servers fixed, ready to go.
July 22, 2007 - 8:31 AM
For a while I was very uncomfortable leaving my current job when I was. We have been having some load problems with our systems and we were frequently experiencing crashes of our application servers. Being the main architect of our systems I felt really bad leaving because of the fact that I know more about those systems than anyone.
After working non-stop for about two weeks on the issue we were able to get everything taken care of. We removed our old oracle 9i (dual 2.4gHz win2k, it could only give oracle 2GB of RAM, sad I know) database and replaced it with a nice new 10g box (2x dual core 3.0gHz RHEL 4.5, 8GB of RAM, all to oracle). We also recently replaced our DBA with one who actually knows what he is doing, bonus.
We also made some modifications to our application. The application itself already caches objects from the database but we didn't use caching anywhere else. Our system uses XSLTs for its view layer and we were loading the XSLT for every transform needed. Rather than loading the XSLT every time we now cache it for up to five minutes. Response times on our servers dropped from anywhere from 3-15 seconds down to 1-2 seconds serving about 300 requests / second (not much I know). The application itself was always designed to have caching implemented in a few of the key places (this being one) but I had no idea we would get this kind of performance increase from it.
Our servers have stopped crashing and not only that they are responding faster than ever and I am pretty sure we can handle as many as twice the number of requests we are (if not more) without adding any new hardware. That was my woot for the week.
