Network
The
Institute of Genomics' network is built on 10 GB/s-compatible network equipment in order to meet the requirements of massive data transfer and exchanges between storage and the computation units. The workstations are connected in GB mode. The network is totally redundant. Two routers are positioned at the periphery of the network.
Data storage and backup
Data storage is a point to which considerable attention is paid. Storage is required to enable:
-
Volumetry:
currently, more than 2 PB of data on line;
-
Performance:
some applications require performance levels of several GB/s. That is also one of the reasons underlying upgrade to a 10-GB/s network.
The service is ensured by a NetApp controller system currently being migrated from 7-mode to the manufacturer's cluster mode. The technology will enhance the availability of the service and optimize its operation.
Various data protection strategies are implemented depending on the risk assessment and life cycle:
-
Equipment protection, redundancy (HA pair, RAID DP);
-
Short-term versioning by snapshots;
copy to a second vault for the medium term (snapvault);
-
Long-term archiving and intermediate-term backup on tapes via a StorageTek SL3000 system (LTO 6).
Computation clusters
There are several computation clusters on site running LSF software (IBM) and Slurm software. Migration towards exclusive Slurm use is ongoing.
The clusters consist of about 60 machines equivalent to about 1,300 cores. Four machines are known as 'large-memory' machines since they have more than 2 TB of memory for specific applications.
One of the requirements of the computation cluster configuration is an optimal combination of the system to ensure the most appropriate compromise between calculation rapidity and minimal latency for interactive processes.
Database
For the Institute of Genomics' own needs the systems team manages several database software packages: MySQL, Sybase, MongoDB and PostgreSQL.
Virtualization
Virtualization, which consists in implementing several (virtual) servers on a single physical machine, enables optimization of the use of the servers whose material resources would be underexploited. The technology also enables consolidation of the infrastructure by facilitating deployment and exploitation.
The virtualization infrastructure consists of 6 Dell Poweredge R620 servers (16 cores, 256 GB of memory) and 4 Dell Equalogic PS6100 bays for 100 TB of disk space.
Instrument steering
While the systems team does not intervene directly on the instruments proper, it manages the access to the machine network and, depending on the case, the configuration of the computers directly connected to it. This often calls for knowledge of equipment operation.