Multithreaded Read & Write Patterns
The multithreaded tests simulate a situation when there are one to four clients accessing the virtual disk at the same time – the clients’ address zones do not overlap. We will discuss diagrams for a request queue of 1 as the most illustrative ones. When the queue is 2 or more requests long, the speed doesn’t depend much on the number of applications. You can also click the following links for the full results:
Theoretically, flash-memory drives should pass the multithreaded reading test with minimum performance loss because they don’t have to spend time moving any read/write heads around like HDDs do. However, there is a performance hit in practice. The second-generation X25-M is the only drive that has no problems with multiple read threads. Its speed even grows up somewhat. Its predecessor copes with the test well enough, too. OCZ’s products are much worse: they only deliver two thirds of their normal speed when reading two and three threads. The Summit, based on a Samsung controller, goes on slowing down at four threads, too, while the Indilinx-based models speed up then: two of them return to their normal speeds and the Vertex Turbo and Agility even set personal records. Perhaps this has something to do with four-channel controller architecture, but the reduction of speed at two and three threads is inexplicable.
Most of the SSDs are not good at multithreaded writing which requires efficient caching. The X25-M are the only drives to deliver stable speed. Being fast at processing one thread, the products from OCZ get slow at two and more threads. As a result, they slow down to the level of Intel’s drives or even lower.