I really like this blog entry by Vince Westin; it absolutely confirms a belief that I have held for sometime! Application vendors have shares in storage companies!
Poorly written SQL statements have probably driven more high-end array purchases than any salesman working for a vendor. Actually I suspect that poorly written SQL statements have driven more Enterprise IT purchases than anything.
It is actually nice to see a vendor rather than recommending an expensive upgrade actually doing the sensible thing and looking for the underlying problems.
Still perhaps Oracle purchased Sun because they absolutely realise that this is the case; they can now sell you both a nut but also a very expensive sledge-hammer to crack the nut.
I suppose if I have one big concern about Cloud is that if we allow people to treat resource as abundant and cheap enough to waste; will code-bloat and poorly written code proliferate?
I am convinced one of the major drivers of storage growth is that desktop storage is now so cheap; it can be wasted but the same is not yet really true of Enterprise Storage; a terabyte of 'Enterprise SATA' is very much more expensive than a terabyte of 'Desktop Storage' but try telling the average developer or user that?
Why optimize DB code when SSD is available to save the day?
I’m being facetious of course but this is the argument I got from app vendors in the past.
The rationale that SSD is likely an extravagant expense in most cases was lost on them.
Martin,
Several years ago, I was at a conference in Amsterdam, and the presenter covered ten things to improve database performance. Improving storage performance was only 8 or 9 on the list. The top seven were all improving the quality of the SQL and server-related settings. I think the problem is that writing tighter SQL is hard, but adding storage resources to boost performance is relatively easy.
Tony P (IBM)
It’s not just a problem about code.
Check those entity-relationship models and compare to transactions. You can also dig a lot of “gold” in there.
No wonder the code follows the same standards.
But I agree, the availability of ever growing resources is a good case for poorly wrritten (and benchmarked) code.
J
Surprisingly, or not.
As a storage salesman, I still get dissapointing responses from customers when I advise them that actually improving their database ‘code’ may actually be a more worthwhile proposition (when there is evidence of this).
One customer even travelled overseas to meet with a SSD vendor in order to discuss improving performance. I knew their SQL consultant, and we has convinced a million dollars spent on H/W would not fix their ‘I/O’ issues. You just can’t help some people.
Paul,
upgrading hardware should often be the last course of action; I’ve rarely seen improvements from changing hardware anything like the improvements made from fixing ‘code’. Throwing money at hardware…just say no!