Storage is the last bastion of overpriced, underperforming hardware. In networks, processors and servers the radical economics of semiconductors has created a lean and mean industry prospering on margins less than half of what the old-line — and now mostly dead — mainframe and minicomputer companies considered their due.
A tech rule of thumb holds that innovation proceeds in three 5 year cycles.
- From invention or lab demo it takes 5 years to a commercial product
- From commercial product to $1 billion in sales is another 5 years
- From $1 billion is sales to commoditization or replacement in the last 5 years
Under that model the storage world is way overdue.
The seminal RAID paper was published in 1988. The first hardware RAID 5 products staggered to market around 1992. Performance and reliability were pathetic, but vendors responded well to repeated floggings by data-losing customers and matters improved rapidly. Except for the Hancock-impaired IBM all the system vendors had competitive RAID systems by 1998. RAID arrays had become not just a billion dollar business in less than 10 years, but a multi-billion dollar business.
So 8 years later, where is the commoditization of the RAID array business? Or the disruptive technology that blows up the industry? Sadly it looks like it will be a while.
The dynamics are pretty clear.
- Storage systems are one of the few highly profitable hardware businesses and every hardware vendor (with the head-scratching exception of Sun) wants to protect those margins with all manner of “added-value” lock-in
- As the very smart Richie Lary has pointed out, efforts to “open” storage create a first mover disadvantage: everyone moves onto your turf and you can’t move onto anyone else’s
- Enterprise customers are very conservative in storage purchases and it isn’t obvious what other segment (consumers?) has the financial clout to drive new technology into the enterprise
For some time I’ve looked to Intel to break this logjam by applying their competitive ferocity and technical chops to creating commodity RAID-on-a-chip, much as they’ve done with low-cost graphics. Thus my disappointment when Intel announced their assimilation by the Borg. Is there no one to challenge the storage oligarchy? Young Skywalker? The open source community? Someone?
Obviously Intel is pre-occupied with a bigger problem than hawking storage to their white box resellers: AMD is kicking their butt in the server business and fixing that has to be job one. Hans Geyer, Intel’s Storage VP, speaks like a man who has had ambitious proposals killed, not once, but several times:
Geyer said that Intel is working with EMC technology instead of developing the technology itself because RAID-5 and other storage-related technologies are not trivial.
“The most important thing is to have a complete solution, and a complete solution does not mean having three programmers and a year later you have the technology,” Geyer said.
Actually most techniques invented almost 20 years ago are trivial today. Why is RAID any different?
An even bigger question: why does Intel think selling the costliest SMB storage out there against Dell’s identical product is going to be a win? Intel’s storage group needs a marketing transplant, ASAP.