Cloud Service SLAs – Better than their Reputation?

You often hear the statement that service level agreements for cloud services (Cloud Service SLAs) are too “vague”. For example, it is claimed that the provision of the service and quality expected by the customer cannot be ensured due to unclear regulations.

However, SLAs must be clearly defined and their degree of achievement must be made transparent via key figures. This is the basis for being able to prove non-fulfillment or poor performance and claiming customer rights.
Are these now too unclearly defined in practice, are key figures missing or are they not meaningful? Are services not being provided sufficiently well as a result, without this having consequences for the cloud provider? Using concrete examples from Microsoft and Amazon Web Services (AWS), we take a look at what is really in the standard contracts / general terms and conditions (GTCs) of cloud providers and whether customers are able to adequately control cloud providers and their services on this basis.

Regulations in Standard Contracts / GTCs

The service level agreements (SLAs) are part of the standard contract with the cloud provider. Microsoft Azure and Amazon AWS offer separate SLAs per cloud service for their infrastructure services.

At this point, the SLA for the virtual server services Microsoft Azure Virtual Machines and Amazon AWS EC2 are considered as examples.

Figure 1: Cloud Services Provisioning Types

The SLAs for Azure Virtual Machines and AWS EC2 specify service levels for availability only. These service levels are differentiated by the type of server deployment: single or multiple server instances that can be replicated in multiple isolated availability groups in a data center. Based on different levels of resilience due to the deployment type of the cloud service, the following service lebels result based on the availability metric:

Table 1: Service Level Availability

The percentage of service availability refers to the service usage within a monthly billing period; at least one server instance must remain up and running for this. Microsoft also differentiates the availability of individual virtual servers according to the type of virtualized data storage used.

Technical support and response times are specified independently of service at both cloud providers with free or additional chargeable support plans. These differ significantly in terms of scope of services, service times and response times. This means that the right support plan can be selected according to the company’s requirements for the cloud services used.

Consequences in the Event of Poor Performance or Non-Performance

If the minimum values specified in the GTCs are not achieved, service credits are agreed in the SLAs as compensation. They correspond to penalty regulations in classic outsourcing contracts. Both cloud providers offer the same service credits, but in slightly different amounts.

Table 2: MS Azure Virtual Machines Availability in Percent

Table 3: MS Azure Virtual Machines Availability in Percent

Table 4: Amazon AWS EC2 Availability in Percent

In some cases SLAs are suspended, which is not uncommon with traditional outsourcing. In the event of force majeure, such as natural disasters, fixed maintenance windows resulting in planned downtime, or non-availability through the customer’s own fault.

Standard Reporting

In order to be able to check and monitor the quality of the service provided, reporting is necessary that shows the degree of compliance with the agreed service levels and the times taken to resolve faults. For this purpose, both cloud providers provide the customer with dashboards that show the service status and provide information on faults and their rectification in all regions (MS Azure Health Status & Dashboard, Amazon AWS Health Dashboard).

Cloud Service SLA Assessment

The contracts of the two cloud providers are comparable in terms of SLAs and offer similar provisioning service options. For these services, availability is defined as a key performance indicator in the SLAs in addition to the specifications in the support plans.

In classic outsourcing, other key figures are usually agreed in addition to service availability to ensure service quality, e.g.:

– Provisioning time / initial provisioning time of a server / virtual machine.

– Maximum downtime (in combination with availability)

– Maximum recovery time

Due to automation of the provisioning process for the cloud services listed provisioning time is usually very short. Maximum downtime and maximum troubleshooting time are limited upwards by the availability target value. In many cases, the service levels offered and reporting by cloud providers will be sufficient.

In the event of poor performance or non-performance of the service, service credits are the sole compensation – but the customer must generally claim these themselves: This means additional effort for the cloud provider manager. Beyond this, no further claims are regulated in the contract: damages from lost profit, lost revenue, business interruptions or loss of business information as a result of non-availability are excluded.

In principle the general terms and conditions of cloud providers are similar to regulations in traditional outsourcing contracts.

Conclusion – Cloud Service SLAs

The cloud services of the cloud providers Microsoft and Amazon shown here as examples are designed in such a way that various customer requirements can be covered. The SLAs reflect this in the availability KPIs and quality levels of the cloud services.

Customers of the cloud providers can check the contractually promised and actually provided services via the cloud provider’s standard reporting.

In each individual case, depending on the application scenario, it must be checked which specific business requirements must be met. Furthermore the GTC of the cloud provider, especially offered service levels have to be checked to see if they fit. Higher service availability is generally also associated with higher costs for the cloud service.

If the regulations of the GTC do not match the requirements of your own company, the following applies: There is generally little scope for individualization in the cloud standard contracts. Therefore, when using cloud services, GTC and especially the SLAs must be examined and possible risks identified and weighed up. In the case of intolerable contractual regulations, a renegotiation should at least be attempted. But success cannot be garanteed as it usually depends on the size of the customer company and its market power.