Bandwidth Upgrade

                My team is in the process of putting some documentation together to justify a bandwidth upgrade. I am asking if you would be willing to reply back to me, with how you decide that it is time to upgrade your bandwidth. On-line or off-line reply's will be acceptable.

Thank You

Daniel Bielawa
Network Engineer
Liberty University Network Services


40 Years of Training Champions for Christ: 1971-2011


"40 Years of Training Champions for Christ: "

You would have thought they would have trained a /Network Engineer/,
or two.. in those 40 years, wouldn't you ?



Start with why you think it's necessary and what happens if mgt doesn't
listen. Bandwidth is like electricity in a sense. Either you have what
you need or you go belly up until some utility company can give you more
juice. If you notice a growth pattern and are trying to get in front of it
that's obviously important. If you are approaching the point where a
single link in redundant pairs can't handle the full load during an outage,
that's important as well. Knowing how to use power point helps too (avoid
animation). Maybe I don't understand the question but this is usually a
no-brainer especially if there is an existing capacity management strategy.

Ideally, when our 95th-percentile hits 65% utilization, we begin the
pricing and planning process and its up on peoples radar. Once the
95th-percentile hits 80-85% we start planning the maintenance and execute
the upgrades. I say ideally, because in a perfect world this would happen
100% of the time.

We try to upgrade when the 95th is at 80-85%, because the 95th-percentiles
is based off 5-min polls, so I am sure traffic is spiking higher at peak



That depends on the network configuration though. If you have redundant
links and one link is at 65% and the other is at 35% or more you won't be
able to get through a circuit flap or outage without dropping packets.

Very true.. It is an open-ended question that can have many answers,
especially without knowing their design...

Yes, lot's of missing pieces here.

It depends on your tolerance for delayed and dropped packets during periods
of high usage, connection media type, speeds we're talking about, who your
users are, and the applications you must support.

Generally if your graphs says 75% peak usage, you should have upgraded. If
you're output drop counters are unacceptable you also need an upgrade.

However cases with a a subrate access network (IE: users capped at 5 meg,
on a 1000 megabit upstream pipe) can get away with running closer to
capacity a lot more than those with a few enterprise "bursty" customers who
can single handidly burst 50% of your upstream), and demand no dropped or
delayed traffic in their SLAs.

Also consider failover issues if you're redundantly connected.

Capacity planning is one of those things that (can) fall into the technical, business, and political aspects of network engineering.

Addressing the technical aspects is pretty straightforward. The first thing is having data to show how much of your bandwidth you're using now. Even something as basic as a set of MRTG (not knocking MRTG at all) graphs.
If you have that, or data from some other package, then that's a good start. If they show things like saturation of your existing pipe, then that's an important data point to share with your management, because flat-topping has a tendency to turn into sluggish performance and unhappy customers pretty quickly. I would agree with other posters that we start looking at capacity upgrades when we get to about 65% usage, and have the upgrades completed before we get to 80-85%. Depending on what size pipes you have now, you can also possibly significantly reduce your cost per Mb/s, though this depends a lot on your location and what sort of comms facilities are in your area. Don't be afraid to call some carriers and get quotes.

Having historical data that shows the growth of your bandwidth needs is even more useful, to a point. The reason I said to a point, is because a straight graph of inbound and outbound traffic doesn't answer questions about what is going on that is driving traffic growth. At that point, you start getting into the realm of traffic analysis.

Addressing the business aspects starts to get into areas that we (people on NANOG) can only offer generic advice, because you would know how to present the business case for an upgrade to your management better than we would. For example, if keeping customer complaints about network performance as low as possible is a business priority, then showing reports of related trouble tickets your helpdesk has received from your user base might be another important data point as well.

Also keep in mind that what might start out as "we need more bandwidth" could turn into other costs as well. A larger pipe might mean you need a new interface card for your router, or other upgrades to the internal network to make use of that larger external pipe. Most managers I've worked with would rather get the bad news (requests for money) all at once, so they only have to 'go to the well' one time, rather than making requests for money to upgrade the pipe, then another request for the new interface card, etc. That also starts to get you into the possibly political aspects of working through your business environment.

Hope this helps.


For the past 17 years I have managed to keep my bandwidth budget the same. I had 1 T1 in 1994 and have multiple GigEs now. I still pay roughly the same price. So, shop it around and see if you can upgrade without affecting your budget. It is much easier to justify when it doesn't change the bottom line.