Picture of computer chip

"Incredible analytics performance"

A new way to scale

Scale CPUs via big iron

This was the original hardware scaling method, whether you are running a pizza box but need a fridge, a fridge but need mainframe, there's always something bigger, and more expensive (both to buy, run and maintain). While larger boxes are back in favour, due to improvements in management and virtualisation it's still an expensive option to gain CPU power.

What happens when you want to double the CPU count? Again? 10x? 100x?

Scale CPUs via pizza boxes

Whether it be formal grids, beowolfs, clusters, MOM/JMS, roll-you-own scaling, throwing more machines are the problem will always increase the raw power you have. However, it generally won't scale linearly, will cost a fortune in hosting charges and generally won't be running anywhere near full utilisation. The darling of the web era (and still used widely) the shine has been lost from simply throwing another rack of machines at the problem (even if they are hyper-dense blades).

What happens when you want to double the CPU count? Again? 10x? 100x?

Why not just use a machine made for the job?

Scale via dedicated hardware

Using ngGrid dedicated hardware designs to make a CPU which is fully optimised for the task in hand, the performance of you application or algorithm can hit new heights. Realistic multiples of 100x - 1000x standard performance are possible for CPU intensive tasks.

Contact us today to find out how we can help.