So You Want to Host My Servers? 8 Requirements for a Data Center

by Aaron Brazell on August 16, 2006 · 56 comments

A good system or cluster of systems is absolutely essential for reliable uptime for a website or service. Even more importantly is the quality of the data center provider that is supporting your servers. This may be as close to any kind of “from the horse’s mouth” as you’re going to get but we have had an ongoing struggle over the past two months or so with our data center provider. We ramped up quickly on quality systems, but over time service and availability has diminished and now we will be looking forward to the next solution.

So what exactly does a company look for when shopping for a data center provider? I have certain pre-qualifying factors for consideration.

Power

Every two years or so, a glitch such as a power outage in the data center is acceptable. I’ll complain, but realistically I won’t move to another provider as long as the issue is addressed as quickly as possible. Twice in two weeks is obscene. Reliable power is the absolute most basic of basic building blocks. Surges in the lines should be moderated by well-rated Universal Power Supplies and the entire center should have at least one backup generator - preferrably two.

Climate Control

Another critical aspect of a data center is a cool, dry climate controlled environment. This goes without saying and should even have to be inquired about. Any competent company doesn’t keep servers in the dank basement of grandmas house.

Eyes and Ears

Unmanaged means just that. Data center people will not be available for support on systems and systems will be completely self-supported. We look for unmanaged servers because they are cheaper. When you get managed solutions, you pay for labor you’re probably not going to use. However, eyes and ears in an unmanaged environment still means that physical presence requirements can be met. Linux servers generally do not need a reboot. Patches can be applied without restarting. In fact Unix boxes were like this long before Windows came around and it wasn’t until the advent of Windows that rebooting equipment became a fact of life. In the off case scenario that a physical reboot is necessary, eyes and ears should be able to do it - even in an unmanaged environment. Hardware repairs that do not allow for me to be onsite should be done at the data center level. Support phone calls in the case of outages should be responded to promptly regardless of SLA.

Remote Monitoring

In this case, “remote” means not on localhost. Remote monitoring means that there is some mechanism in place to alert on site technicians as well as designated self-provided systems administrators to be alerted in the case of basic service outages (http, smtp, pop, ftp, ssh, dns, mysql). I want to know the minute a server goes down for any reason and should not have to pay a load of money to find out.

99.5% uptime

While this may seem extraordinarily high, this equates to 3 hours and 36 minutes in a 30 day month. Thats actually a lot of downtime. Realistically, I want no more than 30 minutes a month. Downtime for b5media means nearly 20,000 pageviews an hour of lost ad revenue, lost traffic and lost productivity. Time is money. I don’t want to waste it on downtime.

Competent Support

While it seems like a gripe, if I call the data center during emergency downtime at 2am, I expect to be able to talk to a competent person who might, if necessary, have to do something with our boxes. High school interns who only answer phones is unacceptable.

Sales Responsiveness

We may currently be in a blog freeze, but we have definite expansion plans and those plans will undoubtedly mean ramping up new servers in a hurry. It happens when you do millions of pageviews of traffic in a month. I need to have a scalable infrastructure and be able to have a new server online and ready to go within 24 hours. We’ll pay. We don’t ask for charity. But if we need it done, we need it done.

Network Redundancy

Without fail, some network backbone is going to go down and in those times, I need to be assured that theres a plan. I want a minimum of 4 Tier 1 backbones coming into the data center. Premium backbones please.

What are your requirements? And email me at abrazell [at] b5media [dot] com for sales pitches in this area. I’ll have server requirements too.

Note: This entry is a part of the Problogger ‘Lists’ Group Writing Project.

Send to Facebook

{ 6 trackbacks }

82 More Lists - Group Writing Project Reader Submissions Part III: 08.17.06 at 8:50 am
Working at Home on the Internet: 08.18.06 at 4:49 pm
Lists Group Writing Project - Categorized: 08.26.06 at 4:39 am
Lists Group Writing Project - Categorized: 08.26.06 at 4:39 am
Michael Anuzis » Lists of Advice: What Would You Like To Know?: 08.26.06 at 11:56 am
Successful Blog - 301 Links in a Story — Chapter 8 Lizzie, Amanda, Arianna, David, and Jeremy: 09.15.06 at 9:16 pm

{ 50 comments }

1 Jeremy Wright 08.16.06 at 6:10 pm: Don’t forget phone support. Something that’s easy to overlook :)
2 Jeremy Wright 08.16.06 at 6:10 pm: Don’t forget phone support. Something that’s easy to overlook :)
3 Jeremy Wright 08.16.06 at 6:10 pm: Don’t forget phone support. Something that’s easy to overlook :)
4 Jeremy Wright 08.16.06 at 6:10 pm: Don’t forget phone support. Something that’s easy to overlook :)
5 Jeremy Wright 08.16.06 at 6:10 pm: Don’t forget phone support. Something that’s easy to overlook :)
6 Aaron 08.16.06 at 6:18 pm: Included in #3, 6 and 7 but yes, you are correct.
7 Aaron 08.16.06 at 6:18 pm: Included in #3, 6 and 7 but yes, you are correct.
8 Aaron 08.16.06 at 6:18 pm: Included in #3, 6 and 7 but yes, you are correct.
9 Aaron 08.16.06 at 6:18 pm: Included in #3, 6 and 7 but yes, you are correct.
10 Aaron 08.16.06 at 6:18 pm: Included in #3, 6 and 7 but yes, you are correct.
11 candice 08.17.06 at 12:42 am: On the more managed side, support of backups is always something to be concerned about. Even if it’s just “is there someone there to plug in this extra hard drive for me.” (Down in disasterworld, everyone has abandoned tape for usb and networked solutions. We learned that tape is useless.)

I’ve had issues with some customers’ offsite data center lately not doing stuff too. We sent them a machine and it took a week to get it put online.
12 candice 08.17.06 at 12:42 am: On the more managed side, support of backups is always something to be concerned about. Even if it’s just “is there someone there to plug in this extra hard drive for me.” (Down in disasterworld, everyone has abandoned tape for usb and networked solutions. We learned that tape is useless.)

I’ve had issues with some customers’ offsite data center lately not doing stuff too. We sent them a machine and it took a week to get it put online.
13 candice 08.17.06 at 12:42 am: On the more managed side, support of backups is always something to be concerned about. Even if it’s just “is there someone there to plug in this extra hard drive for me.” (Down in disasterworld, everyone has abandoned tape for usb and networked solutions. We learned that tape is useless.)

I’ve had issues with some customers’ offsite data center lately not doing stuff too. We sent them a machine and it took a week to get it put online.
14 candice 08.17.06 at 12:42 am: On the more managed side, support of backups is always something to be concerned about. Even if it’s just “is there someone there to plug in this extra hard drive for me.” (Down in disasterworld, everyone has abandoned tape for usb and networked solutions. We learned that tape is useless.)

I’ve had issues with some customers’ offsite data center lately not doing stuff too. We sent them a machine and it took a week to get it put online.
15 candice 08.17.06 at 12:42 am: On the more managed side, support of backups is always something to be concerned about. Even if it’s just “is there someone there to plug in this extra hard drive for me.” (Down in disasterworld, everyone has abandoned tape for usb and networked solutions. We learned that tape is useless.)

I’ve had issues with some customers’ offsite data center lately not doing stuff too. We sent them a machine and it took a week to get it put online.
16 Luck Dog 08.17.06 at 6:07 am: Everyone guarantees 99.5% uptime don’t you know?

Of course they are only up 70 -90% of uptime, and I have yet to ever get a dime back.

When you find the perfect host let us all know!
17 Luck Dog 08.17.06 at 6:07 am: Everyone guarantees 99.5% uptime don’t you know?

Of course they are only up 70 -90% of uptime, and I have yet to ever get a dime back.

When you find the perfect host let us all know!
18 Luck Dog 08.17.06 at 6:07 am: Everyone guarantees 99.5% uptime don’t you know?

Of course they are only up 70 -90% of uptime, and I have yet to ever get a dime back.

When you find the perfect host let us all know!
19 Luck Dog 08.17.06 at 6:07 am: Everyone guarantees 99.5% uptime don’t you know?

Of course they are only up 70 -90% of uptime, and I have yet to ever get a dime back.

When you find the perfect host let us all know!
20 Luck Dog 08.17.06 at 6:07 am: Everyone guarantees 99.5% uptime don’t you know?

Of course they are only up 70 -90% of uptime, and I have yet to ever get a dime back.

When you find the perfect host let us all know!
21 Aaron 08.17.06 at 7:10 am: When you find the perfect host let us all know!

Will do!

Despite your cynical tone, if you have a 70-80% uptime and never got any money back then you have poor judgement on hosts or just plain suck. :) If b5media ever was in such a situation, we’d be out of business.
22 Aaron 08.17.06 at 7:10 am: When you find the perfect host let us all know!

Will do!

Despite your cynical tone, if you have a 70-80% uptime and never got any money back then you have poor judgement on hosts or just plain suck. :) If b5media ever was in such a situation, we’d be out of business.
23 Aaron 08.17.06 at 7:10 am: When you find the perfect host let us all know!

Will do!

Despite your cynical tone, if you have a 70-80% uptime and never got any money back then you have poor judgement on hosts or just plain suck. :) If b5media ever was in such a situation, we’d be out of business.
24 Aaron 08.17.06 at 7:10 am: When you find the perfect host let us all know!

Will do!

Despite your cynical tone, if you have a 70-80% uptime and never got any money back then you have poor judgement on hosts or just plain suck. :) If b5media ever was in such a situation, we’d be out of business.
25 Aaron 08.17.06 at 7:10 am: When you find the perfect host let us all know!

Will do!

Despite your cynical tone, if you have a 70-80% uptime and never got any money back then you have poor judgement on hosts or just plain suck. :) If b5media ever was in such a situation, we’d be out of business.
26 MamaDuck 08.17.06 at 12:12 pm: Ah, the perfect host. When you find it, let me know ;). Our list is up if you’d like to look… have a great day!
27 MamaDuck 08.17.06 at 12:12 pm: Ah, the perfect host. When you find it, let me know ;). Our list is up if you’d like to look… have a great day!
28 MamaDuck 08.17.06 at 12:12 pm: Ah, the perfect host. When you find it, let me know ;). Our list is up if you’d like to look… have a great day!
29 MamaDuck 08.17.06 at 12:12 pm: Ah, the perfect host. When you find it, let me know ;). Our list is up if you’d like to look… have a great day!
30 MamaDuck 08.17.06 at 12:12 pm: Ah, the perfect host. When you find it, let me know ;). Our list is up if you’d like to look… have a great day!
31 Douglas 08.17.06 at 12:26 pm: We have a 100% uptime SLA with our datacenter as far as network connectivity, etc. goes. It isn’t their fault if a server dies, but if their network goes down or the power goes out, it is their fault.

For hardware, we guarantee 99.9% to our enduser clients, which is pretty high. 99.5% is getting to be on the lower end of the industry.

Also, be sure to watch our for certain types of uptime guarantees like the ones where if your site is out between .5 and 1% of the time, you only get a 10% credit, etc.
32 Douglas 08.17.06 at 12:26 pm: We have a 100% uptime SLA with our datacenter as far as network connectivity, etc. goes. It isn’t their fault if a server dies, but if their network goes down or the power goes out, it is their fault.

For hardware, we guarantee 99.9% to our enduser clients, which is pretty high. 99.5% is getting to be on the lower end of the industry.

Also, be sure to watch our for certain types of uptime guarantees like the ones where if your site is out between .5 and 1% of the time, you only get a 10% credit, etc.
33 Douglas 08.17.06 at 12:26 pm: We have a 100% uptime SLA with our datacenter as far as network connectivity, etc. goes. It isn’t their fault if a server dies, but if their network goes down or the power goes out, it is their fault.

For hardware, we guarantee 99.9% to our enduser clients, which is pretty high. 99.5% is getting to be on the lower end of the industry.

Also, be sure to watch our for certain types of uptime guarantees like the ones where if your site is out between .5 and 1% of the time, you only get a 10% credit, etc.
34 Douglas 08.17.06 at 12:26 pm: We have a 100% uptime SLA with our datacenter as far as network connectivity, etc. goes. It isn’t their fault if a server dies, but if their network goes down or the power goes out, it is their fault.

For hardware, we guarantee 99.9% to our enduser clients, which is pretty high. 99.5% is getting to be on the lower end of the industry.

Also, be sure to watch our for certain types of uptime guarantees like the ones where if your site is out between .5 and 1% of the time, you only get a 10% credit, etc.
35 Douglas 08.17.06 at 12:26 pm: We have a 100% uptime SLA with our datacenter as far as network connectivity, etc. goes. It isn’t their fault if a server dies, but if their network goes down or the power goes out, it is their fault.

For hardware, we guarantee 99.9% to our enduser clients, which is pretty high. 99.5% is getting to be on the lower end of the industry.

Also, be sure to watch our for certain types of uptime guarantees like the ones where if your site is out between .5 and 1% of the time, you only get a 10% credit, etc.
36 Josh Maher 08.17.06 at 10:14 pm: Don’t forget Gas for those generators :)

http://joshmaher.wordpress.com/2006/08/14/top-ten-reasons-you-should-not-blog/
37 Josh Maher 08.17.06 at 10:14 pm: Don’t forget Gas for those generators :)

http://joshmaher.wordpress.com/2006/08/14/top-t...
38 Josh Maher 08.17.06 at 10:14 pm: Don’t forget Gas for those generators :)

http://joshmaher.wordpress.com/2006/08/14/top-t...
39 Josh Maher 08.17.06 at 10:14 pm: Don’t forget Gas for those generators :)

http://joshmaher.wordpress.com/2006/08/14/top-t...
40 Josh Maher 08.17.06 at 10:14 pm: Don’t forget Gas for those generators :)

http://joshmaher.wordpress.com/2006/08/14/top-t...
41 Matt Craven 08.18.06 at 10:43 am: Great post - and some key considerations here.

I’m a big fan of redundancy and having remote access into the KVM along with remote reboot capabilities. Makes things a hell of alot easier in the long run to be able to take care of those things..

I’ve been with the same hosting company for 3 years now and can’t complain. I had done the colocation thing for 3 years when we were based out in Boston and enjoyed that as well. At least there in a worst case scenerio, I could drive up to Boston and reboot the damn things myself!

Matt
42 Matt Craven 08.18.06 at 10:43 am: Great post - and some key considerations here.

I’m a big fan of redundancy and having remote access into the KVM along with remote reboot capabilities. Makes things a hell of alot easier in the long run to be able to take care of those things..

I’ve been with the same hosting company for 3 years now and can’t complain. I had done the colocation thing for 3 years when we were based out in Boston and enjoyed that as well. At least there in a worst case scenerio, I could drive up to Boston and reboot the damn things myself!

Matt
43 Matt Craven 08.18.06 at 10:43 am: Great post - and some key considerations here.

I’m a big fan of redundancy and having remote access into the KVM along with remote reboot capabilities. Makes things a hell of alot easier in the long run to be able to take care of those things..

I’ve been with the same hosting company for 3 years now and can’t complain. I had done the colocation thing for 3 years when we were based out in Boston and enjoyed that as well. At least there in a worst case scenerio, I could drive up to Boston and reboot the damn things myself!

Matt
44 Matt Craven 08.18.06 at 10:43 am: Great post - and some key considerations here.

I’m a big fan of redundancy and having remote access into the KVM along with remote reboot capabilities. Makes things a hell of alot easier in the long run to be able to take care of those things..

I’ve been with the same hosting company for 3 years now and can’t complain. I had done the colocation thing for 3 years when we were based out in Boston and enjoyed that as well. At least there in a worst case scenerio, I could drive up to Boston and reboot the damn things myself!

Matt
45 Matt Craven 08.18.06 at 10:43 am: Great post - and some key considerations here.

I’m a big fan of redundancy and having remote access into the KVM along with remote reboot capabilities. Makes things a hell of alot easier in the long run to be able to take care of those things..

I’ve been with the same hosting company for 3 years now and can’t complain. I had done the colocation thing for 3 years when we were based out in Boston and enjoyed that as well. At least there in a worst case scenerio, I could drive up to Boston and reboot the damn things myself!

Matt
46 Aaron Brazell 08.18.06 at 4:17 pm: KVM is a great point. Thanks for throwing that in, Matt.

Josh: Not my problem. :)
47 Aaron Brazell 08.18.06 at 4:17 pm: KVM is a great point. Thanks for throwing that in, Matt.

Josh: Not my problem. :)
48 Aaron Brazell 08.18.06 at 4:17 pm: KVM is a great point. Thanks for throwing that in, Matt.

Josh: Not my problem. :)
49 Aaron Brazell 08.18.06 at 4:17 pm: KVM is a great point. Thanks for throwing that in, Matt.

Josh: Not my problem. :)
50 Aaron Brazell 08.18.06 at 4:17 pm: KVM is a great point. Thanks for throwing that in, Matt.

Josh: Not my problem. :)