lovelace, a simple compute server


A new general purpose compute server, lovelace, has been introduced for interactive use. This has similar performance and capabilities to the macomp17 node in the Maths compute cluster, with 4 CPUs, 64 processor cores and 512 GB memory, but allows you to work with it interactively in real time, rather than submitting your computations to a queue in a managed batch-oriented cluster environment.

Named after Ada Lovelace, the 19th century mathematician and arguably the first computer programmer through her work on Charles Babbage's Analytical Engine, this server fills the gap that was left when the department moved to managed cluster computing eight years ago. Prior to that date, all Linux compute systems in Maths were standalone systems in what was then known as the German Cities compute farm, with systems named after cities in Germany. Anyone with a Maths UNIX account could use these and, using X forwarding over ssh, run graphical software such as Matlab, Maple, etc. With no need to get used to working in a managed cluster environment these proved popular with visitors, users on short term collaborative projects and undergraduate & MSc students using these systems for project work.

Because there is no management of CPU and memory resource usage on lovelace, nor is there any fair usage policy, lovelace is at the mercy of users (this is the main reason we moved to managed cluster computing). So runaway jobs that use up all the memory, users running too many simultaneous computations or being unaware there may be other users on the system, etc may cause problems that will not be automatically prevented by a cluster management system. It is up to users to be aware that resources are finite and to share these with others; lovelace is monitored so that problems will be flagged up and dealt with manually, or users advised if their computations are causing excessive resource usage.

lovelace is intended for those wishing to run interactive graphical software, visitors, users who are in the department for a short period only, UG and MSc students doing projects and for those who want to test a program before submitting it to the compute cluster. It should NOT be seen as an alternative to using the compute cluster for large scale or long term computing work and usage of this system will be monitored to ensure that ad hoc computing resources are always available for as many users as possible. But frankly, due to all the advantages it offers, the Maths compute cluster should be regarded as the primary computing resource in the department.



Andy Thomas

Research Computing Manager
Department of Mathematics

last updated: 24.03.18