CRF Technology

We have never said much about the technology we use. Here is a short digest to answer that call. We don't cover much about our invisible technology here though (meaning software development methods).

We can say we use a wide range of open source and COTS software to develop and maintain our systems and many common programming languages including Prolog. Some of our code dates back to 1998!

What Technology does the CRF use?

• The CRF uses private enterprise class grid computing systems on-premise in the UK and to an extent in Madeira, Portugal in the EU
• Hardware is typically held for seven years and sometimes much longer if it provides value
• Primary software and operating systems are Linux based. SUSE Linux is the main O/S
• SUN Solaris was used for certain systems up to 2012
• We use VMWare COTS products extensively
• Primary enterprise hardware is Oracle and HP using AMD and Intel CPU based systems which suits our custom code. We use Intel CPUs for administrative systems. We use ARM CPUs for numerous appliances and we make use of FPGA technology and GPUs extensively too with CUDA for example
• CRF uses SAN (FC), NAS, DASD and Aberdeen Inc. storage arrays. Large amounts of storage are used by our newer cognitive software ‘Bots
• Experience shows we typically stay with providers we know

Narrow Artificial Intelligence Centres or NAICs

We used to operate Datacentres for an array of computing tasks. During 2017 we migrated to a new environment concept dedicated to supporting our Narrow AI work. After auditing our systems in 2016 we chose to ‘sunset’ some of our general purpose infrastructure after redesigning our private networks to support our new design concept of Narrow Artificial Intelligence Centres or NAICs.

Effectively a NAIC is a custom datacentre where all systems are built to an architecture that supports narrow AI operations with appropriate controls. Test systems operate in their own environments and development cycles have no external time constraints (as they are NOT commercially driven).

Core to our NAICs is privacy as with all Research and Development work risks must be contained. Security is a significant requirement for the NAICs so all the NAIC cores are not connected to the Internet at all.

High-Performance Computing (HPC)

Our HPC is applied to solving all kinds of business problems, not just scientific simulations and calculations

Cloud Computing

We use some cloud or Internet computing for low value, non-sensitive data. Only a small percentage of our work is on any Public Cloud system (such as AWS, Microsoft Azure, Google). Where an operating system tries to 'back-up' any data to a 'Cloud' we try to prevent that by default. For example for field workers using Microsoft Windows 10 we uninstall Microsoft Onedrive etc. and or take other data control measures.

NAIC Private Networks

Our private NAIC networks create a core computing environment where resources can be allocated to different applications dynamically. These resources are owned by us in our private premises.

Big Data Analytics

Our uses for big data analytics are diverse: growth algorithms, demand forecasting, custom analysers, truth engines, recommendation engines, security breach detection etc. benefit from the technology behind big data analytics. These include Apache Foundation tools including Hadoop, Spark, and scalable NoSQL tools like Cassandra. We use our own tools too using modules in various languages from C++ to Prolog.

Deep Learning

For us this is a practical application of artificial intelligence that we use applied to data analysis. Deep learning is improving our research accuracy with video processing, natural language processing, knowledge translation with many other practical applications. Again we use our many of own tools using modules developed in various languages from C++ to Prolog.

Infrastructure ICT Engineering

Commercial ICT is expensive; even the big players say so. Google and AWS both use custom designs to reduce cost and increase efficiency; Take a look at Youtube on this subject to see how these organisations manage their infrastructure today
• Microsoft seem to be one of the big players that at the core uses COTS infrastructure. Again take a look at Youtube on this subject to see how Microsoft manage their infrastructure today

Re-engineering and Unorthodox engineering

• Collaborative Research UK and EU are only small players performing advanced not-for-profit applied A.I. research on a private budget
• For some time now at Collaborative Research we use reengineered older enterprise computing equipment in our research centres and these have been retained in our NAICs
• This reengineering of older enterprise computing equipment means taking the platforms as purchased and increasing the specifications to the maximum or beyond original vendor maximums. An example is using faster modern memory, upgrading CPUs to more capable ones, upgrading BIOS, and using SSD technology. Usually we cluster these servers by type into a functional system.
• We don’t need the fastest, latest shiny form over function infrastructure
• We do need reliability, ease of maintenance and parts availability
• We do need to be able to run all things Linux and all things virtual x86
• We must have value for money and we want to be green
• So reengineering is our chosen way of achieving all of this

“It is often not the best equipped that reaches the top of the mountain” (Gordon Jeffrey 1987)

Just to be clear, we develop our NAI to enhance human ability.