SANs drop the shackles

Computerworld Australia - IP storage made huge advances in 2005, with iSCSI reaching mainstream acceptance. Analyst firm IDC estimates more than 6,000 users worldwide have deployed iSCSI-based SAN technology.

The market is expected to experience 211% compound annual growth for the next three years, topping $2.65 billion in 2009. David Dale, chairman of the Storage Networking Industry Association's IP Forum in Australia, said most iSCSI deployments today are on Intel-architecture servers running Windows.

However, NetWare and Linux environments are also becoming popular for iSCSI, and solutions for small RISC servers running Unix are just starting to emerge.

"The driving force behind a transition from DAS (direct-attached storage) to an IP SAN is often high data growth and the need for operations efficiency. In these circumstances the legacy DAS environment becomes increasingly complex; backup/restore operations become increasingly unreliable, and the storage environment is unable to support the demands of the business," Dale said.

Since iSCSI is a SAN protocol, it was initially seen as an adjunct to Fibre Channel SANs.

"It was usually characterized as a less capable technology that would connect stranded servers into the corporate SAN. On the other side of this argument were iSCSI proponents who viewed it as a disruptive technology that would render Fibre Channel obsolete as a SAN interconnect. The reality lies somewhere between these extreme perspectives," Dale said.

To date, iSCSI and FCP (the Fibre Channel SAN protocol) have often turned out to be complementary technologies -- each having a distinct place in the IT infrastructure as a SAN alternative to DAS. Fibre Channel generally provides high performance and high availability for mission-critical applications, usually in the corporate data center. In contrast, iSCSI has generally been used to provide SANs for business applications in smaller regional or departmental data centers.

Because NAS and iSCSI transmit storage traffic over Ethernet, there was initially some confusion about the technologies.

From an application perspective, Dale said iSCSI-based SAN storage looks just like DAS and so it works transparently with all applications.

"NAS will usually work fine, but since it is a different paradigm it generally needs to be supported by the operating system vendor and qualified by the application vendor," he said.

Since snapshot functionality is so common with iSCSI, most deployments back up from a recent point-in-time copy, which eliminates the traditional backup window.

But second-generation iSCSI solutions will expand on these capabilities.

Dale said second-generation iSCSI hardware initiators add support for active-active multipathing configurations, boot-from-SAN capability, and improved CPU offload. In addition, second-generation iSCSI software initiators improve the support of high-availability environments with multipathing, fail-over and host cluster services support.

The biggest storage array in the world, which is currently under construction, will have capacity equivalent to a stack of iPods three times the height of the Empire State Building. Interestingly, it will be managed with common Ethernet networking tools.

The SAN will support a project called the Human Speechome Project -- for the MIT Media Lab --which is expected to archive and search 1.4 petabytes of data over three years.

The SAN is being built from commodity hardware and uses a 10GbE IP network for data transfer between the backend SAN and hundreds of servers.

Computing infrastructure is expected to be composed of more than 300 Hammer Z-Rack storage enclosures from Bell Microproducts, about 3000 SATA (Serial Advanced Technology Attachment) hard disk drives from Seagate Technology and more than 100 10GbE switches and 400 blade processors from Marvell Technology Group.

The high-throughput switches are needed for the storage I/O anticipated by researchers who believe they'll be processing 700TB of data during every 12-hour analytical run. To achieve the desired performance requirements, 150-drive stripes (aggregated virtual volumes) will be created using the native virtualization capabilities of Bell's Z-SAN. Protection against data loss will be delivered through RAID 10 mirrors (duplicate copies) of the raw video data, transform data, and metadata files.