First, the most important stage of the task 5.2 has been completed.
Existing resources of laboratory had been analyzed and as a result it was taken decision, that the best solution will be purchase of new IT environment. New equipment should be coherent, but open for effective connectivity to and cooperation with existing elements.
Tasks that are provided to implement using lab resources can be defined as follows:
- processing of large data volumes,
- processing, including „massive computing”
- processing of distributed data,
- mix of above,
- testing of configurations and algorithms of above scenarios, which implies the need for quick and easy reconfiguration of the environment.
Given the profiles of the tasks that are expected to laboratory was defined the general characteristics of that environment should have:
- very large overall computational power
- several computational units, homogenous
- each computational units should have resources enough for create multiple virtual units (potentially)
- presence in the environment stream processors (CUDA technology, the most widespread)
- presence of the storage unit with the ability to share resources with other components in a high performance,
- efficient network infrastructure, for communication and data sharing
- proven, efficient virtualization environment, preferably in a recognized standard,
- unified, common management interface
Assumption, of course, after completing the necessary specific minimum values and formulations had been changed in the tender requirements.
As a result of the announced tender we purchased the set, whose structure is shown in Figure 1.
Generally, the structure can be divided into following major elements:
- IBM Flex System with enterprise chassis
- disk array systemacierz dyskową Storvize V3700
- dedicated communication rack enclosure for elements integration and carry
- VMWare vSphere 5, VMWare Center Server
Noteworthy are configurations of listed elements.
Storvize V3700 – storage array equipped with 2, independent, redundant controllers niezależne, each of 2x 10Gb/s FCoE external ports and 2 Eth interfaces for data/management. Device equipped with 24 dyski SAS, 6Gb/s, 15k rpm, dual port.
There are, additionally, 2 redundant Power Source Units. All elements are hot swap.
IBM Flex System consists of:
- chassis with two redundant Chassis Management Modules (CMM),
- Flex System Management Module ( Flex Manager Node),
- 7 modules „Flex Compute Node” of x240 class,
- 2 modules „Flex Compute Node” of x240 class additionally equipped with NVIDIA Tesla M2090 (T20A) module,
- Flex Compute Node x220 module (dedicates exclusively for hosting of virtualization menagement system),
- 2 redundant switches of „converged switch” class,
- 6 redundant Power Source Units, 2.5kW each
- 8 efficient cooling modules with their own logic.
Some details of selected modules:
- Flex Compute Node x240:
- 2x Intel Xeon 8C E5-2690 2,9GHz/1600MHz/20MB
- 256GB RAM, DDR3 1600, ECC, ChipKill
- 2x 128GB SSD HDD, raid1
- 2x 10Gb Ethernet
- service processor
- Flex System Fabric – Converged Scalable Switch (each of)
- 14 10Gb internal ports
- two external 10 GbE SFP+ ports enabled, and six Omni Ports enabled, currently configured for 3 10Gb FCoE and one 1Gb Eth,
- fabric switch – really no internal bandwidth limitations,
- long list of protocols and standards served
The whole completes the license for VMWare vSphere 5 and Center Server covering a set of all computing resources of listed Flex system.
The whole of the purchased equipment was delivered and successfully launched, pre-initialized.
The next step of the task 5.2 will be detailed configuration, installation of virtualization software and integration with other elements of the laboratory.
Figure. 1 Schematic, logical structure of system