Autonomic Systems and e-Business Platforms

Overview: 

The team performs high-level research in today’s eBusiness applications that have to deal with critical IT challenges in areas such as Big Data, Cloud Computing, Business Analytics, High Performance Computing and Sustainable Computing.

Objectives: 

The aim of this team is to conduct research on autonomic and intelligent resource management policies based on Self-Management strategies as the way to improve the adaptability, efficiency and productivity of current middleware layers. In these research fields the team produces top research publications as well as software components and resource management policies that can be applied at middleware level.
 
The group targets execution platforms composed of high-productivity hybrid multi-core systems and advanced storage architectures deployed in large-scale distributed environments.

Research Lines: 

The research agenda of the group for our near future contains 3 main research lines, each one it divided into 4 main topics:

  • Accelerating the processing of big data workloads, including large analytics as well as stream processing, in heterogeneous execution frameworks. During next years the main work of this research activity will be developed as part of these four main threads:

    • Exploring systems and software strategies for leveraging high-performance in-memory key/value databases to accelerate data intensive tasks, with particular attention to the IBM BlueGene Active Storage (BGAS) architecture and the Parallel In-MemoryDatabase (PIMD) as the key/value store.
       
    • Building clouds which via their programmability at multiple layers and the embracing of hardware heterogeneity can host a variety workloads and can optimize resource configuration for these workloads. The project will explore the applicability of the so-called "Software Defined Environments (SDE)" to HPC workloads as it has been previously done with transactional and data analytics workloads.
       
    • Work on the definition and implementation of data management and storage strategies for the IoT scenario that is coming. The goal is to keep track of all the object-generated data, (and provide means to retrieve and update it )and to expose an analytics layer that allows high-level services to retrieve information from the mix of historic and real-time data present in the IoT. The architecture will leverage state-of-the-art NoSQL platforms (both document oriented and graph oriented) and Stream Processing engines.
       
    • Building a hardware prototype (Minerva) as a group platform for running BigData workloads, composed of two differentiated zones. Back-end nodes offer high storage density with HDDs and a SSD caching layer, while front end nodes are SSD dense and offer a high speed interconnection network (Infiniband). The goal is to explore how to map workloads on top of the two zones as to accelerate computation while keeping the cost of the prototype low, leveraging commodity hardware on the back end and high-end components in the front end.
  • Designing resource management strategies for Big Data applications, defining policies that enables distributed data stores to meet high level performance goals, designing a non-centralized highly scalable data store management architecture, and defining interfaces with in-memory databases.During next years the main work of this research activity will be developed as part of these four main threads:

    • Propose novelty resource management strategies as query-driven data model, which focus on adapting the data model to the particular type of accesses implemented by the applications, and query-plan, which focus on automatically decide how to organize the accesses to the data store. We also aim to consider the intrinsic of continuous data streaming with real time requirements. This kind of environment also raises the challenge of defining an execution framework that is able to digest this kind of input data streams.
       
    • Tentatively extend our algorithms to non-centralized key-value multi-datacenter databases. This work will be focused on enabling a non-centralized database to be distributed across different datacenters. This will complicate the design of the management strategies because it adds a new dimension to consider that affects the performance of accessing a database.
       
    • Start to consider network usage of Big Data applications analyzing it and detecting how optical networks can benefit the performance of this application. The output of this analysis will allow us to propose policies to exploit the presence of optical networks, and to improve the automatic query plan.
       
    • Create a set of plugin modules based on our research results in order to be added to state-of-the-art open source NoSQL platforms. After this integration the comprehensive software package will be integrated in the BSC Big Data tools that the BSC department Computer Science will develop with the conjoint work of our research group and the groups of Storage and Grid. 
  • Developing management algorithms for virtualised Data Centres in a large-scale distributed ecosystem running heterogeneous workloads that optimize their operation with respect to energy and ecological efficiency. During next years the main work of this research activity will be developed as part of four main threads:

    • Enhance our algorithms to optimise both the local placement of Virtual Machines (VMs) in physical nodes and the selection of Data Centre for remote placement of VMs aiming for ecological efficiency, by exploiting the usage of green energy and the interaction with cooling systems.
       
    • Enhance our algorithms to optimise both the local placement of Virtual Machines (VMs) in physical nodes and the selection of Data Centre for remote placement of VMs, by focusing on the interaction and information exchange among Cloud layers during the whole service lifecycle for better optimization.(Collaboration with the BSC Grid research group).
       
    • Enhance our algorithms to optimise the local placement of Virtual Machines (VMs) within a physical node in a single datacenter aiming for energy efficiency by exploiting the ARM low-power architecture. (Collaboration with the BSC Architecture and the BSC Grid research groups).
       
    • Given that the BSC EMOTIVE VM management capabilities have been lacking behind the ones offered by commercial solutions such as OpenNebula and OpenStack, the future roadmap for EMOTIVE considers differentiating its VM management capabilities from its scheduling ones. VM management will be maintained basically on its current state to allow EMOTIVE to be used as a controlled middleware to support our research topics. In this sense, it could be used also as a test platform for COMPSs and Big Data applications. Regarding scheduling capabilities, there will be a work to deploy them on top of commercial middleware such as OpenNebula and OpenStack, and porting the resulting environment to the new low-power processors like 64-ARM.
Projects/Areas: 

Current involved projects:

  • Severo Ochoa Distinction (January 2012- January 2016): The Barcelona-Supercomputing Center-Centro Nacional de Supercomputación (BSC-CNS) has been accredited as Severo Ochoa Centre of Excellence, the award with which the Spanish Ministry recognizes leading research centres in Spain and international reference organisations in their respective areas. The award  will enable the execution of an ambitious research project which involves designing the hardware, software and applications to provide future solutions to the social challenges arising in health and climate change. Our main contribution to this research project consists on providing the applications with resource management strategies according to their data management requirements. Project web site: http://www.bsc.es/severo-ochoa/presentation.

 

  • COMPOSE project (Collaborative Open Market to Place Objects at your Service) (2012-2015). COMPOSE is a FP7-ICT-2011.1.2 (ref.  317862) EU Funded Project, coordinated by IBM Haifa (IL) with the following partners: CREATE-NET (IT), Fraunhofer Institute FOKUS (DE), The Open University (UK), Barcelona Supercomputing Center (ES), INNOVA S.p.A (IT), University of Passau (DE), U-Hopper (IT), GEIE ERCIM (W3C) (FR), Fundació Privada Barcelona Digital Centre Tecnològic (Bdigital) (ES), Abertis Telecom (ES), and EVRYTHNG (CH). COMPOSE  aims at enabling new services that can seamlessly integrate real and virtual worlds through the convergence of the Internet of Services (IoS) with the Internet of Things (IoT). COMPOSE will achieve this through the provisioning of an open and scalable marketplace infrastructure, in which smart objects are associated to services that can be combined, managed, and integrated in a standardised way to easily and quickly build innovative applications.

 

  • RenewIT (Advanced concepts and tools for renewable energy supply of IT Data Centres) (2013-2016): FP7-SMARTCITIES-2013 European project coordinated by IREC (ES) with the following partners: BSC (ES), AEA (IT), AIGUASOL (ES), TUC (GE), 451 (UK), DEERNS (NL). The project will develop an advanced energy simulation tool for Data Centres capable of modelling a wide range of novel energy efficiency and renewable energy strategies integrated in urban systems to improve their energy and environmental performance. Our group will contribute with algorithms to optimise both the local placement of Virtual Machines (VMs) in physical nodes and the selection of Data Centre for remote placement of VMs aiming for ecological efficiency, by exploiting the usage of green energy and the interaction with cooling systems.

 

  • BSC - IBM BGAS SoW (2013-2016) is a joint research project between researchers at Barcelona Supercomputing Center (BSC) and the  "Scalable Data Centric Computing" group at IBM Research - Watson Lab. This project aims at exploring systems and software strategies for leveraging in-memory key/value databases to accelerate data intensive tasks, with particular attention to the IBM BlueGene Active Storage (BGAS) architecture and the Parallel In-MemoryDatabase (PIMD) as the key/value store.

 

  • BSC - IBM Heterogeneous Clouds SoW (2013-2016) is a joint research project between researchers at Barcelona Supercomputing Center (BSC) and the  "Middleware and Virtualization Management" group at IBM Research - Watson Lab. This is a project focused on building clouds which via their programmability at multiple layers and the embracing of hardware heterogeneity can host a variety workloads and can optimize resource configuration for these workloads. The project will explore the applicability of the so-called "Software Defined Environments (SDE)" to HPC workloads as it has been previously done with transactional and data analytics workloads.

 

  • Lightness (Low latency and hIGH Throughput dynamic NEtwork infraStructures for high performance datacentre interconnectS) (November 2012 - October 2015) : The Lightness project is a 3-year STREP targeting Challenge 1 “Pervasive and Trustworthy Network and Service Infrastructure” of the Seven Framework Programme (FP7). The project scope lies within the objective ICT-2011.1.1: “Future Networks”. The partners involved in this project are: Interoute, Eindhoven University of Technology, Nextworks, UPC, University of California at Davis, University of Bristol, Infinera and BSC .The main objective of the LIGHTNESS project is the design, implementation and experimental evaluation of a high-performance network infrastructure for data centres, where innovative photonic switching and transmission solutions are deployed. Our group contributes to this project with the analysis of the network usage of Big Data applications and detecting how optical networks can benefit the performance of this kind of applications. Project web site: http://www.ict-lightness.eu.

 

Previous involved projects:

  • IBM SOW-Active Storage Fabrics (ASF) is a collection of components that surround a parallel in-memory database (PIMD). PIMD is a parallel client, parallel server, key/value object store. This research is part of the MareIncognito research framework between IBM and BSC.
  • OPTIMIS aims at optimizing cloud services using techniques that take advantage of an architectural framework and a development toolkit that take trust, risk, eco-efficiency, cost and legal issues into account. Our group contributes in the self-management of Cloud infrastructures using business information.
  • Barrelfish project, which is a new research operating system being built from scratch to explore how to structure an OS for future multi- and many-core systems. The design principles of Barrelfish are motivated by two closely related trends in hardware design: first, the rapidly growing number of cores, which leads to a scalability challenge, and second, the increasing diversity in computer hardware, requiring the OS to manage and exploit heterogeneous hardware resources.
  • VENUS-C is focused on developing and deploying a Cloud Computing service for research and industry communities in Europe by offering an industrial-quality service-oriented platform based on virtualization technologies. Our group contributes with tools that allow user scenarios to exploit the facilities of Cloud infrastructures.
  • NUBA project (Normalized Usage of Business-oriented Architectures) (2009-2012). NUBA is a strategic research  program (MITyC TSI-020301-2009-30) funded by the Avanza2  R&D Plan of the Spanish Ministry of Industry, Tourism and Trade and coordinated by  Telefonica I+D with 8 partners. The aim of NUBA is to advance the state-of-the-art in business models and technology for the real-time deployment  of federated Cloud platforms, integrating infrastructure from different providers, to execute elastic  business services with  the required QoS and minimizing the energy consumption. http://nuba.morfeo-project.org.

PEOPLE

PUBLICATIONS AND COMMUNICATIONS

2010

Goiri Í, Julià F, Guitart J, Torres J. Checkpoint-based Fault-tolerant Infrastructure for Virtualized Service Providers. 12th IEEE/IFIP Network Operations and Management Symposium (NOMS'10). 2010 :455-462.
Torres J, Ayguadé E, Carrera D, Guitart J, Beltran V, Becerra Y, Badia RM, Labarta J, Valero M. BSC contributions in Energy-aware Resource Management for Large Scale Distributed Systems. 1st Year Workshop of the COST Action IC0804 on Energy Efficiency in Large Scale Distributed Systems. 2010 :76-79.
Berral JL, Goiri Í, Nou R, Julià F, Guitart J, Gavaldà R, Torres J. Towards Energy-aware Scheduling in Data Centers using Machine Learning. 1st International Conference on Energy-Efficient Computing and Networking (e-Energy'10). 2010 :215-224.
Fitó O, Goiri Í, Guitart J. SLA-driven Elastic Cloud Hosting Provider. 18th Euromicro Conference on Parallel, Distributed and Network-based Processing (PDP'10). 2010 :111-118.

Pages