Sumit Kumar Bose

Grouping genetic algorithm for solving the serverconsolidation problem with conflicts

The advent of virtualization technologies encourages organizations to undertake server consolidation exercises for improving the overall server utilization and for minimizing the capacity redundancy within data-centers. Identifying complimentary workload patterns is a key to the success of server consolidation exercises and for enabling multi-tenancy within data-centers. Existing works either do not consider incompatibility constraints or performs poorly on the disjointed conflict graphs. The algorithm proposed in the current work overcomes the limitations posed by the existing solutions. The current work models the server consolidation problem as a vector packing problem with conflicts (VPC) and tries to minimize the number of servers used for hosting applications within datacenters and maximizes the packing efficiency of the servers utilized. This paper solves the problem using techniques inspired from grouping genetic algorithm (GGA) - a variant of the traditional Genetic Algorithm (GA). The algorithm is tested over varying scenarios which show encouraging results.

Incorporating fairness within Demand response programs in smart grid

Basic Demand response (DR) programs aim to modulate the demand of electricity in accordance with its supply. The existing DR programs have only been of limited success, though the participation has steadily increased in the recent past. This paper establishes the lack of fairness principles within the DR programs, as perceived by the customers to be one of the key deterrents. Fair DR (FDR) scheme criteria are defined and compared with existing pricing schemes. In this context, a simplified pricing model that takes into consideration fairness criteria for residential category is also proposed in this paper. The proposed pricing model is simulated in Gridlab-D and the results are compared with that of the flat and the price based pricing schemes. Initial results establish that our pricing scheme is fair, it flattens the demand curve over a day and provides a win-win situation for both - the customer and the utility company.

Optimizing Migration of Virtual Machines across Data-Centers

Managing virtual machines (VM) in large scale enterprise grid scenarios, commonly encountered in data centers, is extremely challenging. Currently, live VM migration is based on QoS non-conformance events; migration of a VM is initiated as soon as the aggregate resource (CPU and memory) requirements of the VMs on the physical machine (PM) exceed the capacity available on the PM. However, this paper establishes that, such ‘event-based’ migration can be extremely myopic. To overcome the limitations posed by the ‘event-based’ VM migration, this paper, proposes a new migration strategy based on ‘time-windows’. The paper outlines a mixed integer linear programming (MILP) model and a heuristic solution for generating near-optimal rearrangements of the VMs on the PMs. The paper evaluates the solutions on various VM scenarios which show encouraging results.

Allocating Resources to Parallel Query Plans in Data Grids

In this paper we present a heuristic algorithm for solving the parallel query plan scheduling problem in data grids. The algorithm exploits the binary tree structure of the query plan to determine profitable allocations. It takes care of multiple forms of parallelism while allocating resources to the sub- plans. Previous approaches towards solving the problem either ignores partitioned parallelism by focusing solely on pipelined parallelism wherein the communication cost is minimized or lays more emphasis on partitioned parallelism at the cost of pipelined parallelism. The work in this paper attempts to take care of both pipelining and partitioning mechanisms while optimizing the allocation of resources to the query sub-plans. The algorithm is particularly well suited for allocation of resources to sub-plans in a bushy query plan - which is the most profitable form of query plan in distributed database query optimization.

Semantic Based Text Classification of Patent Documents to a User-Defined Taxonomy

We present a generic approach for semantic based classification of text documents to pre-defined categories. The proposed technique is applied to the domain of patent analytics for the purpose of classifying a collection of patent documents to one or many nodes in a user-defined taxonomy. The proposed approach is a multi-step process consisting of noun extraction, word sense disambiguation, semantic relatedness computation between pair of words using WordNet and confidence score computation. The proposed algorithm resulted in good accuracy on experimental dataset and can be easily adapted and customized to other domains other the patent landscape analysis domain discussed in this paper.