output/results, which has 200GB space, in case it’s running out of space, you should migrate the old results to somewhere else or delete some of them to free up the space.

In the part 2, I will share some advanced operations from the HCIBench console, including how to re-use the existing VDBench client VMs for more tests, how to debug HCIBench from the console and how to leverage the scripts of HCIBench to perform other automation tasks. Stay tuned!