5.2Service Management

A system only delivers value through ongoing operation that continues to provide value, not just through being built. Among the service management practices that ITIL organizes, distinguishing incident/problem/change management and availability calculations (from MTBF/MTTR, and combining series/parallel configurations) are recurring, important domains on the AP exam.

5.2.1ITIL and SLA/SLM

• ITIL (Information Technology Infrastructure Library) is a collection of best practices for IT service management and a de facto international standard. Using the service lifecycle (strategy -> design -> transition -> operation -> continual improvement) or, more recently, the service value system, it treats everything from service planning to improvement as a continuous cycle.
• SLA (Service Level Agreement) is a formal agreement between provider and user that documents target service quality levels (availability, response time, recovery time, and so on). SLM (Service Level Management) is the management process that continuously monitors, reports on, and improves the targets defined in the SLA. The distinction: an SLA is "the content of the promise," while SLM is "the mechanism that keeps the promise over time."

5.2.2Incident, problem, change, and configuration management

• Incident management is the process whose goal is to restore service as quickly as possible after a fault or service disruption occurs (identifying the root cause is not the goal--stopgap response comes first). Problem management is the process whose goal is to identify the root cause and prevent recurrence. Understand the division of roles as "incident management = stop the bleeding, problem management = treat the underlying cause."
• Change management is the controlled process of assessing, approving, planning, and reviewing changes to infrastructure and services, preventing new failures caused by uncontrolled change. Configuration management is the process of accurately maintaining configuration information for IT assets (hardware, software, documentation, and so on), recording relationships between IT assets in a CMDB (Configuration Management Database) that serves as foundational information for incident response and change impact analysis.

5.2.3Capacity management, availability management, and availability calculations

• Capacity management is the process of securing processing capacity (CPU, storage, network bandwidth, and so on) that matches current and future business demand--neither too much nor too little. Its goal is to avoid both over-investment (higher cost) and under-capacity (degraded performance). Availability management is the process of managing whether a service continues to meet its required availability, quantified using MTBF (Mean Time Between Failures) and MTTR (Mean Time To Repair).
• Availability = MTBF ÷ (MTBF + MTTR) (the fraction of time the system is operating normally). For a series configuration (the whole thing operates only if every unit is normal), overall availability = the product of each unit's availability. For a parallel (redundant) configuration (the whole thing operates as long as at least one unit is normal), overall availability = 1 − (the product of each unit's unavailability). Understand the differing idea: parallel configurations shrink the combined unavailability (the probability that everything has failed) by multiplying the individual unavailabilities together.

Take the operating data of a core server as an example of availability calculations. Suppose historical operating results give this server MTBF (Mean Time Between Failures) = 270 hours and MTTR (Mean Time To Repair) = 30 hours. Availability = MTBF ÷ (MTBF + MTTR) = 270 ÷ (270 + 30) = 270 ÷ 300 = 0.9 (90%). Now suppose two units are operated in a series configuration (the whole service stops if either one fails--for example, a configuration where a request passes through different-role units in sequence): connecting a load balancer with availability 0.95 in series with an application server with availability 0.98, the overall availability is the product of the individual availabilities = 0.95 x 0.98 = 0.931 (93.1%)--lower than either alone. A series configuration's key property is that overall availability drops the more units the request passes through. On the other hand, suppose a single server with availability 0.9 is judged insufficiently reliable, so a second server with availability 0.8 is added in a parallel (redundant) configuration. In a parallel configuration, service continues as long as both units do not fail at the same time, so overall availability = 1 − (the product of the individual unavailabilities). The first unit's unavailability is 1 − 0.9 = 0.1, and the second's is 1 − 0.8 = 0.2, so overall availability = 1 − (0.1 x 0.2) = 1 − 0.02 = 0.98 (98%)--substantially higher than either unit alone (0.9). Understanding this contrasting property--a series configuration's availability drops the more availabilities are multiplied together, while a parallel configuration's availability rises the more unavailabilities (failure probabilities) are multiplied together, shrinking the combined figure--is the key to reliably solving availability management calculation problems.

5.2.4The service desk and facility management

• The service desk is the single point of contact that receives inquiries and fault reports from users. Cases that cannot be resolved on first contact are escalated to the appropriate incident management or problem management owner. Facility management is the management area that supports availability from the physical facility side--protection against momentary outages and power failures via a UPS (Uninterruptible Power Supply), and a data center's air conditioning, earthquake resistance, and access control. A UPS cannot sustain a long power outage; its role is strictly to bridge to a switch-over to a generator or to a safe shutdown.

5.2.5Section summary

• Incident management = restore service quickly. Problem management = identify the root cause and prevent recurrence. Change management = assess, approve, control. Configuration management = accurately maintain IT asset configuration via a CMDB
• Availability = MTBF / (MTBF + MTTR). Series = the product of the individual availabilities (drops as more units are added); parallel = 1 - (the product of the individual unavailabilities) (rises as more units are added)
• Distinguish SLA (the agreed quality targets) from SLM (continuous monitoring/improvement of those targets). A UPS's role is to bridge momentary outages/power failures (until generator switch-over or a safe shutdown)