Search

Ability to predict the future performance of a storage system is critical for its efficient management. A modern storage system is a very complex combination of hardware and software elements and inferring its state from those of its individual components is practically impossible. Moreover, the state of the system undergoes continuous changes due to diverse and evolving workloads, frequent configuration changes and natural degradation in components. This paper presents an approach to predict the future response time of a storage system by applying a combination of machine learning techniques on the data of its internal parameters. To the best of our knowledge, this is the first attempt to quantify the response time of a storage system directly from the time-series of its internal parameters. We present a systematic experimental setup for collecting the system data and for measuring its response time while loading the system with different standard workloads. Another novel contribution of the paper is a combination of time-series forecasting model followed by a regression model to predict the response time repeatedly in real-time. The said model is shown to be able to repeatedly predict fairly accurate 15-minutes-ahead response time values using the system data of the past 20-minutes' window. It is shown that the validation calculations typically complete within ~30 seconds on an average using a computer of moderate configurations making the presented model applicable for practically reasonable prediction horizons. The design of the model is such that it can be adaptively tuned online at regular intervals as more system data is observed.

The paper proposes a new simplex-based direct search algorithm for box-constrained nonlinear programming (NLP) problems. As is typical of such methods, the optimal solution is obtained by iteratively transforming a multi-dimensional simplex and no gradient information is calculated. In contrast to unconstrained optimization problems, where the decision about way of transforming an iterate simplex is based solely on the consideration of optimality, in the proposed algorithm, both optimality and constraint satisfaction affect the iterations. First novel feature of the proposed algorithm is a fast way of checking if a planned iteration of current simplex violates any of the bounds and, if yes, a systematic way of choosing an alternative without compromising on the optimality considerations. The proposed procedure, in fact, guarantees that each iterate simplex lies entirely within the feasible region. Second novel feature of the algorithm is the use of so-called free and bounded variables that allow for different dimensions of simplices in different iterations. This feature helps convergence of the algorithm when one or more of the decision variables lie on bounds. The performance of the proposed algorithm is demonstrated on some well known test problems.

This paper is concerned with input adaptation in dynamic processes in order to guarantee feasible and optimal operation despite the presence of uncertainty. For optimal control problems having terminal constraints, two sets of directions can be distinguished in the input function space: the so-called sensitivity-seeking directions, along which a small input variation does not affect the terminal constraints, and the complementary constraint-seeking directions, along which a variation does affect the terminal constraints. Two selective input adaptation scenarios are thus possible, namely, adaptation along each set of input directions. This paper proves the important result that the cost variation due to the adaptation along the sensitivity-seeking directions is typically smaller than that due to the adaptation along the constraint-seeking directions.