The Many Faces of Clustered Storage

Clustered storage is one of the hottest buzz words in the industry, but it means different things when different vendors talk about it. We all know that clusters are good for a lot of things, but the word is used to describe a wide variety of products that often aren't even in the same category. One needs to pay close attention when the C word gets tossed around.

Two recent product upgrades from Isilon Systems Inc. (Nasdaq: ISLN) and Symantec Corp. (Nasdaq: SYMC) highlight the potential confusion. Both describe their "clustering" products as meant for "next-generation data centers." Symantec issued Veritas Cluster Server One, which is meant to improve high availability and disaster recovery for physical and virtual machines. Isilon upgraded its OneFS distributed file system, which creates one "giant Z Srive" on a fully symmetrical storage cluster.

Vendors ranging from IBM Corp. (NYSE: IBM) and Hewlett-Packard Co. (NYSE: HPQ) to ONStor Inc. , ParaScale Inc. , Seanodes , and Caringo have also made recent announcements that featured the words "clustered storage."

"Clustered storage has been around for a while, but it's becoming more trendy," said analyst George Crump, founder of Storage Switzerland. "It is being offered by lots of vendors, and most of them mean they're creating a single pool of storage from a bunch of hardware boxes. Others use clustering to improve disaster recovery. But it has become a buzz word without a clear definition."

On the high-availability and disaster-recovery side, clusters can provide insurance that systems will keep operating and data will be available in case of a failure as data center infrastructure grows. "For most data center managers, the enemy is complexity," said Jason Nadeau, director of product management for Veritas Cluster Server product family. "They want to move to a distributed architecture, they want to virtualize, they want to use x86 platforms. But traditional clustering solutions break when you have thousands of virtual machines. They just can't scale."Symantec says its new software, an enhanced version of its Veritas Cluster Server, uses a client-server architecture that can handle up to 256 nodes per cluster and provide multi-tier high availability to ensure fast recovery by tying together different components of an application (database, applications, data, a Web front-end). The upgrade is designed to improve the management of virtual environments and is aimed at applications from SAP and Oracle, rather than databases. The company claims its approach to disaster recovery can cut capital expenditures and that a less-expensive IT staffer can use a dashboard to shut down, restart, or move applications, reducing operational costs. It costs $995 per CPU.

Isilon, perhaps best known as a clustered storage vendor, means something very different when it talks about clusters. It means nodes that contain high-speed front- and back-ends, multicore processors, 4 GB of cache, and swappable disks, power supplies, and fans. To Isilon, three or more nodes make a cluster and the storage capacity is treated as a single pool of storage.

The company this week released an operating system upgrade that's designed to boost speed and capacity. "OneFS 5.0 creates a single file system that manages everything, no matter how many nodes. That saves massive amounts of management and operating costs," said Jam Wampold, senior director of marketing and communications.

One big change is the ability to work with multicore processors. Many of the company's IQ cluster nodes have been shipping with dual- and quad-core processors, which have basically been sitting idle. The new symmetric multi-processing support takes advantage of the sleeping cores for an instant performance boost. Another boost comes from a new Accelerator-X node that uses two quad-core processors and 32 GB of memory for faster throughput. "We're taking file services into a performance arena that traditional NAS can't handle," Wampold said. The company offers 210 Mbytes/s of throughput per node and 20 Gbyte/s for a cluster, and it can scale up to 2.3 PB in a single cluster.

Wampold cited photo company Kodak as a customer that has benefited from the new capabilities. Kodak was using 360 individual volumes of 1 TB each to handle the uploading of images to its EasyShare gallery. That required a lot of management to create and maintain," he said. "Now they have more than 8 PB and are still using only two storage administrators." The company also will soon name AppNexus, a cloud computing service provider, and Squarespace, a provider of Website publishing and online content management software, as new customers. Its IQ X-Series also has been certified by VMware to work with its ESX Server 3.5.Clustering for servers and storage isn't new, but the need for increased throughput and storage capacity forces vendors to keep upgrading systems, notes storage analyst Tom Coughlin, president of Coughlin Associates. Isilon and Symantec are both using clustering to extend their systems to offer higher performance and more capacity at a lower cost and with greater flexibility. "The Isilon system focuses on their core market of content serving while the Symantec product is aimed at creating a cost-effective data recovery and protection platform," he told Byte and Switch in an email. "Both products extend each company's core strengths and target markets."