I've been writing a story about a fantastically fast SSD for servers and PCs. Now that would speed up Vista. (Read about it on Monday.) It might be true; it might be fantasy.
It's slated to cost $30/GB, meaning an 40GB one will set you back $1,200 (c£600).
The O/S sees it as just another direct-attach disk. Well; load the Windows O/S into it and watch your PC's performance accelerate into hyperspace. Warp drive here I come.
Actually it's in beta test and it's not going to support Vista and XP until some time in the Jan-Mar period of 2008. But we hacks get to foam at the mouth in enthusiasm over new technology. Bear with me as I go with the flow.
I/O-bound and bloated code
Thinking about this card I've crystallised an opinion about Vista. It is a classic I/O-bound application. Microsoft's code saturates the memory-disk channel. What slows it down, apart from it being a terrific example of software bloat, is its dependency on disk I/O. Hybrid drives, ones from the likes of Samsung, with a flash cache promised help because the ReadyBoost feature in Vista can use them. However, early results suggest that they might put a slight spring in the step of a man with crutches but the sticks will still be needed.
The Vista PC with a hybrid drive becomes a slightly faster version of Hopalong Cassidy - better, but no cigar.
But try storing it on, and booting it from, and running it from, an affordable SSD and the software bloat disappears. Lumbering fatty turns into a sprinting skinny kid and your personal productivity rockets up.
The problem with disk I/O is that there is just one slow pipe between the disk and memory. This cannot be fixed; you can't simply add more pipes. it's the wrong thing to do with HDD technology. As far as Vista and similar I/O-bound applications (Quark Express 7 anyone?) are concerned the desktop PC's hard drive is the wrong storage technology at the wrong time.
As far as small and medium businesses are concerned, ones with relatively few and relatively small servers, Vista's (and Windows generally) dreadful multi-tasking performance needs VMware because Windows is so disk I/O-bound and so bad at application switchover. VMware makes the memory/disk pipe function much much better. It makes Windows in servers fly like Netware used to.
Has the hard drive had it?
Imagine what effect an affordable SSD could have on servers? The SME small server users could see their servers getting super-charged. They could avoid going to VMware or use it on direct-attach SSD storage servers and not have to invest in complex, expensive shared and networked storage at all.
What's needed to make Vista PCs fly, Windows servers fly, is the replacement of HDD technology. It's old, it's becoming passe, and applications are becoming too big for direct-attach disks. So, take a deep breath and start thinking about the effect affordable SSDs will have on your desktop and server systems' performance.
If the promise of this technology comes good then spending a thousand dollars on desktops and a sum in the low thousands on servers, could literally transform their power.
A heavily I/O-bound system could run 20 to 50 times faster - if, and it's a big, big 'if', if this technology that we'll describe on Monday comes good.
The card will cost $1200 in early 2008. The way prices are going it will cost $600 in early 2009, $300 in early 2010.
We're in a 3-way bind at the moment:-
- Microsoft cannot make Vista go fast when it is stored on a hard disk drive. Forget it. No way, and we're stuck with Windows.
- The hard drive manufacturers cannot spin their drives faster or speed up drive I/O much.
- And we're stuck with huge I/O-bound applications like SAP, Oracle, whatever.
But, just possibly, we are not actually stuck with hard drives; there might be an escape route, light at the end of the tunnel. That's if it's not too good, much too good to be true. It's the weekend. I'm going home to fantasise and come back on Monday to wake up and find it's all a dream.
But then ..... maybe not.