Hyperconverged infrastructure, a look at Nutanix
Sorting Out HCI
Today’s Hyper-converged infrastructure (HCI) vendors have some exciting product offerings. HCI ultimately provides scalable and flexible storage along with coupling computing resources. Recently at work, our KPI metrics started to show that some of our SAN hardware was having issues keeping up with production workload. So we ended up looking at a few HCI vendors; Simplivity, Nutanix, Pivot 3, and Vmware VSAN. After our initial investigations, it became clear that we weren’t quite ready to step into HCI just yet. Our project scope explicitly called for storage performance. At the current junction, additional compute wasn’t necessary and not budgeted for the project. Since HCI solutions couple storage with computing costs, we ending up investing in an all-flash SAN solution.
However, I was very intrigued with the different HCI platforms. What interested me the most was the ability to scale storage using x86 based systems. During our Nutanix research, I came across their community edition. I decided to load Nutanix CE on my home virtualization server and give it a whirl. There are lots of other great sites with information on how to get the initial setup going, so I’ll focus more on some of my specific findings during my home lab testing.
Nutanix CE Single Node Setup
I started out with a single node cluster as I didn’t have enough spare SSD parts for a three node cluster setup. Unfortunately I can’t review storage clustering, but that will come soon as I’ve ordered two extra 240GB SSD’s. The single node setup consists of the hypervisor (Acropolis) installation on the host; a VM called CVM (Nutanix controller VM), and Prism the management software running on the CVM. A few things I could test with the single node setup were in-line compression, hybrid storage presentation, importing of OVA’s, and overall management of the system. The minimum specifications must be at least 16GB RAM, four CPU cores, one SSD greater than 200GB, and one spinning disk greater than 500GB. For my lab setup, I have a Samsung 240GB SSD drive and a 1TB 7200rpm spinning disk drive. The default configuration for storage after the installation was one storage pool with both the SSD and HDD within the same pool. I enabled in-line compression on the default storage container. I couldn’t setup dedupe because I didn’t have enough RAM installed.
To test file transfer rates, I set up a quick NFS share. To get this going you simply add a whitelist IP range into the Nutanix storage container under the advanced settings. On my windows 10 client I created an NFS share. You will need to know the Nutanix container name which you can find under the Nutanix Prism storage settings. Then run the following in cmd on the client machine:
mount -o nolock nutanix-cvm-ip:/nutanix-container-name windriveletter
mount -o nolock 192.168.254.211:/default-container-50404077655521 P:
I started copying files directly to the Nutanix CE host without the need to set up another VM. I copied a few gigs of family pictures and videos to test throughput. I topped out at around 500Mbps. These results may have been due to disk bottlenecks as I was copying from an older 1TB 5200RPM drive on the Windows 10 client machine. To make sure my NIC or switch wasn’t an issue, I set up a VM ubuntu server on Nutanix and ran Iperf which runs TCP/UDP tests from memory. My tests topped out at the NIC interface speed of 1Gbps running Iperf between my physical desktop and the Linux server. I’m also seeing a total data reduction ratio of 1.67:1, but that’s a combination of data reduction (compression), any cloning, and VM thin provisioning. The compression data reduction ratio is 1.03:1 with a data reduction of 5.89GB. However, most of the large data I have stored are compressed jpg and avi files. Below is a screen capture of the Prism storage web interface:
Nutanix CE Conclusion
The methods I’ve used so far aren’t the best way to store data files, so for my next test I’ll set up a windows server VM and enable file server services. So far, I’m having lots of fun learning the in’s and out’s of Nutanix CE. I’ve started testing backups, imports, and a few other things. Once my SSD parts arrive, I would like to rebuild the setup as a three-node cluster to see how the storage clustering functions across three nodes.