Shivali Agarwal

SmartDispatch: enabling efficient ticket dispatch in an IT service environment

In an IT service delivery environment, the speedy dispatch of a ticket to the correct resolution group is the crucial first step in the problem resolution process. The size and complexity of such environments make the dispatch decision challenging, and incorrect routing by a human dispatcher can lead to significant delays that degrade customer satisfaction, and also have adverse financial implications for both the customer and the IT vendor. In this paper, we present SmartDispatch, a learning-based tool that seeks to automate the process of ticket dispatch while maintaining high accuracy levels. SmartDispatch comes with two classification approaches - the well-known SVM method, and a discriminative term-based approach that we designed to address some of the issues in SVM classification that were empirically observed. Using a combination of these approaches, SmartDispatch is able to automate the dispatch of a ticket to the correct resolution group for a large share of the tickets, while for the rest, it is able to suggest a short list of 3-5 groups that contain the correct resolution group with a high probability. Empirical evaluation of SmartDispatch on data from 3 large service engagement projects in IBM demonstrate the efficacy and practical utility of the approach.

Static Detection of Place Locality and Elimination of Runtime Checks

Harnessing parallelism particularly for high performance computing is a demanding topic of research. Limitations and complexities of automatic parallelization have led to programming language notations wherein a user programs parallelism explicitly and partitions a global address space for harnessing parallelism. X 10 from IBM uses the notion of places to partition the global address space. The model of computation for such languages involves threads and data distributed over local and remote places. A computation is said to be place local if all the threads and data pertaining to it are at the same place. Analysis and optimizations targeting derivations of place-locality have recently gained ground with the advent of partitioned global address space (PGAS) languages like UPC and X 10, wherein efficiency of place local accesses is performance critical. In this paper, we present a novel framework for statically establishing place locality in X 10. The analysis framework is based on a static abstraction of activities (threads) incorporating places and an extension to classical escape analysis to track the abstract-activities to which an object can escape. Using this framework, we describe an algorithm that eliminates runtime checks that are inserted by the X 10 compiler to enforce place locality of data access. We also identify place locality checks that are guaranteed to fail. Our framework takes advantage of the high level abstraction of X 10 distributions to reason about place locality of array accesses in loops as well. The underlying issues, the framework and its power are illustrated through a series of examples.

Behavioral Analysis of Service Delivery Models

Enterprises and IT service providers are increasingly challenged with the goal of improving quality of service while reducing cost of delivery. Effective distribution of complex customer workloads among delivery teams served by diverse personnel under strict service agreements is a serious management challenge. Challenges become more pronounced when organizations adopt ad-hoc measures to reduce operational costs and mandate unscientific transformations. This paper simulates different delivery models in face of complex customer workload, stringent service contracts, and evolving skills, with the goal of scientifically deriving design principles of delivery organizations. Results show while Collaborative models are beneficial for highest priority work, Integrated models works best for volume-intensive work, through up-skilling the population with additional skills. In repetitive work environments where expertise can be gained, these training costs are compensated with higher throughput. This return-on-investment is highest when people have at most two skills. Decoupled models work well for simple workloads and relaxed service contracts.

Does One-Size-Fit-All Suffice for Service Delivery Clients?

The traditional mode of delivering IT services has been through customer-specific teams. A dedicated team is assigned to address all and only those requirements that are specific to the customer. However, this way of organizing service delivery leads to inefficiencies due to inability to use expertise and available resources across teams in a flexible manner. To address some of these challenges, in recent times, there has been interest in shared delivery of services, where instead of having customer specific teams working in silos, there are cross-customer teams shared resource pools that can potentially service more than one customer. However, this gives rise to the question of what is the best way of grouping the shared resources across customer? Especially, with the large variations in the technical and domain skills required to address customer requirements, what should be the service delivery model for diverse customer workloads? Should it be customer-focused? Business domain focused? Or Technology focused? This paper simulates different delivery models in face of complex customer workload, diverse customer profiles, stringent service contracts, and evolving skills, with the goal of scientifically deriving principles of decision making for a suitable delivery model. Results show that workload arrival pattern, customer work profile combinations and domain skills, all play a significant role in the choice of delivery model. Specifically, the complementary nature of work arrivals and degree of overlapping skill requirements among customers play a crucial role in the choice of models. Interestingly, the impact of skill expertise level of resources is overshadowed by these two factors.

Accelerating collaboration in task assignment using a socially enhanced resource model

Knowledge-intensive business processes require knowledge workers to collaborate on complex activities. Social network analysis is increasingly being applied in organizations to understand the underlying interaction patterns between teams and foster meaningful collaboration. The social positions of a worker, i.e. the role played in working with others, can be identified through analyzing process logs to assist effective collaboration. In this paper, we present a novel resource model that incorporates the concepts of resource communities and social positions. We demonstrate our resource model through a real industry process - IT incident management process. This socially enhanced resource model is also used to accelerate the collaboration between various work groups by dedicating collaborative units in the task of incident resolution.

Towards Auto-remediation in Services Delivery: Context-Based Classification of Noisy and Unstructured Tickets

Service interactions account for major source of revenue and employment in many modern economies, and yet the service operations management process remains extremely complex. Ticket is the fundamental management entity in this process and resolution of tickets remains largely human intensive. A large portion of these human executed resolution tasks are repetitive in nature and can be automated. Ticket description analytics can be used to automatically identify the true category of the problem. This when combined with automated remediation actions considerably reduces the human effort. We look at monitoring data in a big provider’s domain and abstract out the repeatable tasks from the noisy and unstructured human-readable text in tickets. We present a novel approach for automatic problem determination from this noisy and unstructured text. The approach uses two distinct levels of analysis, (a) correlating different data sources to obtain a richer text followed by (b) context based classification of the correlated data. We report on accuracy and efficiency of our approach using real customer data.

ReAct: A System for Recommending Actions for Rapid Resolution of IT Service Incidents

In this paper, we present a system called ReAct which, given a problem/incident description, helps the service agents to easily identify set of actions and the possible action sequence to resolve the issue mentioned in the ticket. Th eframework uses unstructured text analysis on historical ticket data to find the next best action steps and uses visualization to help user choose the most suitable option.

Automatic problem extraction and analysis from unstructured text in IT tickets

IT services are extremely human labor intensive, and a key focus is to provide efficient services at low cost. Automation of repeatable IT tasks using software service agents that reduce human effort is therefore an important component of service management. A large fraction of the work done by IT service personnel involves troubleshooting of problems. However, the complexity of IT systems makes automated problem determination and resolution a challenging research problem. Using a database of prior customer problems and solutions, we build a system that extracts knowledge about different classes of problems arising in the IT infrastructure, mine problem linkages to recent system changes, and identify the resolution activities to mitigate problems. The system, at its core, uses data mining, machine learning, and natural language parsing techniques. By using extracted knowledge, one can (i) understand the kind of problems and the root causes affecting the IT infrastructure, (ii) proactively remediate the causes so that they no longer result in problems, and (iii) estimate the scope for automation for service management. In the future, a large cost differentiator for any IT company will often involve being able to build automated service agents from these technologies, which will result in a reduction in human effort.

Distributed Scheduling of Parallel Hybrid Computations

Multicore computing is fast becoming the norm. Improving parallel programming productivity without compromising performance on multicores is a serious challenge facing research community and systems vendors. Towards this end, efficient run-time scheduling of parallel programs helps programmer by dynamically mapping tasks onto processors and scheduling them in appropriate order. Distributed scheduling of parallel computations on multiple places while ensuring low time and message complexity in bounded space is a very challenging problem. We attempt to address this challenge for hybrid parallel computations which contain tasks that have pre-specified affinity to a place and also tasks that can be mapped to any place in the system. This paper presents online distributed scheduling algorithms for hybrid parallel computations assuming both unconstrained and bounded space per place. We also present the time and message complexity for distributed scheduling of hybrid computations. To the best of our knowledge, this is the first time distributed scheduling algorithms for hybrid parallel computations have been presented and analyzed for time and message bounds under both unconstrained space and bounded space.

Continuous Learning as a Service for Conversational Virtual Agents

IT support services are moving towards self assist mode by means of cognitive agents. Such cognitive agents are typically being designed as conversational system. It is important that as the agent interacts with users, it should continuously observe, infer and learn as to what is it that it is doing well, what topics is it not able to handle well and what topics it does not seem to know about at all. In this paper, we have proposed a service that enables feedback based learning in cognitive agents. Conversation systems typically support feedback mechanism for example, some of them may ask the users to vote for the answers, or rate the experience/response that they got for their query. We propose a reinforcement learning based model for the agent to continuously learn and improve. To the best of our knowledge, this is a first attempt in modeling the continuous learning problem in conversational systems as a reinforcement learning problem. We also provide the service design for continuous learning as a service in context of conversational agents. We have evaluated the model against real data to show how the learning is helpful in improving agent’s performance. The model can also be generalized for any supervised classification problem.