We all know that there is a lot of wasted space in most VM storage repositories due to duplicate data. (Or we should.) But what isn’t as well known is the way the disk storage is configured can introduce a lot of delays in its throughput. This is because most storage arrays aren’t aware of the peculiarities of how VMs have to store their data on the specific blocks of their arrays, and the array can waste time seeking and retrieving information from its VMs.
A way around this is to re-align your VMs to make sure their information is stored properly on the array, but until now this has been a fairly complex process. Tintri is attempting to resolve this and other VM storage issues with its second-generation VMstore hardware storage appliance. Most of its storage input/output comes from its built-in flash memory and the new version has dual storage controllers for added redundancy.
The VMstore comes with the ability to do auto-alignment, to boost performance by anywhere from 10 to 20 percent. VMware is currently supported and other hypervisors are planned for the near future. Here is a description of the underlying issue.
The array has some additional features, including the ability to have visual and quantitative (real-time and historical) “end-to-end latency” from the application all the way down to the block storage level. You can now tell where your storage bottlenecks are, and be able to determine if you have throughput or application layer issues. Dashboard can either be run from a browser, as you see here, or fit inside a tab on vSphere’s vCenter.
Pricing isn’t cheap for this unit — $65,000 plus an Ethernet network adapter for 8.5 TB of usable space. It will be shown next week at the VMworld show in Vegas.
