Related Work

There has been surprisingly little published experimental work on OS memory management design. Two works that had a great deal of influence on our work were a series of papers by Bershad [7] advocating the use of ``superpages'' to reduce TLB contention and a paper by Liedtke [3] on performance issues in the development of the L4 micro-kernel. Our initial belief was that TLB contention would be a critical area for optimization and that the monolithic nature of the Linux kernel would allow us to gain much more than was possible for L4.

It is the current trend in chip design to keep TLB size small, especially compared to the size of working sets in modern applications. There are many proposed solutions to the problem of limited TLB reach, caused by the disparity between application access patterns and TLB size, but most of them require the addition and use of special purpose hardware. Even the simpler proposed solutions require that the hardware implement superpages.

Swanson proposes a mechanism in [10] that adds another level of indirection to page translation to create non-contiguous and unaligned superpages. This scheme makes superpage use far more convenient since the memory management system does not have to be designed around it (finding large, continuous, aligned areas of unused memory is not easy).

Since existing hardware designs are set and the trend in emerging designs is towards relatively small TLBs we cannot rely on superpage type solutions or larger TLBs. More effective ways of using the TLB and greater understanding of TLB access patterns must be found. Greater understanding of access patterns and better ways of using TLB would only augment systems with superpages or larger TLBs.