Hi, is there a standard or a practice on how much IP addresses an ISP should
provide to his/her client given that this client has bought only 2Mb of
bandwidth and this client is an ISP?
Thanks
sheng
What type of IP usage does this client expect? And how large is the block
of addresses available to yourself?
If you're in the US...
http://www.arin.net/regserv/initial-isp.html
In a nutshell, the customer needs to be able to populate at least 50% of
the given allocation. If the customer already has allocated space, 80%
of that must be populated before given an additional allocation.
-C
Umm.. don't bother. Let's think this through. 2Mbits/sec of bandwidth
will only sustain about 40 56KB modems doing a simultaneous download.
Even adding in think time and the like, a /24 should be plenty wide enough.
The *BIG* question is how the ISP intends to make any money at that scale.
Figuring even a 10X overcommitment, that's 400 customers at $20/mo or so,
for an inbound cash flow of only $8K/month, with which they get to pay their
bandwidth charge, their tech support, and everything else.
I wish them luck.
> Hi, is there a standard or a practice on how much IP addresses an ISP should
> provide to his/her client given that this client has bought only 2Mb of
> bandwidth and this client is an ISP?Umm.. don't bother. Let's think this through. 2Mbits/sec of bandwidth
will only sustain about 40 56KB modems doing a simultaneous download.
Even adding in think time and the like, a /24 should be plenty wide enough.
nice fantasy land you live in.
in the real world, i'm seeing as much as 32 E1 PRI's (960 lines) connected
by a single 2mbit internet circuit.
that may be crowded to some people on the list, but it is not unrealistic
in some regions.
The *BIG* question is how the ISP intends to make any money at that scale.
Figuring even a 10X overcommitment, that's 400 customers at $20/mo or so,
for an inbound cash flow of only $8K/month, with which they get to pay their
bandwidth charge, their tech support, and everything else.
the BIG question is does the ISP have the bankroll to get going, but that
is way off topic here.
the economics of building an ISP are different from region to region.
Hi Sheng,
Yes there is a formula,
"As many as they need to address their hosts/interfaces." If you look at the
arin website (www.arin.net) I believe you may find some guidelines.
Bandwidth shouldn't be the driving factor but specific addressing needs
should be.
"ISP" can mean so many things these days:)
If they have a 100 or so dial up ports and a dozen or so additional servers
and hosts then a /24 should be more than enough.
With 2Mb of transit they are not going to be any bigger than that?
They may even get away with less addresses depending on what they're doing.
JC
Of course bandwidth != subnet mask. He should give them whatever IP's
they demonstrate a need for in the next three months. Determining and
justifying that
need has nothing to do with how over (or under) subscribed their
bandwidth is.
If they are in fact only selling dialup (not leased lines, not web
hosting),
you might ask how many pops(locations) they plan to have right away,
modems/pop, space
reserved for internal devices (email/corporate lan) and links. You
could
easially justify a couple of /24's with a couple locations and IP's for
all the new PC's in the marketing dept.
KL
Not really...I regularly see 180+ lines filled and BW utilization averages
about 800-900kbps. This is with a couple hundred websites, outsourced
news a dedicated DS0 customer that stays pegged almost all the time and
several colocated servers.
I'll bet this has been hashed here a few times...
James Smallacombe PlantageNet, Inc. CEO and Janitor
up@3.am http://3.am
* Valdis.Kletnieks@vt.edu <Valdis.Kletnieks@vt.edu> [20010612 13:03]:
> Hi, is there a standard or a practice on how much IP addresses an ISP should
> provide to his/her client given that this client has bought only 2Mb of
> bandwidth and this client is an ISP?Umm.. don't bother. Let's think this through. 2Mbits/sec of bandwidth
will only sustain about 40 56KB modems doing a simultaneous download.
You've got an interesting view of the $20/mo. retail dial-up market
economics. Very few access providers have end-users on dial-up with such
thriving Internet habits (though each access providers' customer base
differs).
Even adding in think time and the like, a /24 should be plenty wide enough.
I can't contest this since the original poster provided insufficient
information. A single /24 happens to be a default for some providers of T1
IP transit in the U.S unless the end-user requests otherwise...but not all
providers are quite so lacks. More importantly, the IP address utilization
(and associated utilization time line) is more useful in determining what
size block should be assigned to the client. The available (or in use)
bandwidth is more of a side note than a deciding factor in block size
assignment. It happens to be rare that a dial-up customer can justify
a /24 but it is hardly rare for a T1 customer to have multiple /24s fully
populated. IP address requirements are hardly a direct result of the size
of the pipe.
The *BIG* question is how the ISP intends to make any money at that scale.
Figuring even a 10X over-commitment, that's 400 customers at $20/mo or so,
for an inbound cash flow of only $8K/month, with which they get to pay their
bandwidth charge, their tech support, and everything else.I wish them luck.
You've apparently not paid much attention to how this industry got started
have you? Though, they will need the luck...I'll grant you that... 
Times have changed more than a little in the U.S....but it still happens
here. On an international level there are even more pockets where this
sort of evolution in the Internet industry has yet to be played out (and
can still be exploited for fun and profit).
-jr
* Valdis.Kletnieks@vt.edu <Valdis.Kletnieks@vt.edu> [20010612 13:03]:
> Umm.. don't bother. Let's think this through. 2Mbits/sec of bandwidth
> will only sustain about 40 56KB modems doing a simultaneous download.
You've got an interesting view of the $20/mo. retail dial-up market
economics. Very few access providers have end-users on dial-up with such
thriving Internet habits (though each access providers' customer base
differs).
OK.. I'll admit it - personal viewpoint *may* be a bit slanted, we've been
just a bit ahead of the curve - bev.net launched about a decade ago across
the hall from my office.
http://www.bev.net/project/brochures/about.html
The local people are on-line, big-time, real-time, all the time. At least
locally, we need to over-provision compared to what a lot of other people
are reporting.
> Even adding in think time and the like, a /24 should be plenty wide enough.
I can't contest this since the original poster provided insufficient
information. A single /24 happens to be a default for some providers of T1
Well.. the question was how much space to sell to *another ISP*. Now, we've
seen several numbers that all seem to agree that several racks of modems will
saturate the 2Mb link to the customer ISP for a /24 or maybe a /23 worth
of dialup modem pool space.
If the new startup ISP is providing colocation, expect that either they
have colocated a lot of idle hardware, or that they'll saturate their 2Mb
even faster. I'd expect a /24 worth of webservers should saturate an
uplink even faster than a /24 worth of terminal servers.
You've apparently not paid much attention to how this industry got started
have you? Though, they will need the luck...I'll grant you that...
I *was* paying attention - I was there. 
Just some days I forget there's people still trying to climb onto the
bandwagon we're desperately trying to get *off*.