In continuing my dialog about my insights at the Spring SNW 2008, I did multiple briefings on Tuesday, April 7, but none was more insightful than the sit-down meeting I had with Fusion-io’s self proclaimed Chief Mind Bender Rick White. So while I plan to come back and cover material from some of the other briefings I had on Tuesday, what he revealed to me about what Fusion-io is doing and has on its roadmap is not just disruptive, it’s disturbing. Granted, a lot of what he told me he promised that I not reveal but he told me enough public information to get my head reeling with possibilities.
First, why is it that I think Fusion-io is so disruptive? Well, over the last few years the storage world has been going ga-ga over cheap SATA drives and placing data on the right tier of storage. In fact, one of the best new terms I have recently heard SATA referred to as is the new Value Tier as Permabit now calls it. But Tier 1 providers have merrily been humming along selling million dollar storage systems for Tier 1 apps as Tier 1 storage has largely gone unchallenged for the last 20 years.
Oh sure, a few challengers to EMC have come along since they originally introduced the Symmentrix but the only vendors who have successfully challenged EMC in this space are 3PAR, HDS, IBM and arguably NetApp. But that was only after great cost, new innovative designs and years of investment to grab some market share. Despite all of this investment, I continually hear about organizations trying to balance the cost and performance issues of their mission critical applications running on the 10K and 15K FC disk drives that these systems contain.
Granted, more of these Tier 1 providers are putting solid state disk into their systems but usually at an exorbitant costs ($100/GB). So while organizations can certainly justify deploying some of this technology for some of their applications, they certainly can’t go hogwild in deploying it.
Well, suppose some new vendor who is unaware of these sweet high end deals going on in the background now comes prancing along and offers:
- 320GB or 640 GB of solid state storage for an MSRP of $10,000
- Installs the storage on a single PCI bus of your existing server
- Gets 1,000,000 IOPS per second
- Microsecond response times (yes microseconds)
Does not this begin to argue for the elimiination of externally attached storage systems and possibly all of staff supporting these systems in the process? Hello Fusion-io and hello game changing technology.
Now I am not purporting that everyone should unplug their Tier 1 EMC DMX, HDS UDS or 3PAR T-Class storage systems tomorrow. Far from it. Fusion-io still has to pay its dues by finding partners to put their ioDrive PCI-Express cards into solutions that meet end-user’s need, they need to complete qualification testings and they need to get some customer success stories behind them. But from where I sit as a former end-user, hold on to your pants boys and girls, because this is going to happen a lot faster than any veterans in the storage industry are prepared to adapt or I think are even anticipating.
Fusion-io’s ioDrive is so disruptive that it is disturbing because it changes the whole paradigm of Tier 1 storage at it stands today. Fusion-io now delivers more performance, much less complexity, and an unheard of ROI at a fraction of the cost of these other approaches. Now users are rightly asking themselves, “Why build out an infrastructure that costs hundreds of thousands or millions of dollars, takes months to implement and time-consuming administration when they can pay $10,000 (or $20,000 if you want to mirror data across two cards using Veritas Storage Foundation), slap these babies in your existing server and say, “Problem solved! Next?”
Now I know I don’t know everything there is to know about solid state disk and more than likely Fusion-io ioDrive cards has its share of problems. But I just learned today from Rick that its ioDrive has technology embedded in it so that 3rd party software vendors can monitor the ioDrive for anomalies and proactively replace the ioDrive before it goes bad. Not bad for a technology that many are saying is not ready for prime time. Because if that is not an indicator is is nearing ready for prime time, maybe we ought to subject tape and disk to the same criteria to which we are holding the ioDrive. If anything, sounds like the ioDrive might have already leapfrogged tape and disk in its first release (or whatever release it is on).
And then let’s not even even forget to bring up how disruptive this is to other players in the Tier 1 FC SAN space, specifically data protection, but for different reasons. SSD may actually turn out to be legacy backup software’s new best friend. I was just talking to a user today who needs 11 hours to backup and recover his Oracle databases using backup to disk. Not bad considering where he is coming from when he felt blessed if could complete backups to tape at all. But his Oracle database is on traditional Tier 1 storage. Move that database up to Tier 0 and do a backup to a disk target and he might just drive his backup and recovery times down to under an hour using his CommVault backup software.
Is Fusion-io going to rock the storage world tomorrow at the high end? Absolutely not and more work needs to be done on its end to get its cards ready for primetime. But HP already has an HP-branded version of the Fusion-io card and do you think it is any coincidence that HP is combining its server and storage divisions and just recently bought Lefthand Networks which converts off-the-shelf servers into storage systems. Hmm, makes ones think, doesn’t it? Commodity HP storage plus high performance Fusion-io ioDrive card equals 1 million IOP HP storage system with Tier 1 performance for an entry level price of what? $50,000 with HP support behind it? Sounds pretty disruptive to me.