Cloud supercomputer service for solving demanding tasks in the oil and gas sector, CAE, CFD, bioinformatics, pharmaceuticals, AI, ML.

  • Rating:
  • (0 / 5)
Quick order

COMPOSITION OF SERVICES

  1. Lease of a set of virtual servers in a computing vysokoproizvod tion and computing infrastructure, appropriate characteristics are represented by H nym in Section 3 of this Annex №1 (hereinafter referred to as "virtual cluster") on a computational infrastructure s Competitor:

No

Component

Composition

1

Computing resources

80 servers

(2 x Xeon E5-2680v2 (20 cores), RAM 64-128GB, InfiniBandFDR)

2

Computational GPU pe LAS

5 GPU servers of the following configuration

8-16 cores (2 x Xeon E5-26xx), 128GB RAM, 8 x GTX1080Ti (11GB), 2 x 960Gb SSD

3

Storage

240 TB of disk storage

  1. Virtual Cluster Features:

2.1.      A virtual cluster composed of a plurality of virtual servers with access to the distribution of n Nome parallel shared file storage (hereinafter referred to as "Storage"), the combined high data network and providing the solution disintegrations definiteness tasks in a parallel mode, wherein the virtual server is a special way controllable totality software and hardware, both ensures, with efficient access to computing power of a given configuration (d by Lee and tech stu "node" or " virtual node "), namely one control virtual server (hereinafter referred to as" Virtual Master node ") and a set of alternating symmetrical computing virtual servers (hereinafter referred to as" Virtual Worker assembly "or the" virtual Large Worker node ").

2.2.      Virtual Cluster Configuration on Linux Platform:

●         Virtual Master node ( CentOS 7.1, RAM 2GB, 2 vCPU , 6GB local disk).

●         Virtual Worker node ( CentOS 7.1, RAM 48GB, 20 vCPU , 6GB local disk).

●         Virtual Large Worker node ( C entOS 7.1, RAM 120GB, 20 vCPU , 6GB local disk).

●         The Virtual Cluster on the Linux platform includes one Virtual Master node and from 1 to 30 Virtual Worker nodes.

●         Virtual Worker nodes are provided with physical server cores (CPU pinning ) without Hyper-threading .

●         The storage in the Virtual Cluster is mounted as a GFS2 shared file system with a total volume of up to 40TB, organized on Ceph distributed parallel storage .

●         The virtual cluster as the inter-node connections configured set small laten s t NOSTA InfiniBand FDR with wide channel 56 Gbit.

●         The Virtual Cluster network is isolated from external access by a virtual router vRouter , which performs network address translation, as well as forwarding of allowed ports (SSH) inside the Virtual Cluster. The list of ports can be extended by the dog about the denial of the Parties.

●         On the virtual cluster is set scheduler (SGE or Torque / PBS), cat ing of controls available resources:

  • Virtual Master node;Virtual Worker nodes;Virtua flax Large Worker nodes;shared file system GFS2.

●        Configured on Virtual Cluster OpenMPI for distributed computing.

●        Access to the Virtual Master node and Virtual Worker nodes or Virtual Large Worker nodes are organized by SSH protocol.

2.3.      Virtual Cluster Configuration on Windows Platform:

●         Virtual Win- node (Windows Server 2008 R2, RAM 60GB, 20 vCPU , 50GB local disk).

●         The Virtual Cluster on the Windows platform includes from 1 to 30 symmetric Virtual Win- nodes.

●         The storage organized on a distributed Ceph parallel storage in the Virtual cluster is mounted as a local disk on the first Virtual Win node, and on the rest as a network disk with a total volume of up to 40TB.

●         Inter- node connection between Virtual Win- nodes are organized via an Ethernet network with a channel width of 1Gbit.

●         The network of the Virtual Cluster is isolated from external access by a virtual router vRouter , which performs network translation of addresses, as well as forwarding of allowed ports (RDP) inside the Virtual Cluster. List of ports can be expanded by d of speaking Parties.

●         Access to Win- nodes is organized with the RDP protocol.

2.4.      Configuration of Virtual Cluster on GPU capacities:

●         Virtual or physical server (hereinafter referred to as "GPU-node"), which includes a c t e os:

  • OS : Windows 10 Pro or Centos 7.x, Ubuntu 16.04 and above ;RAM: at least 120GB;The number of CPU processing cores: at least 8 cores;Disk space: at least 200GB;Number and name of graphics cards: 8 cards Nvidia GTX1080ti;Internet access speed: 1 Gbit / s.

●         A virtual cluster consists of one or several GPU nodes.

●         The connection between the GPU-nodes is organized via an Ethernet network with a channel width of 1Gbit.

●         Access to GPU nodes is organized via remote access tools such as RDP, VNC, NICE DCV, SSH.

  1. Terms of service of the Applicant:

3.1.      Virtual cluster on Linux platform

3.1.1.    A virtual Master node in the Virtual Cluster is provided free of charge.

3.1.3.  

3.2.      Virtual cluster on the Windows platform


3.3.      Virtual Cluster on GPU Capacities3.2.2.  


3.4.      The services of the VAT Operator are not taxed in accordance with Clause 1, Article. 145.1 of the Tax Code.3.3.2.  

    No comments yet
    Please rate it

    R&D outsouring

    We specialize in research outsourcing and development outsourcing, using the best equipment and experts

    我们专注于研究外包和开发外包,使用最好的设备和专家

    نحن متخصصون في الاستعانة بمصادر خارجية للبحوث وتطوير الاستعانة بمصادر خارجية ، باستخدام أفضل المعدات والخبراء

    Nos especializamos en la investigación de outsourcing y desarrollo de outsourcing, utilizando los mejores equipos y expertos.

    Contacts

    JURIJA GAGARINA 231, NOVI BEOGRAD, SERBIA, 11070

    +381 62 854 4808

    info@rndoutsourcing.com

    Follow us
    JURIJA GAGARINA 231, NOVI BEOGRAD, SERBIA, 11070 +381 62 854 4808