Performance and Resilience - Can You Have Both?

It is widely accepted that when it comes to outsourcing IT, performance and reliability are the most sought after end-user attributes. However, in looking for a high-speed solution that also offers the highest level of availability, many businesses end up having to compromise one for the other, or must purchase expensive additional services to satisfy the requirement for both. Customers looking for 99.999 percent (the ‘five nines’) availability or higher may need to invest in additional disaster recovery infrastructure, paying associated support and licensing costs as a result.

This article will discuss the true meaning of high availability and debunk some of the misconceptions surrounding the ‘five nines’. It will also explore whether the ‘average’ business really needs such high availability and high performance solutions – and the role service providers play in accurately scoping customer requirements.

The truth about high availability

When it comes to high availability, there are many myths of which customers should be aware, starting with the ‘nines’ system.

Businesses and service providers tend to discuss availability in terms of ‘nines’, as this gives an easy to understand indication of the level of service to be delivered. Generally speaking, 99.999 percent uptime is perceived as a requirement for many businesses, but what does this mean in real terms?

Those service providers offering five nines are essentially promising no more than 5.26 minutes of annual of downtime (that’s just 6.05 seconds per week). Going further still, those offering six nines only afford themselves 31.5 seconds of downtime every year (0.605 seconds per week). The ‘average’ business can reasonably absorb around an hour’s downtime every night without lasting impact on customers or users, so the fact that 99.999 percent uptime is often regarded as the minimum acceptable level is just one of the myths surrounding availability. There are exceptions to this, such as applications serving international markets and systems that need to be accessed 24/7 for backup and bath processing. What matters in all cases is that any downtime is scheduled rather than unscheduled and disruptive. 

The reality is that some service providers offer five nines (or greater) availability to businesses that don’t really need it and furthermore, these providers often cannot deliver to the true meaning of the term. ‘Too good to be true’ guarantees can be vague about what the availability actually applies to, whether it be IP breakout services, operating systems, applications, or virtual servers - it is actually quite challenging to have an uptime SLA that applies across the board.  Secondly, some service providers intentionally design availability SLAs in such a way that the penalty for failure is so minimal (usually service credits or a fee that is a fraction of that customer’s monthly spend), that it becomes irrelevant to their bottom line if and when a failure does occur. 

The final myth is the concept of scheduled and unscheduled downtime.  Many providers will exclude scheduled downtime from their SLAs, which enables them to advertise high availability – claiming that they run at five nines or even 100 percent availability (for unscheduled) despite regular or frequent scheduled downtime.

It is essential that customers bear these common pitfalls in mind for every IT outsourcing decision to ensure that they are not influenced by the promise of high availability without proven ability to deliver.

The need for speed

Ensuring consistently high IT performance will almost certainly be influenced by the level of resilience and availability being offered for the same service. Synchronous storage replication (where a copy of the data is written to a second location at the same time as the original location) has a measurable impact on system performance yet is necessary to achieve high levels of resilience. This is slower than maintaining a single copy of the data and a great example of what appears to be a performance vs resilience dilemma.

However, with significant investment in high speed, low latency infrastructure, such as dark fibre, it is possible to achieve the goal of adequate or appropriate performance with the smallest possible theoretical delay by enabling the storage replication to occur at the highest possible speed. Service providers can also deliver high performance components and infrastructure with no single point of failure. In our experience this works most effectively when using infrastructure from established hardware and software vendor alliances.

The ‘average’ business will often be daunted by the level of expenditure, expertise and time required to deliver even one of the axes of performance or resilience, let alone both. If you are outsourcing, it’s essential to find a competent service provider who has demonstrated an investment and a capability and understanding of the themes explored in this article.

By using best of breed technology and by working in strategic partnership with leading vendors, it is now possible for service providers to offer customers greater levels of performance and resilience than ever before, raising the bar for the ‘average’ business and rendering compromise less necessary.

The Solution

If there is such a thing, ‘average’ customer requirements for both performance and resilience can now be easily met within a single solution. There will of course still be some trade-off among those businesses that do require extreme performance over availability, for example those dealing with very high volume transactional databases. A flexible service provider should be able to adapt their service to accommodate the exact performance and resilience requirements of the particular customer.

So before choosing a service provider or IT partner, businesses should take a close look at whether they truly need extremely high availability and performance, because the chances are they have just become caught up in the ‘five nines’ hype and could easily find a service provider who can offer a balanced solution with availability and performance levels tailored to their individual requirements.