What we do

TANGO is focused research:

Our objective is to provide technologies and tools made to facilitate the adoption of new heterogeneous hardware

analyzing the current market

Analyzing the current market, we have found the importance of exploiting parallelism is of increasing significance, as parallelization has become a dominant method of delivering higher performance and improved energy efficiency.

Embrace Heterogeneous Hardware

use an approach that simplifies the creation and operation of next-gen software by hiding the complexity between the distributed/parallel architecture level and the level of application/software

Optimization of Performance vs Energy

expose performance, energy and other requirements of software applications to be incorporated into the overall development and deployment process, enabling programmers to code and resource managers execute being energy-aware

Research Objectives

we plan to create a number of technologies and tools based on our research objectives:

  • Propose and implement a self-adaptive reference architecture.
  • Extend existing software development models and methodologies for heterogeneous parallel architectures.
  • Develop an energy aware hardware agnostic programming environment.
  • Develop and evaluate a self-adaptive model with identified low power parameters and QoS metrics.
  • Develop hardware power consumption and software energy models.
We are creating TANGO Toolbox

will help you control and abstract underlying heterogeneous hardware architectures, configurations and software systems including heterogeneous clusters, chips and programmable logic devices while providing tools to optimize various dimensions of software design and operations (energy efficiency, performance, data movement and location, cost, time-criticality, security, dependability on target architectures)

  • Fast Protoyping from software emulation & hardware-in-the-loop
  • Abstract & use heterogeneus hardware with Tango Programming Model
  • Spedify application critical and non-critical quality behaviors
  • Optimize energy, Performance and much more
  • Monitor underlying infrastructures, configurations & software systems
  • Define runtime Self-Adaptation behaviors
Tango Toolbox
  • A toolbox based implementation of the reference architecture
  • Reference software development models and methodologies for best practice
  • A collection of reusable IDE plugins, programming models and runtimes
  • An adaptive quality model for holistic system performance
  • Hardware and software energy models
Layer 1 – IDE & Programming Model
Requirements & Design Modelling

Develop a graphic and/or textual language to model hardware capabilities

Develop a graphic and/or textual language to identify the different implementation alternatives of an application

Identifying generic patterns of software requirements and HW/SW design Code Optimizer

Provides software developers the ability to directly understand the energy foot print of the code they write

Programming Model

Integration of COMPSs/OmpSs developed by BSC: COMPSs for higher level / orchestration; and OmpSs at finer level

Use of the COMPSs/OmpSs versioning mechanism to decide which task version fits better the applications’ goals and to interact with the different heterogeneous resources: CPUs, GPUs, FPGAs

Layer 2 - Middleware
Application Life cycle Deployment Engine

Manages the lifecycle of an application deployed by the IDE

Chooses suitable infrastructure according to energy constraints/goals and application performance constraints

Interacts with the resource and job management system e.g. SLURM

Monitoring Infrastructure

Set of APIs/Libraries to facilitate the collection of energy and performance information at runtime level

A monitoring solution that keeps and historic of the data.

Energy Modeler

Predictive energy modeling capabilities and policies depending on architecture

Layer 3 - Heterogeneous Parallel Devices
Device Supervisor

Provides scheduling capabilities across devices during application deployment and operation

Scheduling of workloads of both Clusters: Macro level, including distributed network and data management; and HPDs: Micro level, including memory hierarchy management.

Device Emulator

Provides out-of-band application deployment and operation to emulated HPD resources for the purpose of training application power profiles