High Performance Computing (HPC) at UMBI – Providing researchers with computing power for Big Data analytics
Ang Mia Yang
Zainal Azmir Ramli
Professor Datuk Dr. A Rahman A Jamal
High Performance Computing (HPC) is the use of super computers and parallel processing techniques for solving complex computational problems. HPC technology focuses on developing parallel processing algorithms and systems by incorporating both administration and parallel computational techniques (definition from Technopedia). Analysis of Big Data will require the HPC. With the Malaysian Cohort project and also the various multi-omics projects being conducted in UMBI, the development of the HPC platform at UMBI is really timely.
The application of Next Generation Sequencing (NGS) technologies produces large volumes of genomics data every year at lower cost. This ever-growing volume of big data allows more and more new functional genes and variants being discovered at a fast pace. These plethora of data often requires in-depth high throughput analysis and are often limited due to computational power and memory-intensive applications.
We would like to introduce the UMBI-HPC, a state-of-the-art High Performance Computing Cluster (HPCC), aimed to provide advanced computing support to researchers and students within UMBI, the Malaysian research community and industry. Powered by the all new Intel® Xeon Phi™ processor, UMBI-HPC is able to deliver massive parallelism and vectorization to support even the most demanding high-performance computing applications.
The setup and architecture were established based on a well matured open source technology: OpenHPC cluster manager and SLURM job scheduler, on a latest CentOS 7.3 Linux distribution. Comprising from more than 1,500 compute cores based on v4 CPU, UMBI-HPC is equipped with more than 12 terabytes of DDR4 memory and close to a Petabyte storage system. This 30 nodes workhorse and storage system are connected via 56 Gbps FDR InfiniBand low latency copper interconnects, 10 Gbe, SAS3 and impressively housed within 2 cabinet racks. Analysis tasks could be effectively resolved by segmenting the problem into several parallel tasks, and can be efficiently carried out by UMBI-HPC. We truly believed that researchers will be able to gain a cost-effective competitive advantage by using UMBI-HPC in-house expertise and computing power.
Among the wide range of scientific and analysis software available in UMBI-HPC are BLAST, FastQC, BWA, Bowtie2, GATK, HISAT, Stringtie, Velvet, QIIME, Mothur plus many other utilities and programs. Hosted in the server room at Level 7, UMBI, UMBI-HPC is highly optimized by our expert staff to ensure maximum performance.