hi, Im hust learning about the OC connection types and really appreciate this kind of post. I was chosing a hosting provider and have a choice between oc 48 and oc 12. only question I have
Q: is it possible for someone advertising OC-48 connection for their hosting not really using OC-48 fully? or they have rented out some few oc12 s out of the OC-48 network? in another wirds could they be just twisting words...
Q: Would it be notiable diffeence between OC 48 and OC 12 when running ecommerce site? or when would it be noticable.
I thank you in advance !!!
Bosch,
In respect to your first question:
It is always possible for this to happen. If they are selling OC12s off of their OC48 network, they are not really "twisting words". When they advertize the OC48 they are telling you what kind of bandwidth they have coming in to their router. In theory you are not going to get a full oc48 to your server ever. The host will always run their OC48 to a router which will then run either 100mbps or gige to the switch that your servers will connect to. If they run an OC48 directly to the switch your server is connected to without a router, I would recommend walk, ney run, as their network probably is not setup as it should be.
The answer to your second question is dependant to the host's utilization of the circuit. If they have an OC-12 that is barely utilized you will likely not see too much of a degregation; this is primarily due to the fact that your average user will not be browsing the ecommerce site at 600mbps. If it is a situation where the OC12 is somewhat utilized, I would recommend going with the OC48 host to ensure that you are not part of a choked connection.
How the datacenter engineers its network makes more of a difference than what type of pipes they use to connect with carriers. With sufficiently large pipes (>OC3), it is unlikely that simple changes in usage patterns will congest any links, unless the operators are really asleep at the switch. Latency is not sufficiently different between gigE, OC12, OC48 for anyone to notice, unless you're trying to run iSCSI over it or something.
One differentiator is resilience to DoS attacks. All carrier links to a datecenter network need to support real-time black-hole features, to keep large DoS attacks from affecting too many customers. Datacenters which have a large amount of unused bandwidth on every carrier connection have a better chance of discarding traffic from small to moderate DoS attacks, without impacting most customers. Datacenters which can send the small/moderate attack traffic through a packet scrubber are even better; but that's not a guarantee.
If you're not in a major peering region, then if a datacenter connects locally to many carriers, and your customers are local, then that adds value. Even towns like Miami still have some Tier-1's hauling traffic to DC to exchange with other carriers.
if you colocate in their data center, ask them for "ethernet" connectivity. In fact, that's the norm in the hosting world that i'm dealing with.
as far as OCx, that's telco grade transport (connectivity to a data center). if you're looking for serious colocation, try to find a data center that is located in a "carrier neutral" facility. that means they will give you access to multiple IP uplinks (to the internet). hope this helps somewhat.
I'd appreciate if someone would add comments on current OCx and ethernet applications to this thread. I could but prefer others input for balance and encouraging participation.
For those of you in the US ... we can help you determine the best bandwidth solution for your application(s) at no cost. Meaning .... free of charge for this service.
