Computing Services

CAP-IA

CAP-IA is a high-performance high-availability computing cluster, composed of:

  • 5 computer nodes with 2 Intel Xeon 4210 processors at 2.2GHz with 10 cores, 20 threads and 96Gb of RAM memory;
  • 2 management nodes with Intel Xeon 4210 2.2GHz processors, with 10 cores, 20 threads and 128Gb of RAM memory and 1 storage server; and
  • 2 Asus Geforce Turbo RTX 2080Ti GPUs and 11Gb of RAM memory.

 

Who is CAP-IA for?

IIIA's mission is to make this computing capacity available to both industry and researchers, whose projects require computing and memory capabilities beyond those provided by conventional computers. These requirements are sometimes driven by AI challenges that need to be solved in a distributed manner, and that can take advantage of the parallel computing provided by CAP-IA.  

 

What AI tools are available? 

There are open source libraries for AI tools available, such as Pytorch, TensorFlow, Scikit-learn or Keras.

Additionally, there are several AI tools developed at IIIA that can be used. Highly qualified AI researchers and technicians from IIIA are also at your disposal to help make the most of the infrastructure and available software and, when necessary, will be able to develop customized software for R&D projects, such as:

  • The development of machine learning models that require large amounts of data.
  • The development of applications that require the resolution of optimization problems.
  • The development of applications that require distributed computing, such as multi-agent systems.
  • The development of applications that require intensive testing to verify their efficiency.
  • Empirical research, taking advantage of its ability to abort and launch experiments in a more dynamic and flexible way.
  • The study of the behavior and/or performance of AI algorithms, since we have a homogeneous infrastructure of machines that allow us to ensure the same hardware conditions. 

 

Services & Pricing

Service 1. Basic use of CAP-IA

CAP-IA can be used to solve problems with large computing needs. Only the computing service is offered here, with minimal support on the use of the infrastructure. The service does not provide support on issues like the structure of the data, the software used, or the results obtained.

This service is recommended for researchers and qualified personnel with the knowledge of using such an infrastructure.

This service provides the following:

  • For each CPU used, one will have access to 8 Gb of RAM memory.
  • For each GPU node used, one will have access to 11 Gb of RAM memory.
  • One will have access to 1Tb of disk memory, though extensions may be requested and they will be attended whenever possible.

Pricing:

The cost of using 100 hours of CPU: 13.72€
Minimum price for the State's General Administration, Public Research Bodies, and universities 14.41€
Minimum price for other users 15.78€

The cost of using one GPU node is equivalent to the cost of using 10 CPUs.

Q. Do we need to provide (in all tables) the "porcentaje del coste de amortizaciones sobre el coste total"?! This info is for us or the client?

Q. Do the costs (everywhere) include IVA?!

 

Service 2. Consultancy and Advice on Custom Projects

Consultancy and advice is offered on issues such as:

  • the capacity of AI techniques to solve specific problems,
  • how to apply AI tools in organizations to optimize their processes, and 
  • the design of new AI tools for projects as they allow facing new challenges.

Pricing:

The cost of 1 hour of consultancy and advice (is the 1 hour correct here?): 70.94€
Minimum price for the State's General Administration, Public Research Bodies, and universities 74.49€
Minimum price for other users 81.58€

One may combine the consultancy and advice service with the basic use of CAP-IA service, in which case the final cost must include the costs of the individual services.  <-- Q. This statement is correct?

 

Service 3. Data-Based Decision Support Systems

This service provides the technology, experience and computing capacity for the design and development of decision support models based on data, through the use of AI techniques such as machine learning (ML), deep learning (DL), probabilistic graphical models, case-based reasoning and transfer learning, amongst others.

The computing service allows improving the efficiency of the models created through exhaustive simulation tests.

Pricing:

The cost of XXX (not clear what is accounted for here, which technologies, hours of development, how many hours of CPU?) 110.44€
Minimum price for the State's General Administration, Public Research Bodies, and universities 115.96€
Minimum price for other users 127.00€

Q. What does providing "technology" and "experience" eactly mean here? One of our off-the-shelf technologies? Adapting our existing technologies? Developing new technologies? Are we doing development here? What does the price cover exactly, the use of the cluster PLUS our "technology" and "experience"?

  

Service 4. Optimisation for Intelligent Systems

This service provides the most advanced optimization techniques, through the application of AI algorithms based on metaheuristics that allow us to tackle optimization problems on a large scale.These algorithms include genetic and evolutionary algorithms, iterated local search, simulated annealing and tabu search or the Monte Carlo Tree Search (MCTS) methods that are suitable for solving decision and planning problems. We are also specialists in constrained optimization techniques based on the SAT and Max_SAT constraint satisfaction algorithms that have proven their efficiency in solving industrial problems.

Pricing:

The cost of XXX (not clear what is accounted for here, which technologies, hours of development, how many hours of CPU?) 112.89€
Minimum price for the State's General Administration, Public Research Bodies, and universities 118.53€
Minimum price for other users 129.82€

Q. Why services 3 and 4 are limited our offer to decision support and optimisation? What if other projects pop up that have other requirements? If we are flexible there, then do we want to merge services 3 and 4 to a more general one on providing "technology", "expertise", and "computing capacity" on an array of AI applications?