Current Research

Research at Infrastructure services lab can be divided into four broad categories.

Infrastructure (IT): Traditionally, research in this area has focused on areas like server monitoring and data center activities with the goal of meeting SLAs at minimum cost. However, with the increasing cost of powering data centers (and the associated concerns of GHG emissions), monitoring server power consumptions and managing data centers with power minimization as an added goal has become a subject of intense research.

We have developed a power dashboard which gives a multidimensional view of a data center's power consumption and utilization at the individual server level at various time granularities. The dashboard also analyzes historical power data to predict short-term power consumption. We are developing methods to accurately estimate power consumption of individual servers when power cannot be measured with meters.  This technology will be deployed in production data centers to get real-life utilization and power consumption. We are also pursuing research to study and develop methods for minimizing power consumption in data centers.

Infrastructure (Facilities): Our main focus in this area is the investigation of alternative (clean) energy sources for our facilities such as wind turbines, solar power and biomass. We consider this as an internal opportunity as well as a business offer. Though there are many players and takers for green energy in the market, there is no single tool which can come up with an economically optimum energy mix as a function of energy consumption pattern, available technologies, locality specific constraints, government policies and incentives.

At the Infrastructure lab, we are designing an energy capacity planning tool which   can arrive at an optimum energy provisioning with the details of investment and ROI based on user inputs and preferences.

Further we are looking into the means of increasing the efficiency of the non-IT support equipments. The lab is aiming at an electrical modeling of the data centre which will give a dynamically varying DCIE (datacenter infrastructure efficiency).

Business Processes (carbon accounting and management): With the proliferation of GHG emission calculators, companies have a plethora of options for reporting their carbon footprints. However, a common limitation of these calculators is that they report emissions at a macro level. As a result, while companies know how much of their emissions come from sources such as electricity or transportation, they have no details on the break-up of emissions among the different business processes.

At the infrastructure lab, we are developing an emission management system based on activity-based costing, a standard methodology for allocating costs among the different units in a business. This approach allows one to slice and dice the emission data and report emissions at business process as well as activity levels and simulate the effects of different organizational changes on emission levels.

ADM (Multicore): The non-linear increase in power consumption and the diminishing gains in performance from increasing the processor frequency have led to the advent of multicore processors. A multicore processor combines two or more independent CPU cores in a single chip, with the CPU cores accessing a shared memory. This architecture scales well, while keeping the power consumption at manageable levels.

A significant proportion of applications are not multicore ready – they are single-threaded and cannot exploit the additional processing capacity made available by multicore processors.

TCS Innovation Lab – Infrastructure Services is focused on the following areas related to multicore technologies:

1. Research into compilers and frameworks for parallel processing and multicore technologies
2. Development of tools and methodologies that enable application development and maintenance (ADM) teams to transition single-threaded applications to multicore platforms