Hitchhiker's Guide to Streaming Media: Service Level Agreements
>>>Lookup: Service Level Agreements
Service Level Agreements, commonly called SLAs, form a key part of a contract with a service provider. They should be found lurking in documents with very small print which you should read thoroughly before you sign up for services form that provider that form part of your overall contracts agreed when you take out the service.
The reality is that at the point you sign up for the service it is only fair to say that the small print is too small, and so it gets laid to one side with a resolve to read it at some later point (after you have agreed to the terms of it) and reconcile this with the fact that you think that “because everyone else must have signed it, it must be ok.”
Amusingly, like all contracts, you actually will typically never get around to really reading this until the point where the service doesn’t deliver you what you expect, and doubtlessly the SLA will protect the service provider from any requirement on their part to modify their service to actually provide you with what you thought you were going to pay for. Wonderful.
Here are some of the key terms that you should, if you do nothing else, look for BEFORE you sign up to the service provider’s SLA:
Often, for services such as those provided by content distribution networks and streaming service providers, you will find a lot of figures ending with percentage signs. For example “99.999% uptime” or “99.95” availability.
What does this mean? Well a service provider whose network is only available for XX hours out of YY in a month is actually still delivering you a 99.999% service.
This can mean hours. Have a look at the chart here
And here’s a quote from that site: “Availability is most often expressed as a percentage. Sometimes, people will refer to “four nines” (99.99%) or “five nines” (99.999%). To simplify things, the table shows the minutes of downtime allowed per year for a given availability level.”
So as you can see a “five 9s” SLA will still allow outages of around an hour over each year. An SLA of “Five 9s” can provide a service which should be available for all but 5 minutes in a year. You just have to hope that those 5 minutes are not in the middle of a key live event!
The article quoted above also notes that “if the system was composed of two servers, and each server had an expected availability of 99%, then the expected availability of the system would be 99% * 99% = 98.01%”
So complexity has a significant effect on the SLA offered. If, for example, you are running a fully resilient contribution architecture with two encoders and these were rated with an SLA of 99% each and the network CDN was providing an SLA of 99.99% then you would end up with an effective SLA of 99*99*99.99 = 98%. This could result in up to 175 hours of outages a year. This makes all the promises implied by all the 9s look a little less impressive. IsThat’s potentially more than a week off air.
The other key thing that is often quoted in SLAs is 24/7 support. It is really worth drilling down into what this means. Some service providers consider that having a web page with a simple “submit your query to the void now” button as providing 24/7 support. Other have a room full of guys staring at a wall of monitors or a full-blown MCR (Master Control Room) or NOC (Network Operations Center) with direct remote access to all levels of their service.
Still others will just have a duty engineer who you can try to wake, or hope he is not already tied up.
So, if you can, try to test their support services BEFORE you sign that contract!
Finally, and one of the most interesting sides of SLAs: Talk to other clients about their experience of outages, and ask about how the SLA was honored when it needed to be applied. There is a disturbing culture in all service sectors which use SLAs to defend themselves if accusations are thrown. They use interpretation of the SLA to re-position the responsibility outside of their domain, to pretend that it wasn’t their problem and that they should take no responsibility for the fault.
In nearly all cases there is some technical report produced (which the client cannot disprove) or contractual term stuck to which seems to excuse them from exercising any action according to the SLA. They often ignore the fact that in business the customer is ALWAYS right and so the failure to honour an SLA “in spirit,” even when litigation may fall in their favour, actually results in increased churn on their contracts, since whether or not the service provider was in the “right” or not, a client with half a webcast of black silence who “feels” it is their service provider’s fault will change service provider at the first opportunity.
In the author’s experience, he is only aware of one traditional broadcast facilities provider and one content distribution network who actually proactively have contacted their clients and volunteered information about SLA failures and subsequently refunded service fees against this failure.
“Taking it on the chin” for both these providers resulted in a huge confidence uplift from their clients and renewed and lengthened contracts.
After all, in business, trust is worth a thousand contracts.