Should Server Internet Connection be based solely on Internet Speed?

We are running our own server and our country's ISP is telling us that we should run it with their T1 Copper Line with 2mb DL & UL. Right now it is running on a Fiber Line with 8mb DL & UL.

They say that the T1 would be better since it is connected straight to the internet backbone while our current Fiber line does not.

Right now our users are running on 1mb to 2mb DSL Speeds.

Our systems mainly serve up 10 basic static websites, 1 web service with around 50,000 users (not monthly traffic, its actively using the system 8 hours a day), 1 web service that has a ping for every 5 seconds with around 30 users, an FTP and RDP server.

I am not very knowledgeable about the technical side of internet connection, all I know is that everything that runs through the internet is bytes of data(am I right on that?), so shouldn't it be better that we use a higher connection speed (Fiber)?

Can anyone help us on identifying which internet connection we should use, the T1 Copper Line or The Fiber Line? And Why?

Thanks!

Solution 1:

I can't imagine why an ISP might suggest that, but they clearly have no idea what they are talking about.

A T1 has, and has always had, 1.544Mbps symmetric bandwidth. The usable bandwidth is slightly less than that depending on protocol.

Having a few less hops might reduce latency somewhat, but if your users aren't having latency problems, cutting your bandwidth to a fifth of what it currently is seems like a really bad idea.

T1 is also usually pretty expensive, and often as or more expensive than fibre.

Very strange.

Solution 2:

To answer your question in general, yes there are other factors in play when choosing your internet connection for your hosted services. Internet speed is not the only thing you need to consider.

You need to consider:

1. Bandwidth (i.e. speed)
2. Latency (i.e. the distance you are from your customer base)
3. Peering (i.e. who does your network connect to. This is closely related to latency. Bad peering = potentially bad latency)
4. Redundancy (i.e. what happens if an upstream provider goes offline. This is closely related to peering).

Now, all that said and done, the situation you have described sounds like you're meeting all your requirements right now, so if it aint broke, don't fix it.

For further anecodotal evidence, my former employer has their primary datacenter right in the middle of a large datacenter. They have great bandwidth, they have minimal latency (sub 3ms to some customers), loads of peering options and great redundancy. Everything is great.

But their DR site has a single fiber link, like yours. It has pretty average latency, is not a 1:1 contention ratio, and less than half the bandwidth. It has only one peer for quite a few hops, and no redundancy. But when we had to fail over to DR due to a hardware failure, the complaints that we got from our customers had nothing to do with the internet link.

So all in all it sounds like their marketing and sales guys were just tying to get their hooks into you. I'm genuinely surprised that companies are still selling T1/E1 services for web hosting.