Beijing Institute of Genomics of the Chinese Academy of Sciences (CAS) was founded on November 28, 2003. As the predecessor of the Institute, Beijing Huada Gene Research Center participated in the international human genome monomer profiling program in the past five years and independently completed the “China Super Hybrid Paddy Genome Program” (winner of the collective award of outstanding scientific and technological achievements of the CAS), Silkworm Genome Program, Chicken Gene Polymorphism Map and other major scientific programs. It is also the first to complete the genome sequencing of SARS virus and develop its diagnostic reagent in China.
The applications of the Institute are complicated and all mainstream life science software need to be installed on this platform. This software has its distinct features. For example, Blast has high requirements on system I/O and the program consumes considerable memory resources; NAMD is a computation-intensive application and has high requirements on computing capacity. In order to satisfy the application needs of the Institute and in consideration of future expansion needs, the state-of-the-art life science platform needs to have high performances, high capacity, scalability, adaptability and good services.
Analysis of Application Features
1. Isomeric platform – high performance-price ratio
The computing network of the system is furnished with 16GB memory nodes and 32GB thick nodes that satisfy some special applications. The whole platform network is integrated with 10Gb off-band network, optical fiber network and Infiniband network. All of these are designed according to customer applications and in full consideration of customer needs and distinct application features.
2. Low power consumption to reduce continuous investments
Inspur Tiansuo TS10000's rendering nodes are furnished with high-performance processing chip platform Intel 5000P chip set. Inspur's blade system has a power conversion ratio of 93% to save power consumptions for the client to the maximum extent. When the system is not running a task, remote shutdown is enabled via the monitoring software that Inspur furnishes for the system so that power consumption can be further reduced.
3. Rocks operating system: Applicable for life science applications
Inspur adopts Rocks operating system Version 4.3 for this project. It is not only a software pack with open source codes but also a software pack that can create and maintain high-performance clusters efficiently. Its cluster management software pack and operation dispatching software pack can better manage the cluster system. Meanwhile, Rocks system is also integrated with BIO biological information processing software pack which contains considerable biological data software. It is the best choice for a biological data platform.
The system has 112 blade computing nodes. Each blade cabinet is furnished with 1 Gb through module. There were originally four BNT G8000 switches (48*1Gb RJ and 4*10Gb) and two BNT G8100 (20 CX4 10Gb and 4 SFP+ 10Gb).
120 new blade nodes have been added. Each blade cabinet is furnished with 2* 10Gb switch modules (3 external 10Gb ports). One Extreme X670 48-port 10Gb switch which is expanded to 64*10Gb SFP+ network interfaces and one Extreme X480 48-port 1Gb switch which is connected to 4 *10Gb ports upstream are also available.
The integrated system, the cluster includes 180 dual-channel blade computing nodes NX560, 1 GPU computing node NF5588 and 6 login management nodes NF5280M2. High-performance 10Gb switches are used for the trunk network and high-performance 1Gb switch for the L2 network for off-band IPMI network administration, remote shutdown/starting, restarting and other operations. KVM switch system is used to connect all node servers. Software utilized include Intel parallel environment, TSJM dispatching management, IPMI View and TSMM collective cluster monitoring software, etc. A total of 6*42U standard cabinets are provided (600 * 1050 * 2070mm cabinet).
Teacher Hu of the Customer Project Team gave the following comments on-site provide the best form of recognition for Inspur's project. He said, “Inspur fully analyzes our applications and this solution is very suitable for our applications. Indeed, it is customized for us.” Inspur's solution was recognized by the customer and hence Inspur successfully won the bid. During project implementation, Inspur's project team communicated with the client fully and formulated the implementation plan with the client's feedback. The customer had high praises for Inspur's project progress and implementation.