Once purely the domain of IT and infrastructure organizations, service level agreements (SLAs) have become strategic business tools throughout the enterprise. The proliferation of as-a-service (aaS) offerings places newfound interest and importance in what SLAs should (and do) cover.
Following are five questions you should ask about your provider’s SLAs:
1. What Do Your SLAs Cover?
When organizations purchase multiple solutions from a provider, it is often the case that there are multiple SLAs involved in the agreement. This may not be immediately obvious in the conversation (or in the SLA itself), but it is something of which you need to be extremely aware.
It is not uncommon for one component of a solution, say, a cloud compute instance, to have a SLA that is different from a cloud storage instance, which is different from the bandwidth SLA. In all, a seemingly simple “compute, storage, bandwidth” agreement could include three or more separate SLAs. When you add in firewalls, other networking components, and dedicated servers, these, too, will often carry their own SLAs.
When considering an IT infrastructure solution, be sure to directly ask whether or not the SLA is all-inclusive. If it isn’t (which is often the case), get copies of each SLA and establish a clear understanding of the terms and implications of having multiple SLAs associated with your agreement.
2. How are the 9s measured?
Providers typically measure their SLAs on a monthly, quarterly, or annual basis. The devil here, as always, is in the details. For each SLA, be sure there is a clear understanding how the 9s are measured. Each SLA is going to measure availability differently. Some providers will even have different measurement systems for different service offerings, all of which comprise a single solution for your organization.
3. What do the 9s really mean?
One of the most frequently overlooked elements of SLAs is the value of the 9. When you compare varying SLA measurement terms, not all 9s are created equal. A 99% SLA, when measured monthly, factors in more than seven hours of downtime per month.
Uptime Percentage SLA & Expected Downtime Calculations:
| Uptime | Downtime/Day | Downtime/Week | Downtime/Month | Downtime/Quarter | Downtime/Year |
| 99.999% | 0:00:00.9 | 0:00:06 | 0:00:26 | 0:01:18 | 0:05:15 |
| 99.99% | 0:00:09 | 0:01:00 | 0:04:19 | 0:12:58 | 0:52:34 |
| 99.9% | 0:01:26 | 0:10:05 | 0:43:12 | 2:09:36 | 8:45:36 |
| 99% | 0:14:24 | 1:40:48 | 7:12:00 | 21:36:00 | 87:36:00 |
As you can see, the 9s mean a lot.
4. Does one more 9 really matter to my business?
The answer to this question is unique to every organization. If seven hours per month of unscheduled downtime is something your business can absorb, then a monthly 99% SLA may be appropriate (and cheaper than a 99.99% SLA).
A simple calculation that will put this into a very tangible context is to multiply your revenue per hour gained from the systems exposed to the SLA by the downtime measurement. If you can absorb this revenue hit, you can comfortably operate within that SLA. If you can’t, you need a higher-level SLA.
The higher your availability requirements, the more 9s your SLA needs to cover.
Some may think this is trivial, and in many ways, it is. But it is also critically important, and something we work with prospects to understand every single day. The cost difference between 99% and 99.999% is real. So are the revenue and downtime implications, which are too often missed in an attempt to save money.
5. Is the SLA applicable to all availability zones/data centers?
This question is a bit more specific to high availability solutions (leveraging geographic dispersion for ongoing operations in the event of a natural disaster).
SLAs are often data center-specific or cloud specific. This means that the SLA provided in one data center or cloud is separate from the SLA provided in another data center or cloud. Translation? The portion of your application running in one location has a completely different uptime expectation (and reality) than the portion running in another location.
While this question may seem obvious, it’s is critically important. SLA variance across datacenters is extremely common and is just as often overlooked.
The Implications of Asking
We interact with hundreds of organizations each and every day. In most (if not all) of our prospect engagements, we’re discussing SLAs. We think it’s important to take the time to walk through this issue and be sure that everyone involved understands exactly what is needed, what ServerCentral provides, why we provide it, and most importantly, what it means for their business.
As the complexity and importance of the SLA continues to rise, you can always contact us for help decoding the 9 print.
