Server hypervisors improves utilisation, but what about the attached storage? In fact the server hypervisors are so successful in improving the ability to drive up virtual and physical workloads, that storage can become a casualty. As discussed here earlier.

Tintri is another vendor with ideas on how to fix the VMware bottleneck and make a shilling in the process. The two elements to the Tintri proposition is (1) the Tintri OS and (2) use of flash. I recently did a piece on the many companies with different takes on how to exploit SSD and Tintri has its own unique exploitation of flash.

The Tintri VMstore product is an appliance that houses the processors and two sets of hot-plug bays for flash SSD & SATA HDDs respectively. Connectivity is via 10GbE and the VMstore requires VMware vSphere. In fact many of the attributes which the VMstore management interface measures comes straight from the vSphere API.

The Tintri OS creates a datastore across the SSD and HDD and data is compressed and de-duped for the Tintri filesystem to manage. The philosophy of the OS is to manage the storage to minimise the bottleneck which storage can create. And to maximise data transfer to and from flash. The proposition is that a VMware data pool is optimised which an external traditional disk array needs costly performance resources to mimick.

The customer segment for this type of solution is more than 50+ VMs and more than 6 ESXs. Entry list price is $90K for 13.5TB of usable capacity. Tintri cites that comparable solutions cost more than $500K.

On the issue of the life span of the MLC based flash, Tintri anticipates that on a busy system the flash will last 5 years or more. And that the customers will have moved on to new larger capacity drives by then.

Storage on hypervisors is clearly an area of innovation, and I will spend more time exploring how VMware itself may address this.