AI and big data infrastructure: CASD-IA

CASD-IA infrastructure for addressing the most demanding needs.

The hardware configuration of the platform includes physical calculators, peripheral storage devices and networks with an architecture suited for calculation power, AI and Big Data storage. Using it helps to address the increasing weight of data for hosted projects. Physically located in France, CASD-IA is widely available and can offer all the confidentiality and integrity expected by users and data owners.

The high performance server inside the bubble is only reached through an SD-Box.

The infrastructure size suits hundreds of Terabytes of raw data (aside from compression and replication) and can process several Terabytes in real time or in batches.

Hardware adapted to project needs

Software resources mainly include the Big Data middleware providing appropriate access to the data.

NoSql databases (i.e. beyond SQL: relational, colon, key value, graph, document…), models/ parallelization engines (Mapreduce, Spark), virtualisation systems, management and monitoring tools.

Resources generally set up for projects are based on Hadoop (including Cloudera and HortonWorks distributions), Hive, Impala, Sqoop, MongoDB, Neo4j, ElasticSearch, Parstream, or Redis and AI tools such as infrastructure GPUs “Up to~40k CUDA cores” adaptable and Python, TensorFlow, MXNet, Theano, Keras, Caffe, ConvNet, PyTorch, Anaconda

The deployment of this software infrastructure is also incremental to meet the expectations of “proof of concept” projects and the opportunities from new offers appearing at a very sustained pace.