Friday, 13 June 2008
Service Level Agreement (SLA) is a commonly misused notion in server
hosting and colocation business. Our article Server Hosting and SLA 99,9999
presents a down-to-earth view of this important aspect.
Most of the server
hosting clients require SLA (Service Level Agreement) prior to
signing a colocation contract.
SLA for Server Hosting
SLA specifies in % the
total uptime / downtime of a server availability through the
provider´s network on one side, along with provider´s
penalization due to failing in network availability. The higher the
SLA the higher the price of such service. No surprise if a 0.5%
higher SLA resulting in double the price.
Most of the server
colocation customers are satisfied with a SLA of some 99 point
something per cent. Lets look at a typical server colocation SLA:
„..our common Service Level Agreement covers 99.0 %
availability, upgradable to 99.6 or up to 99.96%“.
Working few seconds with
a calculator is enough to see how these high availability numbers
fade away in reality. With 99.0% availability tech provider is allowed
for a 7hrs 12min outage in a months. With SLA of 99.6% it is
slightly better, but still not many customers will be satisfied with
2hrs 52min downtime per month. And even the „flagship“ SLA of
99.96% still permits the provider for 18min down time per months
without any penalty.
It is necessary to say
that there is no provider of absolutely zero down time. To look
properly after the network some planned outages are just unavidable.
These instances (network maintainance, router installation etc.) are
normaly not covered by SLA, under the provision of advance warnings
and carefull planning to minimize impact on customer servers by
shortening the outage and planning for low traffic periods.
Just to give a concrete
example: at the Coolhousing server hosting data center during the
last 12 months (June 2007 – 2008) there was none of network outage
except of 3 maintainance instances of a total of 40 minutes. This
represents a 100% availability, or 99,9999761% availability when
including the planned service outages.
 |
