I like that analogy and agree for the most part ….
Often when providers are shopping for bandwidth, the first question is price. I am not saying it’s not an important factor but suggesting that folks take other considerations into account such as what you have outlined. For me, latency and performance combined with reliability are key. Also, support for BGP communities is essential. Then comes price … Pricing in IP Transit has become such a commodity anymore – used to be very market dependent but now if you have access to carrier hotels then pretty much anywhere in Canada/USA you will see similar pricing (sometimes a slight amount higher in Canada). If you don’t have access to carrier hotels then that’s a different story :) From: Af [mailto:af-boun...@afmug.com] On Behalf Of Justin Wilson - MTIN Sent: Thursday, June 25, 2015 12:41 PM To: af@afmug.com Subject: Re: [AFMUG] Data center bandwidth One of the things I always use as an analogy for bandwidth are cars. Will a $500 junker car work for you? If so then it’s going to be what you need. Is the ride in that car always smooth? Could be or might not be. Depends on where you are going. Some roads are better than others. Now, do you want something that might get you there faster, or have a smoother ride? or maybe you want a nice luxury car to haul your important clients around in? Do you want more than one car? Sure you do. It’s called redundancy. Do you need two luxury cars? maybe. Do you need two beaters? maybe. Who says what luxury is? Thats marketing speak. Now, can you get from San Jose California to 350 Cermak in Chicago 10 ms faster over Level3 than Cogent? Maybe. Is that important to you? Notice I did not mention any name brands in this? I have clients we have both Hurricane and Cogent and it works great for them. Others not so well. Does this mean Cogent is better than Level3? We could show metrics, all kinds of things to “prove” who is better than others. Much of it depends on what criteria you want to use. The biggest things from a technical standpoint to look for in bandwidth providers are: 1.How many other networks are they peered with? 2.Where are they peered with these networks at? If you have to take long haul 3 states away to hit ProviderA’s POP but only a trip across the state to hit a major carrier hotel with ProviderB r, then your experience might be better with B. 3.HGow oversubscribed is their network in your area. It’s just like your networks. Certain POPs are busier than others. If they have bottlenecks on your local POP that will taint your experience. Seen this happen on every major carrier. It can all quickly become a Ford Vs Chevy or Apple Vs Windows debate very quickly. We all know someone who has had a bad experience with Ford/Chevy/Dodge/whomever and will never buy another. But there is a reason they are still in business. Those work for someone. The key is to make yourself as nimble as possible. The easiest way is, like this thread started out, to get transport back to a “carrier hotel”. Once there you can shop around. Your next step is to have a redundant path to that data center or a redundant path to a different data center. Once you do that you can graduate to connections between the data centers. So you end up with rings of connectivity to those data centers with the ability to change IP providers much easier. You can also keep track of metrics a little better. My router, at the data center, gets 80ms pings within your network. You can’t blame the transport then. Lots of advantages. Justin --- Justin Wilson <j...@mtin.net <mailto:j...@mtin.net> > http://www.mtin.net Managed Services – xISP Solutions – Data Centers http://www.thebrotherswisp.com Podcast about xISP topics http://www.midwest-ix.com Peering – Transit – Internet Exchange
