24.04.2014 Views

D2 3 Computing e-Infrastructure cost calculations and business _models_vam1-final

D2 3 Computing e-Infrastructure cost calculations and business _models_vam1-final

D2 3 Computing e-Infrastructure cost calculations and business _models_vam1-final

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

e-­‐FISCAL: www.efiscal.eu <br />

EC Contract Number: 283449 <br />

utilisation rate of 71.3%. The utilisation rate in the e-­‐FISCAL project has been calculated by taking into account <br />

yearly logical wall clock time <strong>and</strong> available CPU cores at the end of the period (i.e. 31/12/2011). This assumes that <br />

the number of CPU cores reported at the end of the period is available throughout the year. If this is not the case <br />

(e.g. because of an infrastructure upgrade towards the end of the year), the utilisation rate calculated is <br />

underestimated. <br />

Other interesting findings are the high numbers of depreciation rates for the hardware (average 5 years), the <br />

quite good rates of PUE (of around 1.5 median value) <strong>and</strong> the percentage of electricity <strong>cost</strong> (around 15% median <br />

value of all <strong>cost</strong>s). <br />

51% of total <strong>cost</strong>s (median values) is dedicated to personnel <strong>cost</strong>s. It is expected that the personnel <strong>cost</strong>s for a <br />

very large computing centre (in the order of 100,000 cores) can show economy of scale if compared to the same <br />

capacity distributed in smaller sites that are federated together. We discuss this issue in more depth later on. <br />

However, the fact that staffing represents a very significant <strong>cost</strong> category was expected. Intersect 360 in both its <br />

reports of 2011 <strong>and</strong> 2012 (Intersect 360, 2011; 2012) mentions that the staffing represents 23% of top–level <br />

budgets. This amount in not directly comparable to e-­‐FISCAL due to the different methodology employed <br />

(Intersect 360 conducts a census on budgets <strong>and</strong> as such the total acquisition budget of hardware-­‐<strong>and</strong> not its <br />

depreciation-­‐ is included in the 100% calculation); however, the magnitude is similar. Moreover, they claim that <br />

they expect to see in the near future an increase on the money spent on people to manage <strong>and</strong> maintain systems, <br />

as the computers <strong>and</strong> overall IT architectures become more complex. <br />

Finally, our results seem to coincide with earlier findings found in literature as shown in Table 3. However, while <br />

other studies concentrate only on site or centre, e-­‐FISCAL is the first <strong>cost</strong>ing study that comes up with conclusions <br />

resulting from an extensive synthesis. <br />

Reference Cost per core <br />

Comments <br />

hour <br />

Hawtin et al. (2012) €0.075 Study for JISC UK -­‐ Differences between <br />

institutions reviewed <br />

US DoE -­‐ Magellan <br />

report (2011) <br />

€0.015 Hopper system – National Energy Research <br />

Scientific <strong>Computing</strong> Centre-­‐ including storage <br />

sub-­‐ system <br />

Smith (2011) €0.031 Purdue campus, USA <br />

University of <br />

€0.020 Hyak cluster, USA <br />

Washington <br />

Cohen <strong>and</strong> Karagiannis <br />

(2011) <br />

€ 0.09 – <br />

€ 0.14 <br />

Stratified sample of EGI centres -­‐ Assuming <br />

60% utilization ratio – storage <strong>cost</strong> included, <br />

(<strong>cost</strong>s refer to 2009) <br />

Cohen <strong>and</strong> Karagiannis <br />

(2011) <br />

€0.08 – <br />

€0.10 <br />

Stratified sample of EGI centres -­‐ Assuming <br />

60% utilization ratio – storage <strong>cost</strong> excluded <br />

(<strong>cost</strong>s refer to 2009) <br />

Table 3 – Literature review summary in relation to Cost per core hour or Cost per core hour<br />

Option 2: Basic Case Split: We also run the e-­‐FISCAL <strong>cost</strong>ing model by using as inputs the average <strong>and</strong> the median <br />

values of all necessary input data from the HPC <strong>and</strong> the HTC e-­‐FISCAL sub-­‐samples (e.g. the average <strong>and</strong> the <br />

e-­‐FISCAL : Financial Study for Sustainable <strong>Computing</strong> e-­‐<strong>Infrastructure</strong>s <br />

Deliverable <strong>D2</strong>.3 – <strong>Computing</strong> e-­‐<strong>Infrastructure</strong>s <strong>cost</strong> estimation <strong>and</strong> analysis – Pricing <strong>and</strong> <br />

Business <strong>models</strong> <br />

45

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!