The recent changing of the guard at the Storage Networking Industry Association (SNIA) means the group has a new board of directors and a new chairman, Vincent Franceschini.
The SNIA is in the process of writing its three-year strategic plan and Franceschini (who is also the senior director of future technologies for Hitachi Data Systems) recently gave me a quick round-up of the major challenges facing the storage industry.
“Storage solutions need to be further integrated with other layers in the datacentre”, he says, adding that: “The storage industry needs to start collaborating a lot more with other datacentre players. For example, [SNIA has] a working relationship with the Open Grid Forum.”
My conversation with Franceschini touched on numerous interesting topics, including the IBM-initiated Aperi storage standards body (“SNIA and Aperi are still talking and we will have a working relationship on that front”); the possibility of SNIA generating code (“We are not considering an open source model”), and extending the SNIA communication channels via blogs (“Funny that you mention it — we were just discussing using blogs and other tools like wikis”).
Without discounting the importance of the other topics, that opening mention of grid computing seems to me to be the most interesting and promising topic.
“We know that [OGF is] working on a new infrastructure model with some level of orchestration middleware that will touch on storage and data management at some point in time, and we want to make sure that the storage industry is positioned there,” Franceschini says.
Storage grids are a potentially controversial topic, as I realised later on while talking to Gary Francis, senior vice president at Crosswalk. In April, Crosswalk launched iGrid, an intelligent storage grid system. The name is a rather blunt but clarifying self-introduction — if you know what a storage grid is, that is.“People today have a little confusion about storage grids and their definition,” Francis says. That’s probably an understatement. Many storage vendors claim to have a “grid” solution, but the lack of a commonly-accepted definition of a storage grid means the grid solutions group is made up of strange bedfellows: virtualisation solutions, parallel-file system solutions, and various flavours of clustered storage, to name just a few.
The authoritative SNIA dictionary doesn’t even mention storage grids. Searching the site does dig up some interesting documents, however, such as a helpful tutorial that defines a storage grid as: “A unified, managed infrastructure that leverages grid-computing concepts to present a single, utility-like storage system.” That’s a definition that can fit many systems, in my opinion.
Definitions apart, iGrid creates a network of independent nodes between servers and storage, all based on 64-bit commodity servers and running the Crosswalk proprietary iGrid OS.
“We load-balance the access from all servers in case of unequal load distribution or if there is a failure,” Francis says, adding that the storage on the back-end of iGrid appears to the host systems like an easy-to-manage single pool, serving NFS or CIFS files systems very quickly. Crosswalk has a surprisingly technically rich online presentation that better describes how the system works.
According to Crosswalk benchmarks, a two-node iGrid can deliver transfer rates of about 700 Mbit/s to 16 competing applications. It can also scale linearly, so doubling the number of nodes will also double performance.
Last week Crosswalk announced a new node model, the iGrid 7100, and version 2.0 of the iGrid OS.
The iGrid 7100 doubles the cache memory to 32GB and mounts four dual core 64-bit processors instead of the single core of the previous model. Crosswalk estimates that grids based on the 7100 model should improve performance by about 50%.
Is iGrid truly a storage grid? I would hold judgment, at least until the confusion that Francis mentioned clears up. It’s certainly worth keeping an eye on Crosswalk and other interesting grid solutions, but clarification of what the technology is (and is not) is badly needed. Customers are not going to buy what they don’t understand.
Apicella is a senior analyst at the InfoWorld Test Centre