Bibliography Polus : Growing Storage Related Work

Conclusion

Policy-based QoS management has been advocated as a ``silver bullet'' that can help to drastically increase the amount of storage that can be managed by system administrators. It is typically implemented using the ECA rules mechanism. However, as shown in this paper, the policy-based QoS management approach is not gaining much traction because of the associated difficulty in mapping high level QoS goals to low level system actions using the ECA approach.

In this paper we provide an alternative paradigm for mapping high level QoS goals to low level system actions. Our Polus approach leverages the proven AI techniques of learning and reasoning, and combines them with a declarative specification approach. Using this approach, system administrators can specify general rules of thumb to describe their knowledge instead of complex ECA rules containing detailed threshold values. In Polus, these threshold values are derived using learning algorithms. Furthermore, the use of reasoning engine allows Polus to automatically select the right action from amongst the various competing alternatives.

In conclusion, this paper presents a new approach towards how QoS goals can be mapped to low level system management actions. The aim of this approach is to reduce the number of inputs that are required from system administrators. We have analyzed the key concepts of this approach by using Polus to manage a simulated SAN file system and this is the first stepping stone towards the use of Polus-like approaches for managing real storage systems.

Bibliography Polus : Growing Storage Related Work