Cost Estimates

Type of Processing server and disk storage server

The following two plots show the estimated evolution of server costs, based on the CERN models of CPU and Disk servers:

In summary we assume now (2017) a price/performance increase per year of 15% for CPU server and 25% for disk server.

A CPU server at CERN contains two server processors (e.g. Intel Broadwell E5-2630v4), 2 x ~1 TB SSDs (server quality), 128 GB memory (~3 GB per core, SMT on), 10 GB NIC)

  • cpu_server_cost_evolution.JPG:

A disk server uses a CPU server as a front-end and has 2 x 24-bay disks arrays attached via SAS interfaces. The disk size is currently 6 TB (S-ATA server quality) We are using 'server-mirroring' via the CERN developed EOS data management software. The shown cost includes the mirroring (i.e. twice the RAW data costs).

  • disk_server_cost_evolution.JPG:

Each generation of servers CERN is buying, needs to be adapted to various changing boundary conditions: ratio of cores to memory (was 2 GB/core and moved in 2017 to 3 GB/core); 10 Gbit network infrastructure; local disk performance due to more cores = more programs in parallel (move from HDD to SSD)

The following 3 pie charts gives an example of the detailed cost structure of processing nodes:

  • server_cost_structure_Ivy.Bridge

  • server_cost_structure_Haswell:

  • server_cost_structure_Broadwell:

Cost of electricity in the EU 2016

-- BerndPanzerSteindel - 2017-01-18

  • electricity_prices_EU_2016.JPG:
Edit | Attach | Watch | Print version | History: r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r3 - 2017-12-13 - BerndPanzerSteindel
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback