SILECS Platform

Head: Georges DA COSTA

SILECS is a major national infrastructure for experimental research on various aspects of distributed computing, from the myriad of connected objects to tomorrow data centers. This infrastructure will enable end-to-end experimentation at all levels of software and network layers, from event capture (sensors, actuators) to processing and storing data, radio transmission management and the dynamic deployment of IT services. This infrastructure will enable reproducible research on networks that can be programmed at any level.

On IRIT site, it is about the integration of the national platform Grid’5000 and the LocURa4IoT platform deployed at the IUT of Blagnac, as well as their rejuvenation. SILECS is brought to IRIT by Georges Da Costa (SEPIA) and Adrien van den Bossche (RMESS).

Grid’5000 is a large-scale and flexible testbed for experimental research in all areas of computing, with a focus on parallel and distributed computing, including Cloud, HPC and Big Data and AI. IRIT hosted a cluster of 57 servers from 2004 to 2016. Since then, IRIT hosts only one storage bay (managed in the OSIRIM platform). Since then, IRIT remains a full member of the platform and can therefore access the national platform (15000 cores in 8 sites) without limits. Several IRIT teams are currently (Sept. 2020) using this platform for their research (SEPIA, REVA, GIS, PYRAMIDE, TCI), gathering 18 active users. This platform is also used during collaborations with LAAS (4 users in 2 research teams)
LocURa4IoT (Localisation and UWB-Based Ranging Testbed for the Internet of Things) is a testbed dedicated to research in wireless networks, more specifically the performance of wireless protocols for the collection network (first link) of the Internet of Things. Its originality is the ability to study indoor location by radio time of flight. It is based on 50 nodes each composed of a microcontroller, three wireless network interfaces (UWB, BLE and LoRa) and a debugger. The whole system is administered and controlled by a server via a wired supervision network. Three environments are equipped on the Blagnac IUT campus: several offices (simple, with good repeatability properties, fundamental in wireless testbed methodologies), the Intelligent House (allowing interactions and usage tests) and an anechoic chamber (allowing total isolation from the outside world). The platform is used by 8 researchers from 2 IRIT teams (RMESS, ELIPSE), 1 researcher from LAAS and a team from the FEMTO-ST laboratory (UTBM).

The objective of SILECS at IRIT is to integrate these two platforms in order to have an end-to-end experimentation tool, from sensors to large-scale calculation/storage means.

This platform is mainly used in the framework of the Network System Architecture department, but is also integrated in the strategic action Computation, Data and AI and will be able to serve as an experimentation platform for the members of the Intelligent City DAS.

Positioning of the platform in relation to existing platforms (local and national)

Structurally SILECS is a national distributed platform. It is built on the backbone of Grid’5000 which is based on a dedicated physical subnetwork of RENATER which allows to interconnect the different computing clusters. LocURa4IoT is currently a local site, but is in the process of being integrated into FIT (equipex 2011-2020) whose goal is to interconnect different experimental sites around the IoT and wireless networks. In its current version, FIT includes several sites equipped with several tens or hundreds of sensor nodes organized in testbed, but none of them are equipped with UWB interfaces and none of them address the problem of indoor localization by radio time of flight, as is the case with LocURa4IoT. Beyond the interest for the Toulouse site to have a testbed IoT platform to be coupled with its cluster of servers, the LocURa4IoT testbed stands out for its originality compared to existing platforms at the national and European level.

This project is also complementary to more traditional computing resources such as the Toulouse CALMIP mesocenter, which provides high-performance computing resources to other scientific communities: Grid’5000 allows lower-level experiments (by modifying the operating system or the network layer for example). Similarly Grid’5000 also allows to set up algorithms and to test their scalability before deployment on mesocentres (CALMIP for example) or supercomputers (Jean Zay for example). It provides an intermediate stage between OSIRIM, which provides the means to test parallel codes on a small scale, and these large computing centers.

Technical, organizational description, utilization rate

For the moment IRIT’s Grid’5000 node consists of part of OSIRIM’s storage space (100TB). The site is managed at the Toulouse level by Georges Da Costa who coordinates the access of IRIT and LAAS participants. The system administration is ensured by the equivalent of 8 FTEs at the national level. The national platform load is 70% so it is possible to have more local participants. All IRIT members can use all Grid’5000 resources. A charter specify the usage limits of the servers in order to guarantee a good shared use while not being too restrictive. Due to the size of the platform, resources are always available and accessible without waiting (except for specific hardware such as GPU, large memory, many-cores, …). Georges Da Costa has organized several training sessions on the use of Grid’5000 and participated in 2020 in a day “Which computing platforms for which usage”[1] at IRIT (about thirty participants). Grid’5000 days and tutorials regularly take place at the national level (in conjunction with the GDR ASR COMPAS conference for example).

The LocURa4IoT platform was created in 2016 (based on a first “OpenWiNo” testbed deployed in 2014). It is co-managed by Adrien van den Bossche and Réjane Dalcé (remote access to the platform, time sharing and agenda). The maintenance of the equipment on site is ensured by the various researchers and team members present on the site of the IUT of Blagnac. The experiments can be launched during the day or at night, depending on the need for uncontrolled movements during the experiment. For the controlled movements, mobile nodes positioned on two rails allow to move the nodes during the experiment, while knowing, at any moment, the real position. The use varies according to the needs of the teams involved and the projects, as an experiment can last several hours or even several days. The physical deployment space and the range of the nodes can be adjusted so that several experiments can be run simultaneously. Currently, the platform is used at 15-20% of its capacity.

[1] Slides and tutorial available here: https://www.irit.fr/~Georges.Da-Costa/pages/dissemination.html