Dell's Project Hermes aims to move flash capacity onto the serverDate: Jun 28, 2012
Dell Inc. has begun releasing some details about its newest flash cache product, code-named “Project Hermes.” The project includes PCIe solid-state storage cards in servers, with software that reads and writes in memory located in the server. The goal is to move storage closer to computing power.
Project Hermes was made possible by the acquisition of cloud memory startup RNA Networks, which has allowed Dell to transform memory from a fixed storage resource within the server and make it a shared resource among servers. Bob Fine, Dell’s director of product marketing, said “the concept is to imagine putting tier-two and [tier-]three storage in the array and tier-one storage in the server.” Randy Kerns, a senior strategist at Evaluator Group, said Project Hermes is a response to EMC’s VFCache, aiming to spread storage resources across servers rather than rely on the storage-area network (SAN).
To get a better idea of what Dell might be able to do with this technology in the future, SearchStorage interviewed analysts Terri McClure of Enterprise Strategy Group and Andrew Reichman of Forrester Research at Dell Storage Forum 2012 in Boston. Here’s what they had to say:
Terri McClure, Enterprise Strategy Group: One of the hidden gems of the Dell portfolio comes from an acquisition they did roughly about a year and a half ago of RNA Networks. They’ve got some really interesting technology that’s going to allow them to take advantage of end-to-end solid-state memory from the server all the way down to the storage system. They’ve given some sneak previews this week but it’s really exciting stuff. I was excited about it when I saw it at RNA Networks and I’m more excited about it with what I’m seeing Dell do with it today.
Andrew Reichman, Forrester Research: Project Hermes is pretty interesting, and has the potential to be a bit of a hybrid between the two approaches where the capacity itself lives on the server, but you can share the capacity across many servers using a 1 GB, 10 GB, 40 GB or InfiniBand connection between the servers, allowing each of the servers to share the capacity even if it doesn’t live on the server itself. So that’s pretty interesting from a capacity utilization perspective. One of the futures that Dell talked about was the possibility of it being included in the Compellent Data Progression tool so that you could actually have the storage system use that server-side flash as a tier and then you’d be able to control the management of it. I don’t believe that’s available today, but looking forward that seems to be a pretty interesting thing that Dell [can do] with this capacity. In the meantime, its being used as a combination read and write cache for data, so it’s something that could provide a lot of acceleration to applications that need performance boosts either on the read or the write side.
One of the key pieces of the Dell implementation of server-side cache is that it reads and writes. A lot of the implementations we’ve seen so far are read-only. It’s a pretty interesting product, [and it’s] one more way I’m expecting to see flash capacity be deployed throughout the data center -- on the server, in the network and on storage arrays. This seems to be a pretty good step forward and I’m looking forward to seeing what more they can do to integrate it with all of the pieces of the solutions they have.