Reliable Computing

You wouldn't expect your TV to 'crash', or your fridge. You take it for granted that the electricity and water supplies are there on demand, when you want them. It's a talking point when they fail. Why can't your computer systems be the same? The answer is that they can if that's how you design them. Banks, airlines and other businesses aim for 'six nines' availability, or 99.9999%. They take out contracts with suppliers where that's written into the performance guarantees. If you do the sums that's 31.5 seconds of unscheduled downtime a year.


How do your computing systems fare compared with that? Most of the time the answer we hear is 'poorly'. Part of the problem is because people have been conditioned to expect unreliable systems and they don't ask for better; the rest of the problem is because most suppliers don't know how to build better systems either.


BUT ... by applying effort, determination and the use of well-proven techniques that situation can be reversed. Cutter is proud to play its part in that! Aiming for six nines is over-ambitious at present, due mainly to the low quality of mainstream commercial software which hasn't been written with reliability in mind. The underlying hardware systems are, fortunately, highly reliable when properly configured and managed, so the Cutter approach is to design systems which make use of well-known techniques to give the highest availability possible within a practical budget.


Because a chain is as strong as its weakest link you don't get high reliability by focusing on a single part of the system. Each element has to be selected and configured to play to its strengths. Some of the aspects of our approach are these:


  • Reduce systems complexity and management demands. Simple is good. Thin clients with centralised servers focus the weaknesses at points where they can be controlled and carefully monitored. High-quality equipment which would be too expensive to deploy everywhere can be judiciously deployed at the weak points only
  • Use systems monitoring and trend analysis. Problems can be spotted quickly and rectified before they affect large numbers of users, increasing the users' confidence and their perceptions of performance
  • Use high-reliablity software in core sections of the system; software that has been designed to feature reliability over marketing bells and whistles. Core services like email, web serving, remote accessibility, VLEs and databases should be expected never to fail.
  • Partition low-quality commercial software in ways that its failure does not cascade throughout the system. Even if it becomes unavailable the remaining systems continue to deliver. When part of the system is still working, your users perceive that as less drastic.

Cutter is motivated to do this because we don't make our living out of selling hardware or firefighting. Our support contracts are most profitable to us when the systems just work and you don't call us. We aren't stupid, this is good business for us.


Cutter Logo

High Availability

Cutter systems typically experience less than one unplanned outage per server per year. Four times out of five a simple reboot clears the problem. Our monitoring software allows you to review service levels and uptimes and indentify equipment in need of remedial attention.

User Perception

In the case of partial system failure most other services remain available so many users can still get work done. Services are partially centralised for ease of management but distributed across multiple sources of supply ensuring that total unavailability is very rare. Users perceive the system as working nearly all of the time.