In a typical networked business environment a user sends a computing request to a dedicated server, which processes the request and responds back to the client. As the computing needs for the user's application grows the server is replaced with newer, bigger and more expensive hardware. This is a costly and temporary fix because the application's scalability stays dependent on that of the dedicated hardware.

A better, scalable and economical solution is to use the power of parallel computing by dispatching tasks across computer nodes as opposed to CPUs on a single server. Velokri Platform software provides the software infrastructure to distribute work load across many nodes simultaneously. This reduces the computing time for the application and effectively treats nodes simply as computing resources. Therefore, in order to add more computing power to the application one only has to add more computer nodes to the pool of computers in the Computing Engine.

Whether your application is performing CPU intensive Monte Carlo simulation calculations (such as Simulated Annealing) or image transformation calculations in a multimedia application, in most cases you will be able to distribute calculations across computer nodes using our software. This enables your application to run faster, scale indefinitely and require cheaper hardware. For instance, the following figure shows the time taken for matrix multiplication operations in a single CPU dedicated server versus a Velokri Grid consisting of 4 nodes. A matrix of 800x4000 when multiplied with a 4000x800 matrix on a single node utilizing a single CPU the time on a generic (1.8 GHz Intel) hardware is roughly 90 seconds. In comparison, when the same matrices are multiplied using Velokri Platform's CE the time taken is around 26 seconds which is a little over one fourth of the single CPU time. Note, the times for Velokri Grid include time taken in dispatching the tasks and data across the network and all additional overhead.

Matrix Multiplication