I don't know this, but maybe someone can help me understand this better.

On compairing 2 pipes ( oc3 and oc12 ) utilization is at 0%
server setup is the same.

Which one will return the web site to the visitor faster.

I can't figure it out.

both travel at the speed of light. they are only subject to the routing. But if the routing is the same then ...

-Michael

Originally posted by sqposter
I don't know this, but maybe someone can help me understand this better. On compairing 2 pipes ( oc3 and oc12 ) utilization is at 0% server setup is the same. Which one will return the web site to the visitor faster. I can't figure it out. both travel at the speed of light. they are only subject to the routing. But if the routing is the same then ...


This question doesn't return an obvious answer, as there are numerous details that can determine how fast a website is returned to a visitor.

In dealing with the pipe, the OC-12 can theoretically transfer more data at one time than the OC-3, under excellent conditions. Factors such as distance, provider, net congestion, and the enviroment make "excellent conditions" nearly impossible. Further, just because you have an OC-12 or OC-3, doesn't mean that the entire OC-3 or OC-12 is "lit." A lot of providers out there claim that they have dual OC-3 or OC-12, or even OC-192. But, if you run a network statistic on the network, you will notice that their OC-3, 12, or 192 may not be running at the theoretical maximum (meaning no where near it). The reason is that even though a host may actually have OC-3, 12, or 192 lines in their datacenters, they may only have DS-1, DS-3 (E1 or E3 in some countries) lit. Thus, their throughput is only in the 30-45Mbit range, rather in the 155+ Mbit range.

Another factor that can determine speed is the server itself. A 300 MHz pentium can not handle the complete "lit" speed of an OC-3. It couldn't process the information fast enough. A 1GHz AMD couldn't process a complete "lit" OC-12. It can't process the information fast enough. Thus, your speed is also limited to your server.

Routing can be another factor. Even though you said that routing would be the same, the router and switch may not be able to run at the maximum "lit" speed.

Finally, network configuration could cause issues too.

So, basically, there are too many variables to determine which would be faster.

Originally posted by Reg


This question doesn't return an obvious answer, as there are numerous details that can determine how fast a website is returned to a visitor.

In dealing with the pipe, the OC-12 can theoretically transfer more data at one time than the OC-3, under excellent conditions.
Another factor that can determine speed is the server itself. A 300 MHz pentium can not handle the complete "lit" speed of an OC-3. It couldn't process the information fast enough. A 1GHz AMD couldn't process a complete "lit" OC-12. It can't process the information fast enough. Thus, your speed is also limited to your server.

So, basically, there are too many variables to determine which would be faster.

Very interesting, you've covered pleanty of questions ( and opened another huge can of worms for me)

1 question would be then server optimization, for the given:

the pipe selected should be respective of max cpu load (over a given time) of the server.

I'm working on the idea of : theory of balancing bandwidth requirements with server usage.


Now then I have to ask: where do you aquire data that states what motherboard / cpu / memory setup ( we can skip the hard drive and controller card pleanty of optimized solutions for that) gives the most bang for the buck.

to see my line of thinking :
1) depending on server you can host 1 to n ( your willing to give certain amount of cpu cycles to each customer )
2) bandwidth cost ( fix or var. )
3) server cost (fixed )


so with a quick rundown, after you have compiled the data, you should be able to define what server would get what type of service. IE : if your going to minimize the cpu usage on the server, you might want to guild all the cheap customers there ( also keeping bandwidth low and software restrictions high ) while certain servers that require cpu time ( due to software platform ) might be set on simular platforms but price respectively due to the few amount of customers using it.

then after a while you can flip the customer to the right bandwidth pipe to optimize earnings.


now a simple question : is my line of thinking correct or did I fall into the wrong path.

-michael