Developments in network-attached clustering signal new storage approach, says Steve Broadhead
For all the talk of disruptive technologies, few storage suppliers are brave enough to define their products as revolutionary. Why? Because the word implies that age-old IT no-no, the forklift upgrade: throwing away everything you have and replacing it with something new.
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
This is not the approach of storage cluster technology start-up Isilon. Recent tests of its IQ network-attached storage cluster at Broadband-Testing Labs demonstrate the benefits of innovation.
The key storage issue is how to plan for the vast amount of disc space (and often high throughput and concurrency) required by unstructured data and digital content without hugely over-engineering storage requirements.
The real cost of storage comes in overprovisioning as a means of future-proofing. And then there is the problem of data migration – the costly offline time while systems are rebuilt or built from scratch. Clearly there is a need to simplify the whole process of storage planning, capacity upgrades and, not least, deployment and day-to-day management thereafter.
Here is where Isilon believes its IQ cluster offers something innovative – technology that can lead to significant cost savings, as well as hitherto undreamed of levels of scalability. Isilon claims that, once racked, a multiterabyte cluster takes less than 10 minutes to configure and set up, and capacity can be added in less than 60 seconds with no downtime.
Our first test was to check these claims and evaluate ease of use – in particlurar the speed with which a cluster can be configured and deployed.
A wizard-based command-line interface helps set up the initial configuration; an alternative is to use a front-panel screen and controls on the device itself. Although the front panel can be disabled to prevent prying fingers doing any damage, it makes sense as a simple means of initial deployment – and as a visual status checker.
Considering how technical the operation is, in real terms the requirements for setting up the cluster are very simple – nothing more than adding a cluster name for the initial device, then adding IP address ranges for configuring the external Gigabit Ethernet interfaces, setting time zone and date, DNS and gateway address. It is similar to setting up a Windows XP client. Five minutes later you have a storage cluster. Adding extra nodes to the cluster is a one-click task.
“Ah, but can it scale?” I hear you say. Well, up to 88 nodes can be clustered, so up to 528Tbytes of data can be clustered in a single file system. All aspects are replicated as each node is added, with performance and capacity ratios maintained.
This results in aggregate throughput from a single file system peaking at more than 7Gbytes per second. So, yes, it scales. But is it robust? It would be unbearably aggravating to have 528Tbytes of data unavailable. But the IQ’s Flexprotect feature automatically rebuilds cluster nodes in the event of a cluster/drive failure.
We removed the fourth node in our four-node test cluster so that Flexprotect had to rebalance and rebuild the parity on what was a 3+1 cluster into a 2+1 (three-node). It did so in minutes with no human intervention required.
On adding back the fourth node it was automatically reformatted, coming back as a clean node. Then, the automated Autobalance data migration facility, followed by FlexProtect, not only brought the new node online but also striped the data across all four nodes automatically, so data was balanced across the rebuilt cluster. Data is automatically balanced throughout the cluster, removing the need to migrate content manually.
Isilon’s OneFS distributed file system provides the intelligence behind IQ’s clustered storage systems. It combines the three layers of traditional storage architectures – file system, volume manager and Raid – into one unified software layer, creating a single intelligent file system that spans all nodes within a cluster.
It also provides a single point of management, regardless of cluster size. With OneFS, each platform or IQ Accelerator node is a peer, so any node can handle a request, using Infiniband or Gigabit Ethernet for intra-cluster communication and synchronisation.
OneFS clues each node in on the entire file system layout and where each file and part is located. It controls the placement of files directly on individual discs and dramatically improves the performance of the disc subsystem by optimally distributing files across the cluster.
By laying out data on discs in a file-by-file manner, OneFS controls the redundancy level of the storage system at the volume, directory and even file levels.
Drives can be viewed just as if they were shared drives on a Windows-based PC. Moving data around is no more complicated than copying between, say, two PCs. You simply map a drive and commence file operations.
OneFS supports a wide range of file system and protocol types and operating systems. As part of the test we worked with Windows (CFS) and Unix/Linux (NFS) systems. From a user perspective, the system type is completely transparent, just as it should be.
Not that the established players in the storage market are simply treading water and waiting to be hit by a tidal wave of new storage companies. Traditional suppliers cannot afford to stand still.
Nigel Ghent, country manager for EMC UK and Ireland, says, “It is not in storage but at the information layer where the future battleground lies. For example, we have made $4bn worth of recent acquisitions and none of those have actually been in storage.”
That taboo word “revolutionary” may well be appearing on the EMC and IBM websites some time soon after all.
CV: Steve Broadhead
Steve Broadhead runs Broadband-Testing Labs, a spin-off from independent test organisation the NSS Group.
Author of DSL and Metro Ethernet reports, Broadhead is involved in several projects in the broadband, mobile, network management and wireless Lan areas, from product testing to service design and implementation.
Vote for your IT greats
Who have been the most influential people in IT in the past 40 years? The greatest organisations? The best hardware and software technologies? As part of Computer Weekly’s 40th anniversary celebrations, we are asking our readers who and what has really made a difference.
Vote now at: www.computerweekly.com/ITgreats