Posted by Billy Rubin on May 19, 2000 at 21:21:52:
In Reply to: Re: What exactly causes lag? posted by dan on May 19, 2000 at 20:45:51:
> As you probably know, your piece of data jumps across ten or twenty different machines before it'll reach cf.
Spot on baby...machines being ATM routers and switches passing encapsulated IP.
>
> If one of these machines is doing some really intensive stuff (whether it be network intensive [downloading] or processor intensive [compiling a new kernel for unix]) your packet will take longer to pass through the machine.
Bzzzzt... Backbone fiber nodes (see "machines" above) don't compile new kernels or download data..they shuffle data around (cell switching in the ATM world)and send it on its merry way. Nowadays, you rarely get latency in core routers like you used to...IS backbone providers are keeping pace fairly well. Your problems will most likely be "last-mile", or the link between the local phone companies central office and your location.
>
> In a tracert I did during this severe lag, that one piece of data didn't even make it, so your computer will recognise that and send it again, hopefully getting a different routing path. Sadly, each of these retries takes even longer.
>
> The problem with this particular bout of lag seems to lie with Bell Southwestern as my tracert makes it all the way over there, then gives up.
See above, next hop after SW Bell is probably the premises. If the box is thrashing around in death spasms, it may not have time to acknowledge the dead packet it receives. (Look up "TTL" as it relates to traceroute)
>
> I hope everything I've said here was correct, I think it is but I'd welcome any "proper" TCP professionals to correct me
>
This would be IP actually, with some ICMP for good measure (traceroute...or is that UDP?)I'm a long ways from being an expert, but I do get paid, so I guess that makes me a professional. All critiques and mistake finger-pointing welcome indeed.