DC to choose

[Mediumcube]

What criteria have you taken into consideration when choose a DC?

[pingpipe]

A few of the things you should check is confirm that they have multiple upstream providers, onsite power generators (in case electricity goes out), a plan in place in case of emergencies like fire, and their onsite security. Alsi get some test IP's you can ping to check their network latency before you sign up.

[hosting advisor]

Originally Posted by Mediumcube

What criteria have you taken into consideration when choose a DC?

There are many things that you need to consider before going to choose the data center service. But in my opinion the most crucial point is SLA. You should have a good SLA with your data center service provider to get a blown PSU exchanged within few hours at 4 am on a Sunday morning as well.

[fcolor]

Having hosted in Tier 3 or Tier 4 class data center is important. IT is also important your provider to have a redundant network and good onsite support. I also agree that the SAL is very important!

[theseopanda]

It's actually pretty difficult to find a great DC. Rackspace is a good example of a DC that "says" the right things, but puts virtually nothing into practice, I've called them on it on more than one occasion. While NTT DC's are almost all over kill to an extreme.

Good bandwidth is the biggest thing though. Not over sold, with some redundancy.

Here are things you've probably never considered:
- Backup generators fail. There needs to be at least n+1 generators.
- Generators go through fuel like crazy. 9 hours is about the maximum for onsite fuel storage. They need refilling every 6-8 hours to keep them going, therefore having 1 fuel contract isn't enough. These DC's need to have multiple contracts, with effectively a "whoever gets here first, gets the business".
- Generators need to be tested, but not too often. Testing is hard on them. They need to have a proper load bank to test into. Many DC's just flip them on saying they're testing them 6x more than necessary! That isn't positive, that's a negative. Not having a proper way to dispose of all that power is important.
- UPS should last about 9 minutes, enough time to start a generator, anyone who starts claiming masses of UPS time hasn't built out a DC before, they're not there to supply power, they're there to provide a cut over for generators. You need n+1 UPS's as well.
- Testing automatic transfer switches is important
- Ensuring that power comes from multiple plants is important.
- Every circuit in a cabinet should be coming from a different UPS, each UPS should be coming from a different power supply, connected to a different generator. This way if any one thing fails downstream, you're not going to lose out.
- A DC that offers you basically unlimited power, can't cool their DC properly. When you go to a DC and they seem to really "limit" your power, it's because they've done the math and realize that when they start reaching like 50% capacity, they won't have enough cooling. This isn't a matter of putting 1 more AC in there, because AC's are everywhere already. There are plenty of them, it means putting in 10+ more AC's. Finding space for these massive things isn't easy.
- Lots of DC's don't setup dry pipe fire alarms. This means in the event of a pipe leak, servers are going to get wet. Fire pipes should be empty, and should be triggered when pressure drops in the pipes dramatically, such as when a sprinkler head is fully activated, but not when a small leak appears.
- DCs need to maintain proper air pressure system within, otherwise dust will be sucked in from outside through cracks here and there.
- DCs need to be vacuumed all the time. If they aren't, your server will pick up dust. If you go into a really good DC, you should pull out a server 4 years later and have almost no dust in there!
- SLA's are next to useless. They've got an out clause for basically everything, you'll never see a cent


The list goes on and on. For someone without a lot of experience in DC's, they all look pretty decent at the tier 1 level, but there are huge differences. At the tier 3-4 level, you're getting real hit of miss!

[b2netsolutions]

Bandwidth is really important.

[ByteHouse]

Originally Posted by theseopanda

It's actually pretty difficult to find a great DC. Rackspace is a good example of a DC that "says" the right things, but puts virtually nothing into practice, I've called them on it on more than one occasion. While NTT DC's are almost all over kill to an extreme.

Good bandwidth is the biggest thing though. Not over sold, with some redundancy.

Here are things you've probably never considered:
- Backup generators fail. There needs to be at least n+1 generators.
- Generators go through fuel like crazy. 9 hours is about the maximum for onsite fuel storage. They need refilling every 6-8 hours to keep them going, therefore having 1 fuel contract isn't enough. These DC's need to have multiple contracts, with effectively a "whoever gets here first, gets the business".
- Generators need to be tested, but not too often. Testing is hard on them. They need to have a proper load bank to test into. Many DC's just flip them on saying they're testing them 6x more than necessary! That isn't positive, that's a negative. Not having a proper way to dispose of all that power is important.
- UPS should last about 9 minutes, enough time to start a generator, anyone who starts claiming masses of UPS time hasn't built out a DC before, they're not there to supply power, they're there to provide a cut over for generators. You need n+1 UPS's as well.
- Testing automatic transfer switches is important
- Ensuring that power comes from multiple plants is important.
- Every circuit in a cabinet should be coming from a different UPS, each UPS should be coming from a different power supply, connected to a different generator. This way if any one thing fails downstream, you're not going to lose out.
- A DC that offers you basically unlimited power, can't cool their DC properly. When you go to a DC and they seem to really "limit" your power, it's because they've done the math and realize that when they start reaching like 50% capacity, they won't have enough cooling. This isn't a matter of putting 1 more AC in there, because AC's are everywhere already. There are plenty of them, it means putting in 10+ more AC's. Finding space for these massive things isn't easy.
- Lots of DC's don't setup dry pipe fire alarms. This means in the event of a pipe leak, servers are going to get wet. Fire pipes should be empty, and should be triggered when pressure drops in the pipes dramatically, such as when a sprinkler head is fully activated, but not when a small leak appears.
- DCs need to maintain proper air pressure system within, otherwise dust will be sucked in from outside through cracks here and there.
- DCs need to be vacuumed all the time. If they aren't, your server will pick up dust. If you go into a really good DC, you should pull out a server 4 years later and have almost no dust in there!
- SLA's are next to useless. They've got an out clause for basically everything, you'll never see a cent


The list goes on and on. For someone without a lot of experience in DC's, they all look pretty decent at the tier 1 level, but there are huge differences. At the tier 3-4 level, you're getting real hit of miss!

Your right. It depends on the implemenation and what their SOPs are. That's what makes or breaks. There are Tier 4 data centres we've used, and stopped using because although they've got the infrastructure in place their actions are not better than a tier 2 whatever the redundancy et al., they claim and there are Tier 3 data centres that we've used and continue to use that are far better than any Tier 4 data centre simply becauce they've got all the right operations in place.