Storage Tanked?

Monday Oct 27th 2003 by Clint Boulton

IBM has revealed the first fruit of its Storage Tank product; competitors question its proprietary nature from the outset, while some analysts see the product's promise.

Competition in high-tech can take some pretty funny twists and turns. When IBM first dropped hints of its plans to drop a "Storage Tank" onto the industry a couple of years ago, analysts agreed at the time that it would be a strong assault on competition like HP , EMC , and Hitachi Data Systems .

To be sure, when Linda Sanford, then IBM's vice president and group executive for the storage subsystem group, said in November of 2000 that Storage Tank would be the fix for problems associated with operating a heterogeneous storage network made up of equipment from mixed vendors, the solution appeared to be the Holy Grail for customers.

It meant IBM's system would work freely with products from EMC, Hitachi Data Systems, and others that trade in storage infrastructure. It would also mean total interoperability, as Sanford told the public that Storage Tank would bring unity to all the devices used in a storage network, regardless of vendor origin.

But when product managers at those companies saw the way Big Blue presented the finished product a couple of weeks ago, they weren't all that impressed. They called it a large, proprietary file server that seemed geared to compete with smaller storage software and utility computing specialist VERITAS Software , and not the major systems vendors.

So what is this Storage Tank? That was the moniker for technologies developed over the last five to six years in the company's Almaden research facility, but recently it has taken definitive shape in a product called TotalStorage SAN File System, a product designed to let customers share billions of diverse files and to provide one medium of control to manage storage devices and data, as opposed to multiple control points.

In short, IBM would like this product to be the be-all, end-all file system that helps customers better manage their data, which is zipping through the network at multiple points; the bigger the network, the greater the need for something like IBM's new product. But the product doesn't work with just any device or software, as promised a few years ago; it works with Shark hardware and with IBM's AIX version of the Unix operating system, as well as certain Windows operating systems. To be fair, IBM has pledged to offer support for other vendor's products next year.

Bruce Hillsberg, director of storage software strategy and technology at IBM, maintains the point of the SAN File System is to help companies with large data warehousing needs and customers deploying grid computing environments become more competitive by providing an easier way for administrators to manage the massive amounts of data that is stored.

Competitors Focus on Discrepancies Between Promises and Product

Competitors seized upon this gulf between what was promised and what was delivered.

Ken Steinhardt, director of technological analysis at EMC, states the SAN File System is a "far cry from what we expected."

"Their original position in November 2000 was that they had specific plans for an all-encompassing virtual environment that would address every major storage product," Steinhardt told "But this looks like a proprietary IBM file system that doesn't integrate with third-party storage. Sure, it looks interesting, but it looks different from what was originally announced."

Steinhardt also pointed out that the arrival of the first fruit of Storage Tank comes much later than it was originally announced. "In 2000, they said it was right around the corner, but then the quotes changed. Then it was the end of 2001, and then mid to late 2002. Now here it is, and it's near the end of 2003 and the original announcement has been whittled down to something substantially smaller."

Steinhardt also questioned whether or not the file system is something customers will actually want. "It looks interesting, but would I want a distributed environment that may suffer from latency to go into someone else's storage? You should never try and build a proprietary file system."

Page 2: Analysts See Potential, But Have Questions

Analysts See Potential, But Have Questions

Analysts, however, were not so quick to dismiss IBM's new technology.

Enterprise Storage Group Analyst Steve Kenniston finds promise in SAN File System, noting that there are many, many applications that could benefit from this type of technology, and that, when paired with IBM's SAN Volume System and Controller, could yield some solid virtualization capabilities. He sees SAN File System being deployed in a lot of niche places where large file types exist — something that could work well under Oracle's Real Application Cluster (RAC) environment, for example.

"As far as the competition — HP is working on something, but the real thing I believe is what does HDS and EMC do?" Kenniston wonders. "Both big storage companies [compete with the high end that IBM has in Shark] but have no file system software or "real" virtualization software that will scale like Tank. They don't own the IP for it and would have to leverage somebody else's system. That said, they need something, and due to what it takes to build one, it MUST be a buy decision, and there isn't a lot out there. SGI has one and their storage business needs a boost...So it is a good question."

In the meantime, Kenniston feels the more appropriate competitive comparison with SAN File System might be between IBM and VERITAS, which he says is the leading file system across all platforms. EMC's Steinhardt agrees that his company does not seem to be a competitive target with the SAN File System.

Scott Gready Director, Storage Software Technical Office, HP Network Storage Solutions, also questions what problems IBM was trying to address with the SAN File System.

"SAN changes have really gained mainstream acceptance as a way to deploy storage," Gready told "What customers are interested in is how can I deploy SANs in as simple a way as possible. When you create a product, you need to ask the customer: 'What do you want your SAN to do? Do you want the SAN to reliably deliver storage capacity, and do you want to put a lot of array controllers on a network to get lot of flexibility, or do you want just a simple array controller?' Customers don't want SANs to manage their files — they rely on an OS (operating system) to do that."

Accordingly, Gready said, OS vendors consider the file system to be their domain. They require a lot of new components that have to be OS-specific, such as a host agent, installable file system, or meta data server, he said. As a result, "I think the OS vendors are the real ones IBM will be competing with here."

"We don't have customers asking for a SAN to manage files," Gready said. "Customers are worried about how to manage their data in view of compliance regulations — HIPAA, etc. If the problem IBM is trying to solve is file sharing, then NAS has come a long way in terms of becoming a cost-effective file solution. It's become the industry standard mechanism, so you don't have to install proprietary file systems on every server."

Forrester Research analyst Anders Lofgren sees the value in the SAN File System, but he isn't so sure customers will be willing to migrate to the IBM architecture.

"The answer to that depends on where customers are — how far along into SAN deployment they are," Lofgren told "New customers may not need it. I think this will be most attractive to customers who have a number of SANs and want to combine them. They have designed this SAN so you don't have to rip-and-replace infrastructure, which is key. You can install it along existing file systems and move application data over into it to use existing assets."

Lofgren said the question then extends to the long term: "Where is it going? How does the file system they replaced with IBM interact with applications? Can they quantify that with hard numbers? This is definitely a new concept — it aims to meet the challenge in describing and demonstrating quantifiable data."

Lofgren added that while he doesn't believe competitors will answer the problem with the same approach, he knows SAN File System is not for everybody. Instead, he sees virtualization picking up at a much faster clip. "It's just much more approachable to the end user than a global file system."

One thing's for sure, vendors and analysts alike are clearly interested in seeing how the first fruit of Storage Tank fares on the warpath.

Story courtesy of

Back to Enterprise Storage Forum

Mobile Site | Full Site
Copyright 2017 © QuinStreet Inc. All Rights Reserved