Enterprise NAS Requirements

One of the few things just about every analyst firm agrees on nowadays is that file data is growing faster than any type in the data. ESG‘s file archiving market forecast projects that total worldwide file archive capacity will increase from just over 10,000 petabytes in 2008 to over 62,000 petabytes in 2012—that’s a 55% compound annual growth rate.

Between economic pressures and file data growth, it's no surprise that, when we conducted our enterprise storage survey in the fall of 2008, the top NAS features users are looking for have to do with reducing cost in terms of power, cooling, floor space, and data footprint.

What was surprising was to see that 21% of those surveyed would not buy a system without a global namespace and 51% strongly prefer a system with one.  Considering that clustered file systems with global namespaces have seen little enterprise adoption (only 11% of those surveyed had deployed), numbers this high should send a message to NAS vendors: users want the manageability a global namespace provides, including the ability to scale a system across multiple nodes and still manage it as a single entity. 

Will the scale-out clustered file systems that have seen such success in HPC and media & entertainment finally get traction in mainstream corporate IT?  It depends—many are tuned to support massive throughput required in these environments and don't do well in IO-intensive environments.  And features like deduplication and remote copy are not core requirements in HPC or media & entertainment, so some still lack these features.  But if these issues can be addressed, there is a clear opportunity for scale-out systems in the enterprise.

Comments

[Tony Asaro]

Hey Terri,

When was this research done? I was more surprised on the snapshot feedback - I thought many more would have that as a requisite capability.

[Terri]

Hi Tony - Survey was conducted in the September/October timeframe last year.

[random_graph]

Wow, concern for power and cooling efficiency certainly moved up the list over the last couple years. What can you say about the CAGR or power/cooling/floorspace overhead cost vs the CAGR of storage systems? For example, If we assume CPU Ghz per TB are fixed, then as disks get more dense we get more CPU GHz per RU and more wattage per RU. Lots of people have already hit the practical ceiling for wattage per rack.

Short of changing end-user workflows and application behavior (something we've never been successful at), It may be that the most green technology for the datacenter turns out to be de-duplication and compression solutions for storage. Anything that non-disruptively gets more files onto existing spindles.

[Terri]

Excellent points - I have not done the math for CPU GHz per RU or wattage per RU - but we are seeing the impact of the increase in density/power per RU in our surveys. We see power and cooling as a top IT challenge for enterprises that tend to have sufficient floorspace, but in midsize enterprises they are still hitting the floorspace wall before the power and cooling wall.

Agree with you completely - data reduction technology to get more data on existing spindles makes a big impact on power, cooling and space, and is a heck of a lot easier to implement than changing workflows and behavior. We've all become data pack rats and I don't see that changing any time soon.