Search

We provide a formal model for the Change Management process for Enterprise IT systems, and develop change scheduling algorithms that seek to attain the "change capacity" of the system. The change management process handles critical updates in the system that often use overlapping sets of servers, resulting in scheduling conflicts between the corresponding change classes. Furthermore, applications are typically associated with certain permissible downtime windows, which impose constraints on the timing of the change executions. Scheduling of changes for such systems represent a complex dynamic optimization question. In a limiting fluid regime, where changes are assumed nonatomic, we develop a scheduling policy that provably attains the change capacity of the system. We then propose and evaluate an atomic approximation of the optimal fluid scheduling policy, which is well suited for application to a real change management system. Simulation results demonstrate that the expected change execution delay and the capacity attained by the approximate policy is close to the best attainable values, when unavoidable capacity losses due to fragmentation effects are taken into account and is significantly better than a randomized scheduling policy.

Problem determination in a large and dynamic IT service is a challenging task. In this paper we propose a framework for problem determination based on monitoring the event streams generated by the different components of an IT service. We give a generic representation of a problem through spatial-temporal patterns, which is a graph where the vertices capture the location and the time of the matching events, and the edges represent the spatio-temporal conditions between two matching events. The spatial conditions are based on the underlying system topology graph, and the temporal conditions are based on event timestamps.A practical implementation of the above framework will require fast algorithms for detecting patterns. We present efficient algorithms when the pattern graph is a range and a tree, which are then used as building blocks for a hierarchical heuristic for detecting general patterns. Finally, we show that our algorithms perform well in practice through extensive numerical simulations.

In this paper, we consider the change management process for enterprise IT services with the goal of improving the efficiency of this process, i.e., minimizing change completion time and maximizing the "change capacity". The change management process handles critical updates in the system that must be implemented by a set of executing personnel. In presence of such timing constraints and scheduling conflicts between change classes, we argue that addressing the application change management question optimally involves computing solutions to NP-hard problems. Using appropriate simulation models, we evaluate various batched scheduling alternatives to understand the effect of simple process re-engineering on the change completion time, and the change capacity of the system. In addition, we examine the benefit of executor cross-training and degree of conflicts on the performance of the system. Our results indicate that a simple longest queue based scheduling approach works well in a wide range of practical scenarios. Based on these results, we recommend process re-engineering based on longest queue based scheduling with a small degree of executor cross-training.

As more business applications have become web enabled, the web server architecture has evolved to provide performance isolation, service differentiation, and QoS guarantees. Various server mechanisms that provide QoS extensions, however, rely on external administrators to set the right parameter values for their desirable performance. Due to the complexity of handling varying workloads and bursty traffic, configuring such parameters optimally becomes a challenge. In this paper, we describe an observation-based approach for self-managing web servers that can adapt to changing workloads while maintaining the QoS requirements of different classes. In this approach, the system state is monitored continuously and parameter values of various system resources-primarily the accept queue and the CPU-are adjusted to maintain the system-wide QoS goals. We implement our techniques using the Apache web server and the Linux operating system. We first demonstrate the need to manage different resources in the system depending on the workload characteristics. We then experimentally demonstrate that our observation-based system monitors such as workload changes and adjusts the resource parameters of the accept queue and CPU schedulers in order to maintain the QoS requirements of the different classes.