Hi,
this may be a OOO..LD topic which is talked, discussed
or agrued for year. ISP networks may need to be
optimized continuously. But, it seems people have
different view of optimization when they use this word
at different place; sometimes optimization means
adding more access router, add more link bandwidth or
add more servers; while it could be used to point the
requirement on removing subareas in OSPF or simplify
network structure.
Is there a common sense on the target of network
optimization? or is there common startup line of such
work? What should be the model of a optimized ISP
network ( or PoP site) ?
Is there books on this topic?
Joe
You want to optimize for the lowest monetary cost network that still allows you
to meet all the SLA's you've negotiated. And this depends on what you
negotiated - for instance, if the SLA specifies 3 9's of reliability, spending
money to build a 4 9's network is cutting into your profits. Of course, if the
SLA's are biased towards latency or bandwidth, you'll have to consider those.
And remember that there usually isn't one right answer for anything but the
most simple problems - almost always, some constraint will be placed on the
solution. Often it's of the form "The salesdroid just promised XYZ", also known
as the "Don't let your mouth write no check your router can't cash" syndrome.
If it isn't that, it's a financial issue inside the company - there's always the
network you *want* to build, which is almost never the network that your
revenue stream will allow you to build....
Thanks for the response.
You want to optimize for the lowest monetary cost
network that still allows you
to meet all the SLA's you've negotiated. And this
depends on what you
negotiated - for instance, if the SLA specifies 3
9's of reliability, spending
money to build a 4 9's network is cutting into your
profits. Of course, if the
SLA's are biased towards latency or bandwidth,
you'll have to consider those.
There is always someone claims his network could reach
availability 99.9% or so, but I don't understand how a
network availability should be measured or figured
out. Is there any paper on this?
Focusing on SLA of a network, ISP network or PoP site
should not carry only one type of business traffic (
e.g. broadband access, MPLS-VPN, L2 VPN etc.), if we
consider it simply by taking network as a single
system optimization will surely be of no usage.
Looking at PoP site , is there any recommendation on
its design? a layer-2 access model is better than
router based system?
Joe