Design And Implementation Of Network Memory In Linux

Applications such as multimedia, windows systems, scientific computations, engineering simulations, etc. running on workstation clusters (or networks of PCs) require an ever improving of the performance of computer systems. The performance of computer systems has improved dramatically over the last several years. CPU speeds, memory access times and network speeds has improved by a large factor. In contrast, the performance of disks has been relatively constant. Usually, the amount of memory that those applications need is more than any single workstation has available. To make matters worse, the use of multiprogramming and time-sharing further reduces the amount of physical main memory which is available to each application. To alleviate the memory shortage problem, an application could use the virtual memory paging provided by the operating system, and have some of its data in main memory and the rest on the disk. Traditionally, backing store for virtual memory paging is provided by disks, unfortunately, as the performance gap between processor and disk becomes widens, paging becomes more and more expensive. The cost of paging to a magnetic disk becomes unacceptable. There are two solutions to alleviate the situation, one is to increase the size of the physical memory, and another is to avoid using disks for paging by using the memory of idle machines on a network as backing store for those active machines. In a workstation cluster, the global computing power is most of the time underutilized. Reported measurements include 20% for average processor load during the busiest time, one third of the workstations completed unused, even at the busiest times of the day, well over a gigabyte of memory sitting idle on a typical evening. It is therefore possible most of the time to find an inactive workstation suitable for backing store for an active one. We call this strategy as "Network Memory Paging". Consequently, designing and implementing a network memory paging system has three ways, one is to design and implement it in user level, the second is to modify the kernel of operating system to implement this system,...