Monitoring Fundamentals - The Practice of Cloud System Administration - page 333

16.1.2 Service Management
Lastly, here are some service management terms we will use occasionally. They mostly
come from the business world:
• Service Level Indicator (SLI): An agreement as to how a measurement will be
measured. For example, it might define what is measured, how it is measured, and
from what perspective.
• Service Level Target (SLT): A target quality of service; in other words, an SLI's
expected minimum or maximum. Until ITIL V3, this was called the Service Level
Objective (SLO) . An example SLT might be a certain level of availability or the
maximum permitted latency.
• Service Level Agreement (SLA): The contract that states the SLIs, SLTs, and the
penalties if the SLTs are not met. For example, there may be a refund or penalty
payment for any outage longer than one hour. The term SLA is often overused to
mean SLT or any service level requirement.
16.2 Consumers of Monitoring Information
There are many consumers of monitoring information, each of which has different needs.
Peoplewhomonitoroperationalhealthneedtoknowstatechangesimmediately forfastre-
sponse—they need real-time monitoring. Capacity planners, product managers, and others
needmetrics collected overalongperiodoftimetospottrends—theyneedhistorical mon-
itoring data.
A more fine-grained way to differentiate consumers is the Dickson model ( Dickson
2013 ) , which uses three characteristics: resolution, latency, and diversity. These character-
istics can be rated as high or low.
Resolution describes how frequently the metric is collected. High (R+) is many times a
second, minute, or hour. Low (R-) is many times a day.
Latency describes how long a period of time passes before the information is acted
upon. Low (L+) is real-time response. High (L-) means data is stored and analyzed later,
perhaps used for daily, weekly, or monthly statistics. (Note that the use of + and - are re-
versed from R and D. Think of + as the case that is more difficult to engineer.)
Diversity describeshowmanymetricsarebeingcollected.High(D+)meansmanymet-
rics, perhaps many measurements about many different services. Low (D-) means there is
a focus on a particular or small set of metrics.
Consumers can be described by a 3-tuple. For example, (R+, L-, D+) describes a high-
resolution, high-latency, high-diversity consumer.