Troll;
The forge is hosted on two dedicated servers at our hosting facility in the SF area.
As much as I appreciate your troubleshooting, there are many different variables in play here. To not bore everyone, I'll only touch on the most important:
- In the JVM, there are only about 200 threads available. You can think of this analogy like a bus that only has 125 seats available. Once everything's filled up, you're either going to have to stand up and wait for a seat to become available, or wait for the next bus to come around the corner. To add a little more insight, there are only 256 threads available by default in the Apache web server.
- As more people hit the forge, there is more load on the server. (the bus goes slower). Eventually, enough people have their requests filled, and the bus will speed up and get back 'on schedule' again.
While I understand that this is frustrating for everyone, there is only so much that we can do from an application management standpoint -- we can add more threads, however, without additional memory, that would allow more people to access the forge, but everyone's access times would be horrible, as there would not be enough memory to serve everything from memory, and a lot of swapping would occur. Likewise, we can remove some threads, and the forge would be faster, but, we'd introduce more issues (network speed, cpu load, jvm constraints).
The best that I can do is reassure everyone that I am aware of the situation, and that I am actively monitoring it. It would be rather unfair for me to just restart the application server every time the load gets high, as that would interrupt what those 100+ people are currently doing, and then there would be more load as they tried to reconnect.
Ian
