One of the pleasures of being at Tech Field Day (this one is #9) is that you get to encounter start-up companies who are exiting stealth and using the event as their launch vehicle. One such example is a presentation from Infinio, who kicked off the day with an unveiling of their Infinio Accelerator software. This solution provides server-side caching, which is definitely a topic area I’ve been keeping my eyeball on. But rather than leveraging an internal flash device, such as a PCIe card or SSD, Infinio form a distributed caching layer by borrowing memory from the host pool of RAM.
The message of Infinio’s presentation is that the solution is easy to install, try, and buy. Perhaps you could sum that up with “easy to consume.” Focus was given to the concept of how many clicks it takes to install the product (9) and the balance of internal engineering focused on front end interface and the back end magic bits. Ultimately, Infinio states that it take about 30 minutes from download to production, which I think is partially possible by the fact that the installation does not require a reboot or any manual process to deploy. If you do have some trouble along the way, the fully guided installer (which works on Windows, Linux, or Mac) even includes a video to move things along.
Note: All travel and incidentals were paid for by Gestalt IT to attend Tech Field Day 9. No other compensation was given.
One part of the message that I felt was rather sharp is how easy it is to also remove the Infinio Accelerator software. The solution can be safely removed without an interruption or reboot to the host. There’s no need to uninstall a driver, VIB file, or juggle around VM disks. I’m interested in giving this a shot in the lab just to put the claims to the test.
Under the Covers
So how does one tie in a RAM based server-side caching accelerator for NFS based workloads with just software? The Infinio Accelerator is an OVA that is deployed to each host. Each appliance creates a bridge network, builds a vSwitch, and leverages proxy ARP. The storage array ends up thinking that it is talking to the ESXi host, and the host ends up thinking it is talking to the storage array. The accelerator will only take action on NFS requests relevant to the storage – it won’t blindly capture other traffic. Today, the solution is focused on write-through performance acceleration, which ultimately provide an offload of reads from the array, which both boosts read performance and reduces read latency.
Worried about gobbling up too much host RAM? The Infinio Accelerator uses 8 GB of RAM by default, but can be brought down to 4 GB if desired. In my opinion, 8 GB should be relatively trivial in most environments – I think 96 GB of RAM per host, if not higher, is quite common. A mapping table of the cached content is distributed throughout the cluster, and the map table is copied during vMotion to keep it warm. For those new to this concept, warming the cache is just the activity of making sure that cached content leveraged by the virtual machine is available at the target host. This ensures that functions like VMware DRS doesn’t inadvertently harm performance as it shuffle around workloads.
The Infinio dashboard is very sharp and includes some powerful metrics and smooth sparklines for things like response time improved, requests offloaded, and bandwidth saved. I feel that the dashboard could be something useful for a front like engineer or tech and also a business owner or management member who wants to see the ROI of their investment.
For those who enjoy a “single glass of pain” style of management, no such luck for you! The accelerator is handled via a web page, not a client plugin, so you will need to bounce between screens. Notice that next to each host is a “disable acceleration” option for when you need to turn off write-through caching.
The Infinio Accelerator is licensed by vSphere socket, which is much easier to track and maintain for most VMware administrators. After all, we’re already tracking host sockets for VMware licensing anyway, so this steps right in line with that idea. As an added bonus, you can shut down license on one host and use it elsewhere to sort of shuffle things around when needed.
From a support perspective, a purchase includes 3 years of support: phone, e-mail, and eventually live chat.
I’m glad to see some focus towards accelerating vSphere NFS workloads instead of just block storage. I’ve managed, and work with folks who manage, very large farms of vSphere hosts that are tethered to storage using NFS. I’ll be interested to see the product in action, as even though I was shown a demo of the application in action, a demo environment is much simpler and more controlled than a real, dynamic environment hosting chaotic workloads.
However, even with that, I see Infinio having a solution that is unique enough to shake things up, especially in RAM heavy environments or end user compute workloads.