NetApp punches Isilon right in the scaled-out clusters
No short-stroking necessary to crush low hanging fruit
NetApp doesn't do scale-out clustering, everyone knows that. Oh but it does, and it has whacked Isilon firmly in its scaled-out cluster with a SPECSfs2008 benchmark that is 36 per cent faster using half the disks and 116 fewer nodes.
EMC Isilon  was the top SPECsfs2008 NFS dog with 1,112,705 IOPS recorded from a 140-node S200 cluster with 3,360 disk drives and a overall response time of 2.54msecs. That bested a Huawei Symantec N8500, which did 636,036 IOPS; an all-flash VNX 5700, which did 497,632; an IBM SONAS system, which did 403,326; and – almost laughably by today's SPECSfs2008 standards – a NetApp FAS6080, which did 120,011.
Now NetApp has clustered 24 FAS6240 nodes together in an ONTAP 8.1 cluster and run it more than 10 times faster than that lone FAS6080, achieving 1.51 million IOPS with an overall response time of 1.53msecs. There were 1,728 450GB, 15,000rpm disk drives and the FAS6240s were fitted with Flash Cache.
NetApp's UK Systems Engineering director, Jeremy Wallis, said: "At list price we were two-thirds of the Isilon price."
He said that the E-Series arrays were positioned for specific niche markets, such as high-bandwidth business analytics, whereas ONTAP clusters were for core data centre applications. NFS was the main requirement and the company has no plans to run a CIFS version of the SPECsfs2008 benchmark, he said.
Simply put, the ONTAP 8.1 cluster was more efficient at driving its single tier of disks, with exported capacity of 288TB and no short-stroking, than Isilon with its 3,360 disks and a 864TB exported capacity.
El Reg wonders how an all-flash VNX 5500-F  would do? Wallis agreed that it would be possible to build an all-flash system that might surpass the NetApp result but it would likely come at a huge cost and so be ruled out of real-world use.
He added that the scaling in IOPS as NetApp built the cluster from 4 nodes, to 8 and so on in 4-node chunks, was linear with no sign of any performance drop-off. Presumably then it could continue scaling out as there is no architectural limit set at 24 nodes.
NetApp couldn't have a better launchpad from which to impress upon potential ONTAP 8.1 customers that it has finally integrated the Spinnaker technology into ONTAP with sparkling results. Cluster mode works. ®