Conclusions

 

In this paper we explore the use of remote main memory for paging. We describe our prototype implementation of a remote memory pager implemented on top of the DEC OSF/1 operating system as a device driver. No modifications were made to the kernel of the (monolithic) DEC OSF/1 operating system. We run several applications that use our pager on top of a DEC-Alpha-based workstation cluster to measure the performance of the system. The contributions of this paper are:

• We describe how to build a reliable remote memory paging system; we propose a novel parity-based policy that is resilient to single workstation failures.
• We show that reliable paging to remote memory results in substantial performance improvements over local disk paging.

Based on our implementation and our performance results we conclude:

• Paging to remote memory results in significant performance improvement over paging to disk. Applications that use our pager even when running on top of traditional Ethernet technology show performance improvements of up to 96% (see figure 2). Extrapolating from our results, we show that on top of a faster interconnection network even higher performance improvements are realizable!
• Paging to remote memory is an inexpensive way to let applications use more main memory than a single workstation provides. Remote memory paging provides good performance with almost no extra hardware support. The only way for magnetic disks to provide comparable performance is to use expensive disk arrays.
• Reliability in remote memory paging comes at low cost. Parity logging based paging provides reliability at low runtime and memory overhead, performs very close to NO_RELIABILITY and much faster than disk paging.
• The benefits of paging to remote memory will only increase with time. Current architecture trends suggest that the gap between processor and disk speed continues to widen. Disks are not expected to provide the bandwidth needed by paging unless a breakthrough in disk technology occurs. On the other hand, interconnection network bandwidth keeps increasing at a much higher rate than (single) disk bandwidth, thereby increasing the performance benefits of paging to remote memory.

Based on our performance measurements we believe that remote memory paging is a cost-effective and performance-effective way to execute memory-limited applications on a network of workstations.