FAST '04 Proceedings of the 3rd USENIX Conference on File and Storage Technologies
Automatic Model-Driven Recovery in Distributed Systems
SRDS '05 Proceedings of the 24th IEEE Symposium on Reliable Distributed Systems
ICAC '07 Proceedings of the Fourth International Conference on Autonomic Computing
| Hi-index | 0.00 |
Configuring a database is much easier than it used to be, but it's still a chore that many of us would rather not perform. The investment in hardware and people necessary to make a large, powerful database available is prohibitive for short-term tasks. As a result, applications that might use a database do without; queries take much longer than they need to; useful business analytics is not performed; and people waste time learning how to make the database fast, rather than focusing on their primary business function. Prato solves these problems by offering customers a private, virtual, DBMS appliance that can be sized up to several hundred nodes, and made available on demand, in minutes. The initial Prato prototype is built around a main-memory DBMS on which decision support queries run an order of magnitude faster than on a traditional DBMS. Future versions will make the database resilient to a wide range of failures; completely self-managing; and capable of supporting multiple back-end database types. The research problems are centered on how to automate Prato's control system in the face of a large-scale distributed system that has many failure-prone components, supports multiple users with varying degrees of sophistication and demands, and has to handle events in a completely lights-out manner. Towards this end, we employ a policy-based approach to the design and implementation of Prato's control system.