Overview

FIT4Green aims at contributing to ICT energy reducing efforts by creating an energy-aware layer of plug-ins for data centre automation frameworks, to improve energy efficiency of existing IT solution deployment strategies so as to minimize overall power consumption, by moving computation and services around a federation of  IT data centres sites.

The FIT4Green optimization layer will not compromise compliance with Service Level Agreements (SLA) and Quality of Service (QoS) metrics, and will operate on top of current data centre management tools to orchestrate the allocation of ICT resources and turn off the under-or unused equipment.

The project will consider the deployment of whole IT solutions, from client devices to the data centres where applications and services are dynamically allocated, including the impact of the networks that provide connectivity and re-deployment capability.

 The project rationale builds on the following shortcomings of current systems:

  • Virtualization, consolidation and data centre automation techniques provide a means to achieve flexibility of IT solutions; however reduced energy consumption (achieved as a side effect of the reduction in the number of servers) is generally not considered among the Key Performance Indicators (KPIs) of the deployment options;
  • Current Service Level Agreements do not include any metrics related to environmental footprint.

FIT4Green will explore both optimal deployment solutions satisfying existing SLAs with minimal energy consumption, and next-generation energy-aware SLA statements - with minimal tradeoffs with respect to current metrics so as to provide high overall improvements for in KPIs which include environmental sustainability. 

The FIT4Green approach will be potentially applicable to any type of data centre;the project will run pilots using three representative data centres types: Service/Enterprise Portal, Grid and Clouds - provided by project partners with real test beds to validate models and policies based on the plug-ins developed in the project. Multi-cycle pilot trials with increasingly complex phases will be applied, following the spiral lifecycle model, on three federations of data centres, each one representing a different computing style: service/enterprise application portal by ENI, grid inside a supercomputing environment by JÜLICH and cloud computing lab inside HPIS.

 Motivation

Saving money in the energy budget of a data centre, without sacrificing SLAs is an excellent incentive for data centre owners, and would at the same time be a great success for environmental sustainability. As shown in the diagram below, AMAZON has evaluated its data centre expenses, showing that server costs account for 53%, while energy related costs total 42% (direct power consumption 19% plus amortized power and cooling infrastructure 23%).

Dennis Pamlin, Global Policy Advisor of WWF-Sweden, published a report titled “The potential global CO2 reductions from ICT use: Identifying and assessing the opportunities to reduce the first billion tonnes of CO2”, where he highlighted different IT solutions and the beneficial impact on GHG (green house gases, including carbon dioxide, i.e. CO2) emissions. These opportunities include, among others, a set of IT solutions: e.g. Smart Buildings, Smart Transportation and Communication, Smart Commerce and Services, Smart Industrial production; named with attribute “smart” as prefix, meaning “with low carbon footprint”. This report clearly shows a path where the adoption of these “smart” IT solutions will enable a high magnitude of GHG potential reduction.

The obvious deductions are the following:

  1. ICT itself is a large power consumer (and therefore a GHG emitter)
  2. IT solutions have a huge potential impact in reducing GHG emissions in many sectors

FIT4Green project is therefore seeking, through the federation of data centres, an aggressive strategy for reducing the energy consumption in ICT, with a potential double long term benefit for the environment:

  1. FIT4Green reduces the emissions of ICT equipment employed for today’s IT solutions
  2. FIT4Green will reduce emissions of the new “smart” IT solutions designed to reduce the environmental impact of other sectors

In the last period many ICT players have been proposing focused solutions to optimize single components (low consumption servers, CPU speed scaling, power save modes, efficient cooling devices for data centres, etc.). These solutions enable the energy footprint reduction of single devices, considering them as isolated components (holistic view), therefore lacking any savings obtainable through global optimizations. Consolidation and Virtualization techniques produce energy saving effects through the reduction of the number of active servers, but deployment strategies are fairly static and not guided by energy savings principles.

FIT4Green goes beyond this state of the art with the global analysis of IT solutions deployment needs and the optimized exploitation inside a federation of data centres with different energy related characteristics, considered as a global resource pool. By distributing computing and resources among them, FIT4Green is able to capitalize on additional degrees of freedom with potential high impacts on the optimization strategies at federation level using, for example:

  • data centres at different latitudes and the relative implications on cooling requirements based on external temperature, availability/possibility to recycle heat through co-generation devices (different latitudes in the same hemisphere – capitalizing on temperature range variations - or different hemisphere – for seasonal changes)
  • data centres in regions where different sources of power generation are available, at different costs and GHG emissions
  • data centres inside different time zones with respect to the clients, allowing a different balance/mix of computing tasks.

 Technical approach

Taking a global view on IT solutions (rather than a holistic view on single components) and applying global optimization throughout the whole ICT based process FIT4Green technical approach includes the following topics:

  • FIT4Green will introduce an optimization layer on top of existing data centre automation frameworks – integrating them as a modular “plug-in” – to guide the allocation decisions based on the optimized energy model based policies.
  • FIT4Green will develop business models for a sustainable ICT (“smart ICT”, using a title in line with other segments’ initiatives in the same direction) from the energy consumption and GHG impact perspective, including the shift towards environment aware SLAs.
  • Energy consumption models will be developed, and validated with real cases, for all ICT components in an IT solution chain, including the effects due to hosting data centres in sites with particular energy related characteristics, like alternative power availability and energy waste/recycle options, etc.
  • Optimizations, based on policy modelling descriptions able to capture the variety of deployment that are possible for a given application or a set of applications, integrated with semantic attributes supporting the evaluation of energy consumption models, will guide the deployments decisions on which/when equipments needs to stay on, where/when applications should be deployed/relocated, also capitalizing on the intrinsic non linear behaviours of energy consumption grow  with respect to the load of ICT components

Obviously, by moving applications and services to alternate data centres force the network traffic between client and servers to follow different paths, which have different impacts on the global energy consumption of the full process: both networks operated by telecommunication operators and local area networks will be considered in the global optimizations schemes.

In this count, the typical distribution of clients is very significant as well: services with a global scope, like for instance well known search engines or Web 2.0 applications, receive requests from all over the world (following some distribution patterns bound to local times); on the other hand a public administration service provider will very likely receive requests from a much more limited geographical area (and time span). The effect of the deployment of such services inside a federation of data centres implies a completely different parameterization of the model for the computation of the energy consumption: long term relocation of the public administration services to a data centre in the opposite hemisphere will force all users to have a much longer network path, while such effects do not show up so strongly if the services has an almost homogeneous distribution of clients around the globe, like in the first example. The picture changes again if it’s possible to quickly relocate services, for instance with a “follow the sun” pattern; the energy impact of the service relocation needs to be considered in the overall computation for the optimal solution.

Scenarios will be developed for several IT solutions deployed through there major categories of computing styles:

  • Service/Enterprise Portal: a set of public services and/or private enterprise applications, typically accessible through Web interface, for a large variety of user agents
  • Grid: a supercomputer system accessible through the Grid middleware
  • Cloud: a lab with cloud computing infrastructure based on open systems frameworks

There will be one pilot site for each computing style. Service/Enterprise Portal, Grid and Cloud pilots will support both single site and federated sites scenarios: multiple collaborating data centres inside the same organization for the Portal pilot; federation of supercomputer systems for Grid and open cloud federation of multiple labs for the Cloud.

Each pilot will implement the relative scenarios, measure the overall power consumption, apply the optimizations, evaluate the results and assess the expected impacts; this process will be iterated three times to allow energy models and optimizations to be refined based on the results collected from the real test beds.

The final step of FIT4Green project is to further investigate on the lessons learnt in the development of energy models, optimization policies and plug-ins development for the various computing styles tested in the pilots, and rationalize them in a set of guidelines for the development of future IT solutions, which will intrinsically consider energy consumption and environmental footprint as essential design principles.

 

Scientific and technical Objective summary

The following list summarizes FIT4Green scientific and technical objectives:

  • Research on energy aware extensions to business models and SLAs for whole IT solutions
  • Design of the system architecture for “smart ICT” solutions based on the single and federated site's real world technological and business requirements
  • Research and development of energy consumptions models of the integrated full chain of components in an IT solutions, including client, servers and network equipments
  • Research and development of energy aware deployment optimization policies, considering the differences between sites of federated data centres
  • Development of energy control mechanisms, as modular plug-in for the data centre automation frameworks of the pilot sites (but designed to be extensible to any framework), to integrate energy consumption models and optimization policies for the different computing styles both in single and federated configuration.

Validation of all the research outcomes through extensive multi-phase pilot test in real data centres representing various computing styles