Monday 6th July 2009 by James Skinner

Not so many years ago, when for most companies the web was little more than a way to publish a brochure without paying for the printing, the subject of 'uptime' tended to be of interest only to the big players - online banking, eBay, Amazon, the BBC and so on. These were true 24x7 operations with money, market share and reputation at stake if their service went offline.

What's the difference, then, between 99% uptime, and 99.99% uptime?

Perfection

It's actually quite a lot, especially when you look at the cost of getting the reliability of your hosting platform up from 99% (or 'good') to 99.99% (or 'perfect'). It's the difference between 14 minutes per day of outage, and 0.1 minutes per day. That's six seconds to be precise. Only 36 minutes per year.

The reality, of course, is that if a webserver falls over, and from time to time they do, 36 minutes might not be a lot of time to get it back online again, and if one uses up that whole 36 minutes, there's a full year to wait before another glitch is acceptable.

Times have changed, however, and social media brings with it an audience who have things to say, and who want to say them 24/7. In most cases an outage at busy times, which is often outside business hours, will be met with anything between mild annoyance and openly critical rage. And the volume of complaints from your online community will rise exponentially with the amount of time the outage lasted, and what your response to it was.

Seven posts per minute

Putting this into context, a high-traffic community forum can generate huge volumes of user-generated content. Four hundred posts or more per hour is quite commonplace on the busiest of forums, for example. So on a busy site, the extra investment in seeking out that final 0.99% can be worthwhile.

At SiftGroups, we've run our own hosting infrastructure for the best part of a decade. Like many at the time, we opted for co-locating our own hardware within a data centre, not least because the software was relatively sensitive to the hardware configuration and keeping everything under our control was desirable. The move to open source software offered us the opportunity to review things, however, and there are new options available that bring that sought-after ultra-reliability into our reach.

The cloud

Cloud computing is purported to be in its infancy (we disagree, but that's a subject for another day) but there are compelling offerings starting to emerge from some big players, not least Amazon, leveraging the infrastructure and know-how that they gained since being one of the first to demand supreme reliability for their commerce platform.

The ability to 'build' a new server, from scratch, in a matter of minutes is of great interest to organisations who need to scale rapidly and perhaps provide extra capacity for short periods. Over at AccountingWeb, for example, our sister company Sift Media need to deal with significantly higher than normal volumes of traffic over the course of the Budget.

The question remains, though: can the cloud providers keep up with demand? Ultimately they have a finite set of resources, and are constantly trying to keep ahead of their clients appetite for virtual servers and SAN space.

Managed services

For organisations which want the flexibility of the hardware being someone else's problem and the SLA to be enterprise-grade whilst ensuring that they are in control of "their" kit, the managed service offering is highly attractive.

With virtualisation technology allowing multiple virtual webservers to operate on a single pair of high-availability servers, we've made the move to a managed service arrangement to enable us to offer a dedicated, scalable and resilient environment for clients who need to cater for seven posts per minute, to those who need to cater for seven a day.

The holy grail of 99.99% uptime may seem like a pipe dream at times, but as social media applications demand more and more reliability and resilience, the net cost of reducing downtime becomes lower and lower, which means that more organisations can demand the very best in reliability rather than simply aspiring to it.