Tintri Briefing Note
Most primary storage scale-out systems are tightly coupled clusters made up of identical servers acting as storage nodes. As the organization requires more storage performance or storage capacity it adds new nodes. In many cases these nodes must be identical to the existing nodes and it adds performance and capacity in lock step. The tightly coupled approach does bring linear scale and simple expansion but it does so at the expense of flexibility. Tintri’s VM Scale-out announcement asks IT professionals to rethink the status quo when it comes to scaling storage.
Scale-out storage systems, especially as the data center adopts more flash, face several problems that may limit their appeal. The first problem is that storage media changes over time. Flash media in particular is evolving rapidly, increasing in both durability and density while decreasing in cost. The flash media used today will be easily circumvented in the future. While many scale-out systems can mix media types in their clusters it is not as simple as just dropping in new nodes. Often new nodes must deliver the same capacity as older nodes. In an era where flash density is increasing rapidly this may cause some configuration issues for these vendors.
The second problem is that tightly coupled scale-out architectures require the addition of storage and performance in lock step. The real data center never scales this way. Usually one of the two storage attributes takes precedence over the other, which one it is varies from data center to data center.
Finally, for many data centers, one all-flash or hybrid array provides all the performance and capacity they need in a single system. They want the ability to scale in the future but don’t want to have to pay for that ability today. The need for more capacity and/or performance will be a long way off.
The Tintri Answer
Tintri’s answer to these scale-out challenges is a loosely coupled cluster. “Nodes” in the Tintri eco-system are actually stand alone storage systems. What Tintri adds in this announcement is the ability to centrally manage these systems and automatically move VMs between nodes as performance or capacity demands dictate. Tintri leverages its already rich performance analytics to make sure these virtual machine movements are both needed and directed to the most viable storage target.
This feature is actually similar to the way a VMware cluster scales compute. No one “scales” a single server by hot adding CPU or memory anymore. You simply add another server when you need it, and vMotion and DRS software optimize VM placement for you. Tintri VM Scale-out works like this as well. The software moves VMs across storage nodes to optimize performance and capacity, and when you add new VMstores allows you to take advantage of that new storage immediately.
But Wait, There’s More
In addition to the VM Scale-out announcement, Tintri also announces new capacity options for its existing VMstore T5000 All-Flash arrays. These systems can scale from 17TBs to 308TBs of effective capacity. Tintri is also updating its operating system to support larger file sizes (vDisks can now be 64TB), improve its deduplication capabilities and provides remote VSS backups for Hyper-V.
Every Tintri customer we have spoken with has nothing but positive things to say about the solution. The only concern tends to be scale. This announcement resolves that concern by leveraging the existing capabilities of the solution, instead of shoehorning a legacy scale-out model into the architecture. The result is a scalable architecture that provides maximum flexibility. VM Scale-out combined with the new hardware and updated OS keep Tintri as a top consideration for virtualized environments.