Enterprise performance
I’ve been working with a lot of enterprise software during my whole career and it keep surprising me that almost no enterprise tool has any consideration for performance. I understand that the pricing is usually per CPU, which makes performance a less interesting feature for sales. But it goes too far.
How is it possible that you are advised to run 1 Oracle WebLogic Portal instance on 1 CPU? I know portal is not comparable to PHP, but a factor 1:1000 is a bit to much for my liking. And Oracle is not alone in this: IBM, SAP, all of these are slow and slower. And it’s weird since we (the IT boys and girls) have fixed this a long time ago.
There is a beautiful concept called HTML caching. Yes, I know: that’s not nice, architectural incorrect, not flexible enough. Indeed, heavily personalised pages might not be the primary candidates for full page caching, but most of the websites out there still use a fairly static homepage. Cache that page!
At Componence we’ve (finally) delivered a standard caching solution to put in front of Oracle WebLogic Portal (or BEA Weblogic Portal and the result is shocking. The performance improvement is upwards of 1:20. Give it a try: http://www.wk-vet.fr. Yes, that’s portal, and it’s dynamic.
Dear enterprise architects, please stop being such principle asses and apply the most common fixes to your performance problems.
Hi Martin,
Exactly ‘principle asses’. In this case even our internal technical guys thought it was a bad solution
For more elaboration and images:
http://hhvo.wordpress.com/2008/11/01/continuously-pursuit-to-optimize-weblogic-portal-performance/
He-he )))
I liked “principle asses” as well.
I think there may be a good reason why those guys in BEA, Oracle, IBM, etc. don’t do this: they just stopped development of real portals some many years ago and doing only “academical” work, i.e. implementation of frameworks. So it’s like teaching about engineering when the last thing you really engineered was 10 years ago or maybe even “never”.
@Alex
I can see that caching a full page kills the portal functionality, which was the argument of the technical people involved. But sometimes you don’t need the portal functionality, sometimes a page is just static. That’s a point often forgotten by the ‘technically smart’ people.