20/80 or maybe 80/20 rule

Have a question or want to start a discussion? Post it! No Registration Necessary.  Now with pictures!

Threaded View
What is your understanding of the 20/80 rule? I have seen it as 20% of
traffic is one the local subnet, 80% is remote, but have also seen it the
other way, 80% local, 20% remote.

My frickin' BCMSN study guide shows it both ways at different spots in the
book. Also, it's not clear to me if this is intended as a design guideline
or a "fact of life" to be dealt with.


Re: 20/80 or maybe 80/20 rule

Quoted text here. Click to load it

My understanding of the 80/20 (20/80) rule is that it is one of those
networking myths that some marketer used in a presentation somewhere and it
then became embedded in networking folklore - of course I don't have
references for this so this maybe just a myth I've created.

I would say that pre-Internet (and remember most "traditional" network
design had it's origins solely within the Enterprise) the original idea
would have been to design your network/internetwork so that as much traffic
as possible was kept local. Enterprise WAN bandwidth was scarce and
expensive. And operationally this made sense too - servers and printers,
etc, were all on the local network with the WAN used for interbranch
communications (internal email say) and perhaps nightly backups to a central
archive or some such.

The Internet and the Web changed all that.- now it was possible for
customers and the public generally to access the services you made available
to them. Also the Internet made bandwidth available for internal
communciations via VPN, etc.

So the balance has shifted, but to say it's 80/20 or 60/40 or 50/50 and in
which direction, is a guess - each enterprise and system has its own
characteristics which means initial network design analysis is more critical
then ever - generalisations don't help and in my view should be avoided.

And besides I think you'd be very unlucky to have this a question in any
cert exam.


Re: 20/80 or maybe 80/20 rule
Quoted text here. Click to load it

it actually goes all the way back to bridging, and it cam out of local
campus design.

WANs to LANs even now have a much bigger difference in performance - very
few people (or company offices) have an Internet feed that has 20% of the
performance of their local 10/100 or 1000 Mbps LAN.

the idea was that most traffic stays in 1 collision domain and only 20%
crosses a bridge to go somewhere else.

this was when bridges (= 2 port switches) were software driven beasties and
couldnt cope with a full 10 Mbps thruput.

and even if they could handle "wire speed", with the same speed LANs
everywhere and 2 port bridges, if your design has 100% remote traffic over a
bridge hop or 2 the aggregate thruput of your campus drops to 10 Mbps or
even less.

it went out of the window when multiport bridges and then switches became

Plenty of big sites ended up with large central star switches or even
routers with just about every LAN hooked up.
Quoted text here. Click to load it

stephen_hope@xyzworld.com - replace xyz with ntl

Site Timeline