On 6 August, 11 August and 15 August 2008, Google’s enterprise e-mail system, Apps Premier Edition, shut down and disrupted some organizations and users that rely on the service. One of the outages was systemwide, affecting nearly all users for approximately two hours; the other two outages affected part of the user base for less than 24 hours.

On 28 August 2008, Google said it would credit all its customers with 15 days of service (even if they were not impacted), which is more than the contracted penalty for not meeting its 99.9% service-level agreement (SLA) commitment. Google also promised it would improve its communications regarding future outages by implementing a dashboard and by offering post-mortem calls.

Business executives who source applications through software as a service (SaaS) or cloud models often assume that the provider will deliver the service with minimal to no unplanned service interruptions. The prevailing perception is "turn it on and it is always there" — but the reality regarding service availability is not so simple. Within the cloud, the traditional complexities of managing data centers, infrastructure and applications still persist, but the service provider is responsible for uptime. The critical difference between premise deployments and SaaS/cloud architectures is that the latter are typically far more complex and must support thousands of customers simultaneously.

While organizations can set expectations with their service provider via an SLA, there are no guarantees those SLAs will be met. In addition, it is likely that no contract penalty will be able to compensate for the lost business or damaged company image associated with an unplanned mission-critical service outage. When Google misses SLAs, it offers what is typical in the industry: a certain number of days of “free” service (often 30 days; Google offers 15 days). Uptime is only one characteristic of service quality; Gartner believes that service delivery SLAs are insufficient if they are overly dependent on a single metric. To be effective, SLAs should also take into consideration other factors, such as performance, privacy, data rights, business continuity/disaster recovery and recovery time

In light of the e-mail service failures, Google has promised to build a client-accessible dashboard, which it states will be available in a "few months," and will provide data on the nature of the problem and the estimated time-to-resolution. We believe this overdue action — along with the need for a better change management communication process — reveals Google's immaturity in selling commercial applications as well as the overall immaturity of the SaaS market.

Before signing up for service, enterprises should:

• Review the vendor's uptime past history.
• Prepare for the contingency of subpar performance by implementing plans to migrate to an alternative service, and understand the costs, resources, time and other implications of doing so.
• Contractually reserve the right to cancel services without penalty, should SLAs be consistently missed.

• "IT Operational Considerations for Cloud Computing” — Gartner advises IT operations groups on how to evaluate the currently proliferating cloud services with respect to operational risks. By Donna Scott
• "The State of Google Apps” — Gartner explores the business model behind Google Apps and the usage model with which it is entering enterprise operating environments. By Tom Austin and others

(You may need to sign in or be a Gartner client to access the documents referenced in this First Take.)