One of the biggest lessons learned by many of Salesforce.com's 2.1 million global subscribers, following the cloud computing giant's recent catastrophic power outage, is that support problems don't disappear when applications or infrastructures move to the cloud.
As the cloud-based service provider's servers remained inoperative for an agonisingly long period, there seems little doubt that a sizable number of its 86 000 corporate customers were revisiting their service level agreements (SLAs) and quickly reading through the fine print detailing Salesforce.com's support targets.
A failure of this magnitude highlights the need for end-users to first align their own internal support infrastructures with their cloud provider's before signing any agreement. This is of growing importance, as providers continue to host ever-increasing numbers of mission-critical applications. This effectively raises the stakes when it comes to selecting an appropriate cloud service and dependable service provider.
By addressing these issues, individually, with a range of service providers ahead of time, there is a good chance that providers without solid, bullet-proof failure plans will be eliminated from the short-list.
Of course, even a watertight SLA is not a guarantee of a service provider's reliability, but more of an indication or guide as to how the company will react in a time of trouble.'10 000% guaranteed SLA'. When cutting through the marketing hype, it is clear GoGrid is offering a 100% uptime guarantee, which if there is a failure to deliver, will be compensated by the customer receiving 100 times the fee paid for the uptime denied.
How dependable is GoGrid's promise? The company's track record may well hold clues to the answer. However, not all vendors are free with this information. In fact, sourcing long-term reliability numbers is often a thankless task. Customer testimonials are better bets, while more trusted comparisons are available from organisations such as CloudSleuth and CloudHarmony.
If the operational success rate of a cloud service provider is so difficult to define, what are the indications of a sound basic infrastructure?
Even a watertight SLA is not a guarantee of a service provider's reliability.
Power outages are not the only cause for concern. How many times have South Africans been told by Internet service providers that “all the international bandwidth is down”?
For SA-based organisations, a suitably backed-up cloud service model could be based on that of a (hypothetical) provider which has a prime local data centre, with a second in the US or Europe and a third backup in the Far East or Australia.
If power and bandwidth are key criteria for 100% uptime, then so is data security. With significant advances in this area, many providers are now able to implement deep multilayer security architectures. But they could move up a level by introducing completely different vendor equipment in their prime and secondary data centres.
This would have to encompass everything from server technology to networking and storage platforms to protect users (and themselves) from any of the 'single-vendor' bugs and other increasingly common vendor-related potential vulnerabilities out there.
In addition to security threats, providers should also strive to achieve compliance with security-related standards. One gaining the most relevance is PCI (Payment Card Industry), a set of standards developed to protect credit card information during and after a financial transaction.
Today, choosing the best, or most appropriate, cloud provider for an application presents organisations with multidimensional challenges. These challenges will only increase over time, mirroring the increase in the numbers of service providers computing in the market – many of them now beginning to specialise.
Their marketing messages promise new levels of data protection, security, system performance and interoperability to suit specific applications. Some are ecologically focused and boast reductions in data centre power consumption and cooling. Many of these providers have adopted the cloud computing approach known as 'follow the moon'. The basis of this model is that the provider will have a number of physical data centres in several locations, and run the applications that are active from the 'day side' of the world in centres on the 'night side', thus taking advantage of lower power and cooling costs.
Perhaps the ultimate data centre will one day be located on the moon. Cooling shouldn't present a problem and power will be freely available from consistent solar energy.
Our comments policy does not allow anonymous postings. Read the policy here