Performance of Remote Memory Paging Over the Ethernet

Next: Scaling the Input Up: Performance Results Previous: Performance Results

Performance of Remote Memory Paging Over the Ethernet

In our first experiment we evaluate two methods for paging:

REMOTE_MEMORY, which uses only main memory of other workstations as a paging device. In this experiment only one remote memory server was used, because it was enough to provide the additional amount of memory our applications needed. The measurements were done on an (almost) idle Ethernet to ensure repeatability.
DISK, which uses the local DEC RZ55 disk for paging.

The completion time of the applications is plotted in figure 2. We see that in all cases the use of REMOTE_MEMORY results in significant performance improvements. For example, for the GAUSS application, the REMOTE_MEMORY results in 112% faster execution time than DISK. Even for the MVEC application which performed very little paging, REMOTE_MEMORY results in 39% faster execution time.

The reason for the somewhat surprising performance improvements is that paging to remote memory over an Ethernet interconnection network is simply faster than paging to the disk. Even though, both the disk and the Ethernet have similar data transfer rates, REMOTE_MEMORY does not suffer from seek and rotational latency as DISK does. The average page-in/page-out service time was measured to be close to 9 ms for REMOTE_MEMORY, and close to 17 ms for DISK.

Our experimental results verify than even when the network data transfer rate is as low as the disk transfer rate, the performance of REMOTE_MEMORY is significantly higher than the performance of DISK. Since architecture trends suggest that modern high speed networks provide much higher data transfer rates than modern disks, the performance improvements of REMOTE_MEMORY over disk are bound to increase.

Evangelos Markatos
Fri Mar 24 14:41:51 EET 1995