Isilon made its first major product launch since EMC acquired the clustered NAS vendor for $2.25 billion late last year, rolling out new versions of its performance S-Series and capacity X-Series NAS, and upgraded its OneFS operating system and SyncIQ replication software. DDN said it will soon enter the enterprise NAS market with its NAS Scaler system.
Isilon bumps speed and feeds, enhances software
Like previous Isilon S-Series and X-Series systems, the S200 and X200 are 2U models and support clusters up to 144 nodes. The S200 and X200 are compatible with Isilon’s previous systems, so they can be added to any current customer set up.
The S200 upgraded from Intel Harpertown to dual Intel 2.4GHz Westmere chips to speed performance. Isilon increased the number of maximum drives in the S200 to 24 from 12 in the previous S-Series, bumped up maximum node capacity to 14 TB from 5.4 TB, and also increased maximum memory to 96 GB from 16 GB, maximum IOPS to 1.4 million from 703,584, and maximum throughput to 84,960 MBps from 44,640 MBps. In addition, the S200 went from 3.5-inch SAS and solid-state drives (SSDs) to 2.5-inch 6 Gbps SAS and SSDs.
The X200 upgraded to the Intel E5504 Nehalem CPU from the Harpertown chip in the previous X systems. The X200 remains at 12 3.5-inch SATA or SSD drives, but doubled maximum node capacity to 24 TB. It increased maximum IOPS to 309,312 from 288,000, and throughput to 35,712 MBps from 28,800 MBps.
Isilon added 10 Gigabit Ethernet (10 GbE) support to the X200. The previous S-Series systems support 10 GbE.
Isilon’s OneFS 6.5 operating system supports native CIFS for the first time, as well as NFS 4.0 and Kerberos NFS3. Isilon has also integrated its SnapShotIQ software with SyncIQ 3.0 to speed replication performance, and added capacity monitoring to its InsightIQ 1.5 analytics application.
Isilon also modified its SmartPools tiering software to use SSD for file data. SmartPools previously only read metadata from SSDs. Now the Isilon systems can use SSDs as a tier for low-latency applications such as virtual machines and databases.
List price for the S200 starts at $57,569 per node and the X200 starts at $27,450 per node. OneFS 6.5 is standard with any S200 or X200 purchase, and SyncIQ has a list price of $4,950 per node.
Beta customer gets performance boost from CIFS support
Ancestry.com has approximately 3 PB of data on 18 clusters of Isilon X-Series systems. While it hasn’t upgraded to the new hardware yet, its storage team beta tested Isilon’s new operating system. Travis Smith, Ancestry.com’s senior manager for storage systems, said the biggest benefit is a performance boost from OneFS 6.5’s native CIFS support.
“They changed the way the CIFS stack worked,” Smith said. “With the older versions, the CIFS stack spun off a single thread per host that would access a cluster. Most of the time that was fine, but you would run into issues when you had a bunch of systems trying to get out a whole slew of small files. It would stack up requests in the queue and cause latency. Now the CIFS stack is multithreaded, so you don’t have as much queuing when you have hosts requesting tens of thousands of tiny files.”
Smith said he also likes that Isilon now uses SSDs for file data. He said SmartPools will come in handy when he's ready to add new systems to his current clusters, some of which include three-year-old hardware. “As we start to grow clusters and add new systems into our older clusters, SmartPools start to make a lot of sense for us,” he said.
DDN provides ‘traditional’ NAS
DDN positions NAS Scaler as a content platform for big data and rich content workflows. Similarly to Isilon’s systems, NAS Scaler can scale for performance and capacity separately. It supports up to 16 cluster nodes with a maximum cluster capacity of 2 PB and a maximum file system size of 256 TB. It also includes a data protection suite that includes asynchronous replication, snapshots and clones.
NAS Scaler starts in a 6U configuration that scales to 120 TB maximum (96 TB usable) capacity. It includes a 10U system with 192 TB usable capacity, an 18U configuration with 384 TB usable, a 32U configuration with 480 TB usable, and a 52U (two 26UB boxes clustered) to 960 TB usable. NAS Scaler supports SSD, SATA and SSD drives.
DataDirect Networks plans iSCSI support and the ability to run NAS Scaler as a virtual appliance in 2012, said NAS Scaler product manager Anand Singh Bisen.
DDN sells mostly block storage systems for high-performance computing (HPC), and the private company claims it had $180 million in revenue in 2010 and expects to generate $240 million this year. It does have other file storage systems, but they handle mostly HPC applications while NAS Scaler is aimed at traditional unstructured data.
“We had a hole in our product portfolio,” Bisen said, “and that was lack of enterprise NAS.”
No pricing is set yet for NAS Scaler, which is expected to be available around the middle of the year.
“This [NAS Scaler] is for people who have a serious amount of data,” said David Floyer, chief technology officer of the Wikibon research and advisory organization. “It’s aimed specifically at throughput, not large IOPS. You would not typically use a small NAS device, and NetApp doesn’t scale up that far.”
Enterprise Strategy Group senior analyst Terri McClure said NAS Scaler is more optimized for small file transactions than DataDirect Network’s other file systems. She said NAS Scaler will compete with Isilon’s S-Series in some cases.
“This gives DataDirect a high-performance, small file transactional play,” she said. “Markets like media/entertainment and seismic data typically have classic NAS applications that require file solutions. DataDirect could only satisfy part of their needs with the file systems they had. NAS Scaler is good for certain workloads that would go against Isilon’s S-Series while [other DDN NAS systems] GridScaler, ExaScaler and xStreamScaler would be used for higher throughput.”
Big deal about big data
Isilon and DDN are among the vendors throwing around the big data tag for its storage products, although they address only part of what really constitutes big data computing.
“A lot of people are trying to climb on the big data bandwagon,” Wikibon's Floyer said. “This [scale-out NAS] is large amounts of data -- big blobs -- but not big data in the sense of using Hadoop or scale=out mechanisms for distributed data over a large number of places. They’re more about large amounts of data in one place than truly big data.”