Some Data Centre/Colocation Service Level Agreements (SLA’s) can be extremely misleading, complex and are often quoted during the sales process to imply reliability.
Most Data Centres have extremely robust infrastructure; but don’t assume an SLA will offer you much protection.
SLA’s usually refer to three areas of the data centre infrastructure;
· Power
· Atmosphere (temperature, humidity, particulate level)
· Connectivity
Each of these then has the following:
· The Uptime Promise
· Method of measurement
· Compensation level
1. Uptime %
You would imagine this is the main focus of the agreement (well it can be for the marketing department). It is usually expressed as a % of availability over a time period, eg 99.995% & this is sometimes converted into the number of hours/minutes/seconds of downtime per month that triggers the SLA. It can, however, be meaningless depending on the following:
2. Measurement Method
a) Power is straightforward you would think, it is pretty much on or off. However a momentary loss, change of voltage or frequency can cause huge problems for server infrastructure with potential data loss & corruption, but not trigger the SLA even if it is 99.995%. Most facilities offer two power feeds & the SLA is usually built around both feeds failing, if one is up & your infrastructure fails, it’s your problem.
b) Cooling
The SLA will give you a window of operating temperatures, for example 22-26 degrees C.
Firstly you need to see where these are measured from, is it in front of the racks (which you can measure) or in the cooling supply chamber under the floor or overhead (which you can’t). Temperatures can vary within any data centre aisle as some racks will draw much more power and air than others, this can mean your rack is much hotter than others but the aisle stays ‘within SLA’. Is it measured as an average? Some facilities have client portals that give full visibility of environmental data. If your data centre doesn’t, it is worth asking for access to it. Just remember to ask where the sensors are & do they relate to your SLA?
c) Humidity
Many of the newer cooling systems supply filtered outside air. The downside to this is that there is less control over humidity, particularly as water evaporation is often used to cool the air, which increases the moisture content. High or low humidity can be terminal for servers (if you’ve read this far I’m sure you won’t need to ask why), consequently it is frequently left out of some SLA’s. If it is in there take a look at where & how it is measured.
d) Contamination
Not common in many SLA’s but it should be, the same questions apply – how and where is it measured? I’m sure you have opened a server at some point to find a layer of fine dust inside – not ideal.
e) Connectivity
Connectivity SLA’s can be really tricky because of the potential complexity of networks themselves so I will write a separate article, as quite often these are supplied by third parties.
3. Compensation
For most outages, the potential financial & reputational loss to your company or your clients is not proportional to what you will get as compensation. Compensation is usually a % of your DC charges. So navigate through the SLA & prove the DC was at fault, it will usually add insult to injury by offering a pitiful amount.
In Summary – Do your homework on the facilities infrastructure, have a critical look at the SLA & set up your own monitoring.