Our Informatics support team provides support and architecture design for Sanger's traditional high performance compute (HPC) clusters and developing flexible compute infra-structure. Our traditional HPC environment currently provides:
- Approximately 20,000 cores of computational resource, using IBM Spectrum LSF DRM (http://www-03.ibm.com/systems/spectrum-computing/products/lsf/)
- 13.5Pb of lustre clustered file systems, (http://www.ddn.com/products/lustre-file-system-exascaler/config-specs/, http://www.intel.com/content/www/us/en/lustre/intel-solutions-for-lustre-software.html)
- 18Pb of iRODS managed storage repository (http://irods.org/consortium/). This is currently cross site replicated and provides management of all sequencing data created at Sanger since 2009.
- 160Gb/s network backbone. In total our datacenter network contains enough cabling to cover the distance from lands end to John'o'Groats !
- Our clusters are currently running the Ubuntu operating system.
- We provide support for both locally created code as well as for the for a range of centrally managed applications.
- We provide consultation supoprt and assistance for all aspects of cluster use and run a cluster course here at Sanger for new starters.
Such traditional platforms have been the bedrock of computational processing within the bio-informatics field for the past 20 years. They are well known and have performed well to date. As Sanger moves towards the next 5 years, it does so with a clear vision towards bio-informatics research and developing healthcare applications (http://www.sanger.ac.uk/about/who-we-are/sanger-institute/our-vision). Supporting cutting edge research of this nature requires that we provide platforms that provide an increasingly cost effective, flexible and scalable platform with opportunities for resilience and self service.
In June 2016, our team delivered a commercial Hybrid OpenStack and VMWare based SciaaS platform (sciaas.sandbox.sanger.ac.uk). This service provides seemless, self service access to effective, resilient, flexible and cost effective compute based services to new spinout companies from Sanger.
We will be delivering an internal private cloud Q1 2017 to our scientific researcher community. This flexible compute platform will be based around RedHat OpenStack and CLoudForms and will provide:
- 6,000 cores (SuperMicro)
- > 50TB memory (SuperMicro)
- 3PB NVMe accelerated Ceph storage (SuperMicro)
- ~1TB SSD per compute node for high speed ephemeral storage
- 100GB/s point to point software defined networking (Arista)
We will also be supplying a range of supporting applications and images to support our research community as they migrate their workloads to a more distributed compute platform.