The taste of Pesto

Abstract

The Pesto distributed storage platform is geared towards a computing model where private machines play a pivotal r ˆ ole. We argue that no centralized solutions are acceptable in its design and that it supports allocation of separate tasks to separate system components found in its target environment. Hence, Pesto separates trust from responsibility, storage from access control policy, and replication from consistency control. Pesto is designed around a few clean-cut abstractions, that make the above separations possible and efficient to implement. This report gives an overview of the main tasks typically supported by a distributed storage system, how Pesto supports these, how Pesto separates these from each other and what benefits such separation of concerns provides.