header image

Why Is It Faster?

WireCache Doesn’t Have to Write

A relational database must make compromises between the data structures it uses in memory and those located on disk, because it must be able to transfer data between memory and rotating storage. Even modern fast disks are thousands of times slower than random access memory. Thus it makes good sense for an RDBMS to optimize its data structures for disk access.

The WireCache Transparent Database Accelerator does not have this problem. Since it never modifies the data except to replace it, WireCache can take advantage of data structures and algorithms that are highly optimized for in-memory read-only access. The performance impact of this shift is huge. Queries run in WireCache many times faster than in an RDBMS even when the latter is tuned for all data to be memory resident.

Hardware Optimizations

An RDBMS must run on a variety of hardware/OS combinations. This forces RDBMS developers to write to a hardware lowest common denominator.

In contrast, the WireCache Accelerator code runs nowhere but on the WireCache Appliance. It is highly optimized for the Appliance hardware, explicitly taking advantage of high-speed L2 caching, pre-fetching and other advanced processor features. This allows WireCache to run more efficiently than an enterprise RDBMS, even on low cost, commodity hardware.

The DA3000 is an Appliance

The DA 3000 is not a computer with interfaces and users. It only runs a single application. It does not have to be user friendly internally because all interaction with the appliance is through a simple web interface. It utilizes a stripped down light-weight Linux operating system that is dramatically faster than an off the shelf OS.

This simplification of the operating system shortens the code path and increases the speed of all operations.

take the tourschedule an analysistalk to a consultant
© 2007 WireCache. all rights reserved.