Hosting and Support

Alhough our software development services are structured to maximise client independence, clients often prefer to outsource the hosting and administration of their applications. So for the applications we develop, we offer a hosting, support and administration service. This service is not aimed at clients with high revenue servers because currently, some 4 in 10 alarm calls are answered by someone on-call rather than someone already at their desk. This is dictated by our size. Instead the aim is to provide the most effective protection for the money.

For your service to be online, three things must be operational; The program that provides the service, the server it is running on, and the network that connects the server to the outside world. Keeping servers up isn't the problem as server reliability can be massively enhanced by redundancy. If one server in a cluster fails or needs to be taken offline for maintenence, another will automatically take over its workload. Individual server uptimes are often quoted as high as 99.99 percent, meaning that at any given time, there is a 1 in 10,000 chance of finding the server down. On this basis, outages will be some 53 minutes a year which we think is too optimistic. However, using a more cautious 99.9 percent uptime estimate, with only two servers in a cluster, the odds of at least one being up at any given time is an impressive 99.9998 percent. This gives odds of an outage visible to the user at any given time, of 2 in a million. With a third server the odds are 3 in a billion and with a forth, 4 in a trillion. This is the basic idea of the cloud.

Our software is not run on typical clouds, real time software generally isn't. The reason why is that clouds are implimented at the OS level and mirror everything. For performance, the mirroring is done at the application level. That way it can be limited to only new and updated data. So our 'cloud' is a cluster of standard Linux servers. One clear benefit to this approach, is that it is easier to implement a cluster over multiple data centers - thereby applying the principle of redundancy to data center networks. There will always be a performance hit to this, but this is much worse if everything is mirrored. Application specific mirroring makes it a viable option.

We rent our servers from hosting companies around the world, so your servers can be based anywhere you like. For the money, the hosting companies with their large data centers, are the best people to keep servers and networks up. We have added to the automation of the data centers and we would argue that where human intervention is necessary, it is best done by developers who benefit considerably from knowing the programs in depth.

Fees

The minimum configuration is either a shared or whole front line server, a comparable space on a server earmarked for backup, and with both servers located in the same data center. A very common configuration exactly doubles up on this with the two pairs of servers located in different data centers. The administration fees vary but as a ball park, are around the cost of the minimum server configuration. Fees do increase with larger clusters, but not in line with cluster size. As a guide, an entry level server is around $100 per month.

Minor Service Hosting and Private Linux Accounts

Many applications we write do not require a dedicated server to host or anything remotely approaching that level of capacity. Where this applies, multiple applications from various clients run on the same cluster, bringing significant cost savings. While our hosting services are mostly reserved for our bespoke software and $100 programclients, we do offer standard (non-superuser) Linux accounts more widely. Mostly these are used to develop and run server programs based on the HadronZoo library, by people who do not otherwise have an online server. However they can be for any legal purpose. Hosting for minor applications and private Linux accounts both start at $12.50 per month, with the basic package usually sufficient for most purposes.

Cyber Security

We do not disclose our cyber security techniques, so these are discussed on a case by case basis.