Re: Bandwidth Customer ratios

Mitchell B. Wagers ( (no email) )
Thu, 11 Dec 1997 15:18:01 -0800

I'm with you, I don't like "shorting" bandwidth, which is what happens when
you say you can fit 96 33.6's on a T, because that is way over the total
bandwidth available...

At 06:08 PM 12/11/97 -0500, you wrote:
>I believe it all varies based on your usership, and you shouldn't put a
>hard, fast number on it. If you can feed 150 "ports" on a T-1 and not
>saturate the T at peak usage, then do so! If you can only support 60
>users because most are download hogs or using video on demand, time for a
>new T, in spite of that 96 recommendation (which seems to me to be based on
>a HORRIBLE ratio... A T-1 is ABOUT 28 56k modems, or about 46 33.6k modems.
> Now, I don't know about YOU, but let's assume everyone's using 56k
>(unrealistic -- 33.6k is FAR more common, even now). Let's assume some
>overhead, and call that 24 concurrent downloads at 56k. Are ONE in FOUR
>of your customers doing a SUSTAINED download at any given time? Not
>here, they're not.
>At 05:36 PM 12/11/97 -0500, you wrote:
>>-> A silly question for someone,
>>-> Has anyone found out a decent formula for figuring out how much
>bandwidth is
>>-> needed to provide a decent service to "X" amount of dialup customers?
>>-> than the obvious phone call stating "Man downloading is sooo slow?" I
>>-> haven't figured out a really good way of telling. I have to factor in a
>>-> nominal number of webpages that we host also. Any kind of generic
tips or
>>-> thoughts would be helpful.
>>Boardwatch magazine printed an article a while back and I believe they
>>suggested 96 dialin lines per T-1 of bandwidth into the Internet. This of
>>course whas when we were using 33.6kbs modems, not ISDN and 56k series
>>Jeff Binkley
>>ASA Network Computing
>> ----------------------------------------------------------
>> NTISP Mailing List
> ----------------------------------------------------------
> NTISP Mailing List