Site Loading Problems

Ric said:
Sorry about that. We got a spike in traffic after the FA Cup draw, and I wasn't around to sort it out.
Click to expand...
No need to apologize, this site so integral to how fans communicate it's gets frustrating sometimes. Thanks.
 
super_city_si said:
Need a new host. Its very unreliable, even now after something like the FA CUP draw it craps out.
Click to expand...

Our host is Amazon who run one of the largest websites in the world.

The simple fact is that no website in the world is able to cope with going from 50% of capacity to 150% of capacity in 20 seconds. Even Reddit, a global website which does five million pageviews a day, 11 full time server engineers and millions of dollars of funding cannot cope with this. Neither can anybody else.

Capacity is not the problem, if we were millionaires we could run 200,000,000 people on the site at any one time. The problem is the speed at which the users online number grows. The traditional way to handle this is to know that that traffic is coming and then scale the architecture at the backend up - so essentially you temporarily increase the capacity. This is what we do on match days or when there's an inkling of an incoming transfer and it's what every other website does.

We have automated systems which detect traffic flow on the servers and database that we use then automatically increase the capacity by adding new resources into our server farm. Unfortunately creating a new server, loading it with the programs and files that we need and putting it into the mix takes somewhere in the region of 90 seconds from dead to fully serving content. We have something that sits in front of all of our servers, tests the amount of traffic sent to each of them then works out the best way to send a request so that you get the quickest response from the site possible. When a traffic spike hits unexpectedly when one server reports that it is starting to overload, a new server gets created and put into the mix. The problem is that as the overloading server reports to the load balancer, the load balancer then shifts less traffic to it or more traffic to the other servers. Now instead of 4 full capacity servers running the site we have 3 full capacity servers running the site, a couple of servers working their way into the mix and one that is overloading. In a scenario when the traffic spikes, all that's happened is that you've shifted the burden to less servers which overloads them too until another reports it at capacity and the load balancer then switches all traffic between two, then between one then all servers are out of commission.

I'll compare site A and site B. Site A runs at a capacity to support 2000 users normally. Site B runs at a capacity to support 20000 users normally. A traffic spike for Site A is when an extra 500 users log on above what it normally online - this is a full 25% of what they are supposed to usually run. The same traffic spike for site B is when 5000 users come online which very rarely happens to anybody.

The obvious way to fix this is to increase capacity over and above what we do need and sit having lots of it doing nothing for most of the time. The problem with that is that capacity costs money, everybody blocks advertisements which pay for that capacity (and then tell us how horrible the message asking them to actually make a contribution to the site by completely ignoring those ads like everybody else does is), and it's not worth spending and extra hundreds of pounds per months to run at a capacity which you'll need for 30 seconds twice a month.

Internet traffic is like beach tides. Nobody drowns because of the amount of water incoming, they drown because it happened too fast to react to.
 
Damocles said:
Our host is Amazon who run one of the largest websites in the world.

The simple fact is that no website in the world is able to cope with going from 50% of capacity to 150% of capacity in 20 seconds. Even Reddit, a global website which does five million pageviews a day, 11 full time server engineers and millions of dollars of funding cannot cope with this. Neither can anybody else.

Capacity is not the problem, if we were millionaires we could run 200,000,000 people on the site at any one time. The problem is the speed at which the users online number grows. The traditional way to handle this is to know that that traffic is coming and then scale the architecture at the backend up - so essentially you temporarily increase the capacity. This is what we do on match days or when there's an inkling of an incoming transfer and it's what every other website does.

We have automated systems which detect traffic flow on the servers and database that we use then automatically increase the capacity by adding new resources into our server farm. Unfortunately creating a new server, loading it with the programs and files that we need and putting it into the mix takes somewhere in the region of 90 seconds from dead to fully serving content. We have something that sits in front of all of our servers, tests the amount of traffic sent to each of them then works out the best way to send a request so that you get the quickest response from the site possible. When a traffic spike hits unexpectedly when one server reports that it is starting to overload, a new server gets created and put into the mix. The problem is that as the overloading server reports to the load balancer, the load balancer then shifts less traffic to it or more traffic to the other servers. Now instead of 4 full capacity servers running the site we have 3 full capacity servers running the site, a couple of servers working their way into the mix and one that is overloading. In a scenario when the traffic spikes, all that's happened is that you've shifted the burden to less servers which overloads them too until another reports it at capacity and the load balancer then switches all traffic between two, then between one then all servers are out of commission.

I'll compare site A and site B. Site A runs at a capacity to support 2000 users normally. Site B runs at a capacity to support 20000 users normally. A traffic spike for Site A is when an extra 500 users log on above what it normally online - this is a full 25% of what they are supposed to usually run. The same traffic spike for site B is when 5000 users come online which very rarely happens to anybody.

The obvious way to fix this is to increase capacity over and above what we do need and sit having lots of it doing nothing for most of the time. The problem with that is that capacity costs money, everybody blocks advertisements which pay for that capacity (and then tell us how horrible the message asking them to actually make a contribution to the site by completely ignoring those ads like everybody else does is), and it's not worth spending and extra hundreds of pounds per months to run at a capacity which you'll need for 30 seconds twice a month.

Internet traffic is like beach tides. Nobody drowns because of the amount of water incoming, they drown because it happened too fast to react to.
Click to expand...
So to cut a long story mate,why have we been having problems this morning ?
 
oakiecokie said:
Ah its just that a couple of times this morning it was running slow and then I got the message "site is not loading,check with site maintenance" which I assumed was Blue Moon.I had no problems getting onto other sites.
Click to expand...
Getting the same message
 
Damocles said:
We haven't. You might have been having problems, but the site is running fine.
Click to expand...
What do I need to do then Damocles as I`ve just tried to get onto another page on Blue Moon and this happened again :
SITE PROBLEMS.
I logged out tried to get back on 3 times with the same message.
Much obliged if you can help but please "an idiots guide" is preferable.
 
You must log in or register to reply here.
Unread Posts

Don't have an account? Register now and see fewer ads!

SIGN UP
Back
Top
✖
  AdBlock Detected
Bluemoon relies on advertising to pay our hosting fees. Please support the site by disabling your ad blocking software to help keep the forum sustainable. Thanks.