2.1SLA/OLA/UC & service-level design

For a service manager, the service level promised to customers is not "the higher the better." Raising an availability target by even one notch can send the cost of redundant configuration, monitoring, and maintenance staffing soaring, and setting an unachievable target invites chronic SLA breaches. This section builds on the three-layer structure—the SLA agreed with the customer, the internal OLA that supports it behind the scenes, and the UC with external suppliers—to learn how SLM (service level management) keeps them aligned and how to design and maintain service-level targets under constraints of cost and feasibility.

2.1.1The three layers: SLA, OLA, UC

• An SLA (Service Level Agreement) is a document agreed between the service provider and the customer that defines the quality levels of the service delivered (availability, response time, resolution time, etc.). It is an external promise specifying the service level the customer actually experiences, and each SLA target cannot be met unless it is backed internally by the OLA and UC described below.
• An OLA (Operational Level Agreement) is an internal agreement between operations teams within the same organization. For example, the service desk, server operations team, and network operations team agree on internal roles and response targets such as "first-line triage within N minutes" and "recovery work within N hours." The OLA is the internal backing that makes the SLA achievable, and it is not made with the customer.
• A UC (Underpinning Contract) is a contract made with an external supplier (a vendor, carrier, cloud provider, etc.). It secures, as a contract, the outsourced portions needed to meet SLA targets (maintenance response time, line availability, etc.). If the UC's levels fall short of the SLA targets, the SLA can no longer be met due to external factors.

2.1.2Maintenance and improvement by SLM

• SLM (service level management) is the process of agreeing, monitoring, and reviewing SLAs, reporting achievement, and continually maintaining and improving them. SLM periodically measures actual results, checks whether each SLA target is being met and whether the OLA and UC underpin those targets, and, where gaps exist, drives corrective action or revision of the SLA.
• The essence of SLM is keeping "the targets agreed with the customer" aligned with "the internal and external agreements (OLA, UC) that underpin them." If an SLA target is raised, the corresponding OLA response targets and UC maintenance levels must be raised at the same time; otherwise only the external promise moves ahead and becomes impossible to keep.

A service manager, having received a customer request to "raise the monthly availability target for the core-business operations service from 99.5% to 99.9%," is deciding how to redesign the service-level target. In terms of allowable monthly downtime, 99.5% is about 3.6 hours per month, but 99.9% shrinks to roughly 43 minutes per month, making the required speed of maintenance and recovery an order of magnitude stricter. Before casually agreeing to this new target as an SLA with the customer, the manager must first confirm it can be underpinned internally and externally. Internally, the manager reviews the OLA that defines internal response targets from fault detection through first-line triage and start of recovery; since the current "first-line triage within 30 minutes" level cannot possibly stay within the 99.9% allowable downtime (about 43 minutes), the manager evaluates whether the OLA response targets themselves can be shortened through monitoring automation and strengthened on-call staffing. Externally, the manager checks the UC with the vendor to whom hardware maintenance is outsourced; if the current contract is "next-business-day on-site maintenance," recovery from a component failure could take nearly a full day, revealing that 99.9% is physically unattainable under this UC. The UC must therefore be revised to a higher maintenance level such as "on-site within 4 hours," or the design switched to a redundant configuration that keeps the service running through a single-component failure—both of which increase cost substantially. The judgment the manager should make here is not "promise 99.9% exactly as the customer asked," but rather to present the OLA/UC backing and the cost increase to the customer and re-agree on a target level commensurate with the cost the customer is willing to pay. Setting an SLA target with no backing merely invites chronic SLA breaches, and the role of SLM is precisely to adjust to an achievable, cost-effective service level while keeping these three layers aligned.

2.1.3Section summary

• SLA = provider-customer, OLA = between internal operations teams, UC = contract with an external supplier; who the agreement is between is decisive
• SLM keeps SLA/OLA/UC aligned and continually maintains and improves them by measuring and reviewing actual results
• Set service-level targets by weighing availability, cost, and feasibility; never promise a high target that has no backing