ICT for Research

Computational Infrastructure

Computational infrastructure plays a significant, empowering roll in the research and development activities of some research groups at Process Engineering. In addition to individual personal workstations, researchers (including post-graduate students) have access to two, on-premise, enterprise-standard physical servers to run demanding computational processes.

Two Dell PE2950 and two Dell R610 servers have been virtualized in order to exploit their full potential to a maximum. These virtual platforms provide computational resources, as well as file storage, process historian, SQL databases, as well as software license services to researchers at Process Engineering. Four Windows 2008 R2 servers and one Windows 2012 R2 server run as virtual machines on these four physical hosts. All hosts deploy VMware 5.5 as virtualization layer. Researchers access these servers through concurrent Remote Desktop login sessions.

Each PE2950 presents 2 four-core XEON 5410 CPUs at 2.3GHz and up to 32GB RAM. The R610 hosts boast 2 twelve-core Xeon X5680 CPUs with hyper-threading (giving 24 logical cores per host) and 96GB RAM each. The virtual machines are easily reconfigured in response to changing computational demands, reducing capital expenditure and maintenance time.

computationalResourcesProcEng(BlankBackground - wide)

Diagram of computational infrastructure available to the Centre at Process Engineering

Professional, secure data storage and file serving are of cardinal importance. For this purpose, a Synology RS2414+ NAS server provides 19 TB network access storage to the Department of Process Engineering, with Microsoft Active Directory access control, ensuring professional levels of operational reliability and data security. The NAS server also acts as data store to the VMware hypervisor, enabling High Availability and Load Balancing features. The NAS server offers data storage and computer backup storage to the entire Process Engineering department. Secondary backup of content on the NAS server is provided by the IT department’s IBM TSM system.

Major computational software frameworks at the Department include MATLAB, Aspen Plus, Statistica, and OSIsoft’s PI System. Various toolboxes have been developed in-house for the MATLAB environment, complemented by a number of external, third-party toolboxes. Statistica is provided free of charge to the University, under sponsorship by Statsoft SA, while PI System is also free of charge, sponsored by OSIsoft, Inc.

To pre-empt large increases in computational demand that may arise occasionally, the Department has access to the University’s HPC cluster, while continuously expanding their own computational facilities. HPC virtual machines in Microsoft Azure are available to individual researchers for short term scale-out computational demand.

The Department aims to build technology partnerships with prominent infrastructure providers in order to enhance the available computational and storage facilities, such OSIsoft (PI System) and Microsoft (Azure cloud services).

For more information and technical support: Dr J P Barnard (System Engineer: ICT for Research at Process Engineering)

© 2011 Copyright Stellenbosch University | Department of Process Engineering