Cooperative caching and prefetching in parallel/distributed file systems
Toni Cortés · 1997
After the results obtained by the centralized version of our cooperative cache, we are ready to present the distributed version.This new design will solve most of the problems found in the centralized version but while still maintaining the main ideas presented in the previous chapter.Scalability is the first problem we want to solve with this distributed version.As it was already explained in the previous chapter, a centralized cooperative cache can obtain quite good performance on machines with tenths of nodes, but if larger machines are used a distributed version is needed.In the previous version, we wanted to prove that exploiting physical locality was not mandatory to achieve a high-performance cache.Once this idea has been proved, there 79