Distributed computing

This is a great and very old concept, in fact a lot of scientific papers uses this method of computing to get their results before the universe ends. The idea is to use a network of computers and distribute small tasks as equally as possible, relative to their processing power and resources of course. Ideally the result would be something like adding an array of additional CPUs and GPUs to your computer. But we don't live in a ideal world, this software for instance requires as width of bandwidth as you can get because what will most likely happen is that your network will choke on all the bandwidth demand and that will limit all the speed/performance benefits that are potentially there. But don't worry that's fine in fact it's great because we are using what is available to us.


In the last few years there's been a lot of talk about harnessing the power of GPUs for the purposes of general computation. When used right it can dramatically improve overall speed of computation or maybe it would be better to say, increases computation bandwidth. However because of its different design it is not as easy as saying, just throw a bit of data to the GPU and let it do its job and thus everything will run faster. Many times you have to factor in other variables such as how fast can I expect returned results and how fast can the GPU start to work on the dataset. In these terms GPUs are still less reliable relative to CPUs. But they do provide a brilliant additional option in making things snappier and faster when used right in different situations.


As of writing this (2013), most of the internet still runs on IPv4. But all modern Operating Systems should and probably do support its successor version 6 of the protocol. The fact is that current version of protocol just does not support enough addresses for billions of networking devices out there. The actual number is 2^32 (4 x 1 byte) but big parts of them are not publicly available. So look forward to the future where we have much more space for expansion (8 x 2 bytes).

Last edited May 1, 2013 at 3:15 PM by nejclesek, version 4


No comments yet.