Network storage's sad state

For a couple of weeks in January, the storage bulls were off to the races. But after staging a mini-run, they ran out of gas when reality set in. They realized the blunt fact that short-term prospects for storage subsystem companies are just not very good.

Despite developing amazing technology, the storage industry has descended into commodity status. Past margins posted by storage companies are simply that--past. The cost per megabyte of storage capacity in high-end subsystems has dropped in the last year from approximately 10 cents to 15 cents a megabyte to roughly 2 cents to 4 cents a megabyte.

It is possible that the situation could rebound, but not in an environment where customers remain reluctant to make large storage purchases.

Another worrying sign is the exodus of employees from leading storage-industry companies. While EMC, Network Appliance and Veritas would have us believe they have simply been prudently cutting nonessential employees, the truth is that many senior, experienced employees, disillusioned with underwater stock options, are bolting for the exits. And the longer the current market conditions persist, the more the talent will walk.

The macroeconomic conditions have also contributed to the problem. There's no question that corporate layoffs have hurt storage networking. The amount of storage a corporation needs is directly linked to the size of its payroll. It stands to reason that a company in the middle of reducing the size of its work force is less likely to expand storage capacity than a company that is adding staff.

Many claim that storage is a strategic market, but the truth is that it's a dependent one. The heady days are long gone, and customers have since found better ways to manage their data and stretch existing storage capacity. What's more, storage sales follow the sales of systems and applications, especially line-of-business applications such as relational databases. And in light of dim projections from Sun Microsystems and Oracle, it doesn't appear that there is going to be much growth in either systems or applications.

In light of dim projections from Sun Microsystems and Oracle, it doesn't appear that there is going to be much growth in either systems or applications.

The situation might be better if the cost of network storage products were lower. Both network-attached storage and storage-attached networking technologies have excellent potential to solve serious IT operating problems. But network storage products, particularly SANs, are relatively expensive to purchase compared with server systems. Forget about the huge cost of ownership benefits; there is no way most IT organizations can justify spending $80,000 to $100,000 on a SAN for a couple of $16,000 Windows 2000 servers.

This is why most SAN-related revenue comes from connecting Unix systems to storage; the cost of Unix systems is higher than that for Windows servers, and, by comparison, the cost of a SAN is easier to justify.

The installation of a SAN is another matter. This is more like a forklift upgrade than a "slip in a little change here and you're ready to go" upgrade.

A customer would need new host adapters and cabling as well as to deal with the transfer of data from SCSI storage to Fibre Channel storage. For that reason, IT organizations are less likely to make a change to SAN technology until they upgrade their servers. And in an environment where customers are trying to stretch the usefulness of their current server systems, SAN installations will be held hostage to a thaw in IT budgets.

Ethernet-based storage will certainly take down the cost of SANs, but it has been disappointing to watch the snail's pace at which the industry is moving to deliver Ethernet based storage solutions. The selection of iSCSI as the technology of choice has forced the deployment of new TCP acceleration chips. That's fine for companies already developing iSCSI chips or with deep TCP technology they can leverage. For everybody else it means added time and cost.

Against that unsettled backdrop, here's my thumbnail sketch of how the key players in this market are faring:

McData
Trying to stay relevant by suing Brocade Communications Systems is a bad strategy. McData's SAN Navigator acquisition pits the company against the SAN software world and will drive potential partners to work with Brocade, QLogic and Inrange Technologies. In addition, McData's focus on the high end is going to raise expectations to unsustainable levels that the company cannot satisfy over the long haul.

Brocade Communications Systems
There is no indication that Brocade will be able to make the transition to iSCSI, when that happens. But its new 3200 8-port switch is the smartest product the company has introduced in years.

EMC

Hitachi has feasted by cloning and improving on EMC systems. But what will Hitachi do when EMC's not there to build systems anymore?

It's unclear who the new king is going to be, but it probably won't be EMC. The testing and quality levels that allowed the company to dominate are too high for it to survive in the post-crunch world. EMC's subsystem controllers will be supplanted by more efficient--and more valuable--functions running in network boxes. Still an extremely important and influential company, EMC is running out of time to shift its business emphasis.

Hitachi
Hitachi has feasted by cloning and improving on EMC systems. But what will Hitachi do when EMC's not there to build systems anymore? It's even less prepared to deal with a changeover to network value than is EMC. In true Hitachi style, the company will try to milk the current situation for all it's worth and then shut it down, following the same scenario it followed in the mainframe computing business.

IBM
IBM runs third behind Hitachi and EMC, which means its Shark business is at risk. With IBM's disk drive business faltering, there are not a lot of spare resources available to bail out Shark. Instead, IBM is putting its weight behind storage software. This won't succeed, but it will be interesting to watch for the next four to five years. Although Big Blue won't dominate any of the markets it enters, the company functions as a marketing catalyst that helps create new opportunities for everybody else.

Sun Microsystems
Maybe Sun will see more success under storage head Mark Canepa. He seems to get the point that storage is a real business opportunity and that the whole storage world has been setting its sights on the unprotected Sun market for the past decade. The Hitachi manufacturing deal will be successful for Sun in the near term. Eventually, Sun will need a new marketing strategy if it wants to be a serious storage player. But given the company's identity fixation, that may never happen.

Emulex
The host bus adapter arena is turning into a gritty little business, and iSCSI will make the situation worse. Margins are going to get squeezed, and there is little that Emulex will be able to do about it. Short term OK, long term weak.

QLogic
Similar to Emulex, but in better shape through a more diverse product line in SCSI products and switches. QLogic will be able to offer cross-product pricing and delivery options that will be difficult for any of its competitors to match.

Adaptec
Out of Fibre Channel, but back in the game with iSCSI, Adaptec may be able to regain prominence in the I/O industry. If the company succeeds, it will be in stealth mode, as Adaptec does not have the marketing leadership to make a big deal out of anything.

Alacritech
An early leader in TCP offload technology, Alacritech may be able to win business in the short term. It will probably wind up as one of the more important acquisitions made by either IBM, EMC, Sun, QLogic, Network Appliance or another company.

Network Appliance
Still the leader in network-attached storage with the most valuable software in the market, Network Appliance is trying to adjust to shrinkage. Its marketing capabilities are suspect, and the company seems to have had a problem finding its identity during the recent market maelstrom. Of all the larger mainstream network-storage companies, Network Appliance has demonstrated the best ability to innovate and could be a big winner as iSCSI comes out. But that success would be at the expense of its DAFS initiative (Direct Access File System, a new file-access protocol designed to take advantage of standard memory). Still, it's probably better to have two viable strategies than none. Short term OK, long term still very good.