Performance Improvement?

[StangFlyer]

Well, I'm glad to hear that most of you seem to be seeing good performance from the site like I do when testing.

Fox Body, you're "independent" testing is good, especially if you are only connecting at 24 Kbps! You must not have a very clear phone connection being out in the "country" or something. I just did the same pages you did on my own dial-up account. On the 56K modem built into my laptop I connect to my account at 52.3 K/s. It took under 10 seconds to load all the same pages you just quoted. On average, it looks like it's taking 8 seconds for site pages to load. Going to the homepage of MW only took 6 seconds (for example). Saturday afternoon I was an hour away from here and I used someone's cable modem to surf around the site. Load times were 2 or 3 seconds per page. Of course, most of the time I am on the actual internal network with the servers. So, I'm hitting them at 100 Mbps and loading is almost instantaneous for me. The only delay I ever see is from the ads on occasion that get loaded from the ad providers we work with.

Mustang Works now lives on a large scale hosting environment consisting of 16 dual processor IBM servers with multi-drive SCSI raid arrays. The servers are on a multi-100 Mb network (front side and backside networks) using all Bay Networks enterprise level switching equipment (gigabit interconnectivity) and goes into a proxy/firewall system that connects to our fractional T3. Eight of these servers are web servers that are configured in a cluster behind Alteon hardware load balancers, and pull content from central file servers. They function as one to spread the request and serving load across all eight servers for performance, high scalablity, and redundancy/failover. Our load balancers can also handle 1024 connected web servers and 200,000 requests each second per box too. So, we can grow and scale at a whim now. We have default machine images set-up that can be used to load machines with our default web server configuration. With them we can load additional web servers and have them online in 15 minutes per box. Additionally, we have multiple, dedicated database servers with each having seven drive U160 SCSI raid 5 arrays that are SMOKING fast. This "web farm" has been slowly built over the past six months and is now done. It's an enterprise level, scalable web solution not only meant to power MW, but other projects in the works and client sites / projects of MW's parent company, McClain Technologies, Inc.

So, when I said I didn't know what else to throw at MW to improve performance (aside from perhaps a slightly lighter weight site design)... I wasn't kidding! If you'll recall back last year when we had the big Cobra abuse thread fiasco and were suddenly receiving millions of page hits per day (at its peek, 1300 connection requests per second!)... well, if something like that ever happened again we could now "scale" up to handle that kind of onslaught of traffic.