Hardware Infrastructures

ITB-MILANO

General

CNR-ITB bioinformatics computational resources present at ITB-Milano consist of more than 700 CPU-cores, more than 270 TB of disk space and more than 1700GBs of total memory, in an dual and quadri infiniband interconnected computational clusters. The architecture provides both an advanced HPC computational infrastructure and a distributed cloud-like virtualization facility for aggregating virtual servers, in turn providing the CNR-ITB bioinformatics services exposed to the Internet.

Interconnection

Infiniband Quadriband Cable

Cluster nodes are interconnected among themselves via a dual interface: 1Gbit ethernet + a blazing fast DDR x4 Infiniband (40 gigabit/sec). The infiniband connection allows for maximum scalability of the parallel algorithms (MPI), extremely fast disk sharing and  overhead-free access to remote memory of nodes (RDMA). The connection to the internet is in the process of being upgraded to 100mbit/sec. This will allow larger visibility of our Institution and easier remote collaboration, allowing more partners to leverage our hardware architecture.

Memory

Cluster nodes have a 24 or 16 GB memory each. This is an adequate amount for most bioinformatics algorithms known today. Experiments are being made for a virtual shared memory computational resource which can enable large single-core (nonscalable) algorithms to run on a very large amount of memory.

Cluster Nodes

 

Processors

Our computational cluster consists of more than 80 six-core Intel® Xeon® L5640 Westmere, 50 quad-core Intel Xeons E5420 Harpertown based on Penryn microarchitecture and 15 Intel® Xeon® E5420 Harpertown 4 core, for a total of more than 700 cores. These processorssupport the Intel hardware virtualization extensions VT-x and VT-d (Intel Virtualization Technology for Directed-IO). Due to the large amount of L2 cache and good virtualization support, these CPUs proved very appropriate for the algorithms and virtualization needs of the Bioinformatics Group at CNR-ITB.

Shared login

External and internal collaborators can use a convenient remote real-time desktop on a shared centralized server which proved usable all around the globe, and even mostly usable through HSDPA connection. Collaborators can also have private development servers (mostly virtual and in redundant storage).

Virtualization

Bioinformatics Workstation

 

At the Bioinformatics Group of CNR-ITB we count more than 25 virtualized servers hosting the various bioinformatics projects we have running ,exposing services to the outside world through the Internet. The linux-standard server-oriented KVM hypervisor was used as vi rtualization engine. Virtual machines are running on an experimental cloud-like architecture consisting of distributed virtualization on centralized redundant storage. This architecture provides provides redundant storage together with the possibility of migrating virtual machines on less loaded hardware nodes (which is user-initiated, for now).

 

 

ITB-BARI

ITB-BARI InfrastructuresBioinformatics server infrastructure present at ITB-Bari consists of various servers and clusters dedicated to HPC and HTC computational jobs, database servers and for web and web services deployment of bioinformatics tools (database and software) developed by the research unit.
In details, the server infrastructure contains a total of more than 100 CPU cores (one of the server is equipped with 2 GPU Tesla too), about 200Gb RAM and more than 80TB of disk space. A dedicated cluster of 5 nodes is directly connected to the Roche 454 pyrosequencer in order to provide the computing resources needed for a sequencer run.
All servers are equipped with CentOS Linux Operating system and are hosted in the IT-infrastructure room of the "CNR - Area di Ricerca di Bari". The internet connectivity is guaranteed by the "Area di Ricerca di Bari" which is directly connected to a GARR PoP node (100Mbps).
Various bioinformatics frameworks, packages and software are hosted in the server infrastructure. In addition to the general bioinformatics tools (like Emboss, Blast, NCBI toolkit, etc) and most popular bioinformatics language environment (like BioPerl, Biopython, BioJava, R/Bioconductor, etc), various installed software are focused on the management and analysis of Next Generation Sequencing NGS data (like SAMTools, Chipster, 454 Sequencing System Software, Bowtie, etc)


 

TOP