CC BY-NC-ND
33
Estimated aggregate cost of ownership of a data processing center
Lyubov A. Pirogova
Project manager LEAN Lamoda Company
Address: 10, Letnikovskaya Street, Moscow, 115114, Russian Federation E-mail: lbaydalina@gmail.com
Vladimir I. Grekoul
Professor, Department of Information Systems and Digital Infrastructure Management National Research University Higher School of Economics Address: 20, Myasnitskaya Street, Moscow, 101000, Russian Federation E-mail: grekoul@hse.ru
Boris Е. Poklonov
Associate Professor, Department of Information Systems and Digital Infrastructure Management National Research University Higher School of Economics Address: 20, Myasnitskaya Street, Moscow, 101000, Russian Federation E-mail: bpoklonov@hse.ru
Abstract
Under current conditions, we see growth in demand for IT outsourcing services. This implies the activation of design and construction processes for data processing centers (DPC). Since a DPC is a complicated and expensive system, there arises the issue of justifying selection of the future project based on the estimated costs of designing and operating data processing centers.
This paper analyzes one of the possible complexes of measures to estimate costs for development and operation of data processing centers. The analysis identified main groups of capital cost in development of data processing centers which were not fully taken into account in assessments of the total volume of capital investments in previously proposed methods. The article proposes regression models to evaluate processing center construction projects based on two measures. We propose to estimate the capital cost as a function of the projected floor space of service platforms and projected number of server racks. On the basis of the models developed, analysis of the construction sites of processing data centers was conducted. This showed the model's suitability to real data. The main groups of operating costs for DPC maintenance were established, and a regression model of their evaluation was proposed. Based on the regression equation, we propose to calculate the processing center's power consumption depending on the area of the service platform or the number of server racks. The operating cost of the data processing center is determined by the power value. Analysis of information on the operating cost of various data processing centers is in fairly good agreement with the calculations obtained on the basis of the model developed.
The proposed models make it possible to evaluate with reasonable accuracy the project characteristics of development and subsequent operation of a data processing center.
Key words: data processing center, regression model, construction, representativeness of the sample, cost structure, operating costs, capital costs.
Citation: Pirogova L.A., Grekoul V.I., Poklonov B.E. (2016) Estimated aggregate cost of ownership of a data processing center. Business Informatics, no. 2 (36), pp. 32—40. DOI: 10.17323/1998-0663.2016.2.32.40.
Introduction
Under current conditions of market relations, we see expansion of the tertiary industries sector along with an increase of the outsourcing share. Services like IT outsourcing are becoming very attractive. The main producers and suppliers of IT outsourcing services are modern data processing centers (DPC), which provide a sufficiently wide range of different IT services for consumers. By using a DPC, the customer can make effective administrative decisions under conditions of limited abilities to attract financial resources for development of the company's own IT infrastructure while finally ensuring a stable and breakeven point in the company's business. Thus, it can be assumed that the demand for IT outsourcing services will be growing. Therefore, the task of developing tools for pre-estimated costs to implement such expensive and resource-intensive projects as development of data processing centers becomes a subject of great current interest.
The appearance and development of DPCs are directly linked to a multiple increase of processed and stored information volumes, the need to ensure high operational capability of mission-critical applications and business continuity processes.
Based on the implemented functions and core requirements for data processing objectives and processes, a DPC can be defined as a complex solution intended for high-performance and reliable data processing, storage and transmission having a high operational capability. The solution also includes an engineering infrastructure comprising a significant share of costs both in the course of the center's establishment and operation, i.e. in the aggregate cost of the system's ownership. On the other hand, the DPC is a combination of a large number of software and hardware platforms of various kinds — servers, data storage networks, operating systems, workload management systems and data backup built in according to specific business needs of its owner.
Based on the high level of complexity of the data processing system, it is necessary to select a set of measures for estimating costs on a reasonable basis which may occur in the processing center's development and operational phases.
Similar problems have already been solved. The solution results are presented in papers [1—3]. In the solution, the indicated problems of foreign experience were primarily considered. Let us consider the data relating to national DPC development projects.
1. Structure of cost for DPC development
Analysis of papers [4—6] enables us to identify the following four main groups of capital costs:
1. Building construction. A high-quality DPC (beginning with level Tier 3) should be located in a freestanding building with special characteristics. For this reason, the construction cost can differ from similar projects for building storage premises. However, the building can be taken on lease. In this case it should be brought into compliance with all technical requirements.
2. Grid connection. Data processing centers are distinguished by large amounts of power consumption. Therefore, they need a separate power input from the power plant. If for level Tier 1 and 2 DPC one power input line is sufficient, a Tier 3 DPC requires one active and one standby power input line, and a Tier 4 DPC needs two active lines.
3. Optical cable. It is important to note that every year the server throughput capacity is growing. In this regard, requirements for optical cables and their cost are increasing. It is assumed that there is increased demand for link capacity of communication lines (assuming an increase by a factor of about 4).
4. DPC engineering systems. The cost of backup power supply, procurement of uninterruptible power supplies, provision of the cooling system, raised floor, routing of electrical networks and purchase of equipment (racks, etc.) can be referred to this article.
Having summarized the investigation results on the cost structure [1, 7, 8], the following components of the data processing center construction costs can be identified:
1. Building construction (~ 10-15%);
2. Grid connection (~ 20-25%);
3. Optical cable (~ 0-5%);
4. DPC engineering systems (~ 60-70%).
Capital costs are generally determined by DPC surface area (associated with a number of racks) and Tier level reliability. The data provided in article [9] makes it possible to estimate the cost parameters for construction of the engineering infrastructure for a certain representative project of a data processing center (Table 1) and evaluate the dependence of these parameters on the DPC reliability level.
However, these costs do not fully reflect the total capital investment in building a DPC. If we take into account construction of additional premises required for secured assurance of center operation reliability, the cost
Cost of DPC Тable 1.
Cost of DPC construction
Tier II level Tier III level Tier IV level
Cost of 1m2 of DPC
$10579 $13 941 $25 767
Cost per rack
$26 447 $34 852 $64 417
of construction of 1 m2 (main area) increases by a factor of 2.2, and costs per rack increase by a factor of 2.4. The cost of 1 m2 of one level Tier DPC construction can vary significantly depending on the total surface area of the center. In addition, the proposed evaluations do not make it possible to extend this data to the DPC project of another configuration and do not allow us to take the
Original data sample for
construction region into account, which also has a significant impact on the cost.
Therefore, estimation procedures based on the cost detalization are untenable for project evaluation in the initial phases.
2. Regression model of capital costs
All the cost components listed above are directly or indirectly related to such characteristics of the data processing center as the surface area or number of racks. In this connection, it is reasonable to develop a model which would enable us to conduct the project assessment via these two measures.
To meet the target, data on 70 processing center construction projects was collected in Russian for the period from 2008 to 2014 (Table 2).
Тable 2.
DPC construction project
Name Year City Project cost Total area, m2 Area of service platform, m2 Number of racks, pcs. MW ' Level, Tier
Irkutsk-Energosvyaz 2014 Irkutsk 2.5 bln. Rub 10000 3200 1300 NIA 3
Government of Chelyabinsk region 2014 Chelyabinsk 27 269 000 Rub 12000 NIA 1600 16 NIA
Ministry of Health of Tula region 2013 Tula no information available (NIA) NIA NIA NIA 0.8 NIA
Gazprom Neft 2013 St. Petersburg NIA NIA NIA NIA NIA 3
Irkutsk region 2013 Irkutsk 30 bln. Rub NIA NIA NIA 30 NIA
VimpelCom 2013 Yaroslavl 4 bln. Rub 15000 3000 1200 10 3
Rostelecom 2013 Moscow 30 mln. US$ 11500 10000 NIA 40 3
Sibirtelecom 2012 Novosibirsk 70 mln. Rub 215 NIA 60 3 NIA
Inoventica 2012 Vladimir region 90 mln. Rub 300 60 0.45 3
Rostelecom 2012 Stavropol NIA 280 250 20 NIA NIA
Electronic Moscow 2012 Moscow 114.5 mln. Rub 530 250 93 1 NIA
Dataline 2012 Vladivostok NIA NIA 1000 509 NIA NIA
Transinfo 2012 Moscow NIA NIA 600 200 NIA NIA
I-Teco 2012 Krasnoyarsk NIA NIA 120 40 NIA NIA
Ikselereyt 2012 Moscow NIA NIA 580 200 NIA NIA
Stack 2012 Kazan $37 mln NIA NIA 376 2,5 NIA
Stack 2012 Moscow NIA NIA 250 30 NIA NIA
Storedata 2012 Moscow NIA 250 125 30 0,3 NIA
Rostelecom 2012 Sochi 1 bln. Rub 2000 400 92 NIA NIA
Rostelecom 2012 Kaliningrad 33.5 mln. Rub NIA 150 20 0,1 NIA
Megafone 2012 Orenburg NIA NIA 270 110 NIA NIA
Fianco 2012 Krasnoyarsk NIA NIA 370 80 NIA NIA
Fianco 2012 Ekaterinburg NIA NIA 56 12 NIA NIA
Inoventica 2012 Tatarstan NIA NIA 300 60 NIA NIA
Sberbank 2011 Moscow $1.2 bln 16500 5000 1500 25 3
DataSpace 2011 Moscow $85 mln 6000 3000 1000 4,8 3
Name Year City Project cost Total area, m2 Area of service platform, m2 Number of racks, pcs. MW ' Level, Tier
Rostelecom 2011 Vladivostok 110 mln. Rub 320 100 15 1,2 3
Yandex 2011 Moscow NIA 4500 NIA NIA 8 NIA
Linxdatacenter 2011 St. Petersburg 20 mln. Euro 7500 NIA 250 NIA 3
KROK 2011 Moscow $100 mln. 5000 2000 800 NIA 3
BSTelehouse 2011 Moscow NIA NIA 1000 75 NIA NIA
Selektel 2011 St. Petersburg NIA NIA 800 250 NIA NIA
Megafone 2011 Khabarovsk NIA NIA 390 50 NIA NIA
TEL-Hosting 2011 Moscow NIA NIA 350 60 NIA NIA
Permenergo 2011 Perm 14.6 mln. Rub 44 34 14 0,06 NIA
Bank "Neiva" 2011 Ekaterinburg 6.9 mln. Rub 34 25 4 0,04 NIA
OBIT 2011 St. Petersburg 15 mln. Rub 400 NIA 60 0,3 NIA
Bashneft 2011 Ufa 342.76 mln. Rub 400 NIA NIA 0,56 3
Oversan Mercury 2010 Moscow 400 mln. Rub 950 500 200 4 3
Oversan Luna 2010 Moscow NIA 120 50 0,5 NIA
Megaphone Samara 2010 Samara 930 mln. Rub 6912 2400 720 8 3
MDM-bank 2010 Moscow 100 mln. Rub 350 100 50 0,5
Miran 2010 St. Petersburg 80 mln. Rub NIA NIA 100 3,5 3
Storedata 2010 Moscow 60 mln. Rub NIA 250 100 1 NIA
Sibirtelecom 2009 Novosibirsk 124 mln. Rub 900 300 70 1,5 3
General DataComm 2009 St. Petersburg $5 mln 2000 500 NIA NIA NIA
Komkor (Acad Telecom) 2009 Moscow 400 mln. Rub NIA NIA 140 NIA NIA
Dataline 2009 MR - Korovin high road NCA + 217.5 mln. Rub NIA 2700 800 7 2
Dataline 2009 Moscow -Borovaya NCA + 122 mln. Rub 1855 900 360 4 2
IT-park 2009 Kazan 3500 1000 294 5 3
Megafone (Synterra) 2009 Kazan 100 mln. Rub 229,5 170 48 0,5 NIA
PiN Telecom 2009 St. Petersburg 18 mln. Rub 200 NIA 38 NIA NIA
ISG 2009 140 mln. Rub 700 NIA 150 3 NIA
Trastinfo 2009 Moscow NCA + 176 mln. Rub 3000 1600 800 6,4 NIA
0KB Progress 2009 Moscow NCA + 4 bln. Rub 480 480 100 NIA NIA
Infobox 2009 NCA + 7.9 mln. Rub NIA 600 NIA 25 NIA
Selektel 2009 Moscow 4.5-5 mln. US$ 500 300 80 NIA 2
Uralsvyazinform 2009 Ekaterinburg 300 mln. Rub NIA 432 250 NIA NIA
Dataplanet 2009 Zelenograd NCA + 9.8 mln. Rub 170 160 40 0,3 NIA
Raduga-2 2009 St. Petersburg NCA + 2.2 mln. Rub NIA 60 20 NIA NIA
Rostelecom 2008 Ekaterinburg 10 mln. Rub 155 100 36 NIA NIA
Troika Dialog 2008 Moscow $10 mln 200 NIA NIA 0,5 NIA
Peter-Service 2008 $20 mln 480 480 50 0,3 3
0BIT 2008 St. Petersburg NCA + 10.1 mln. Rub 400 300 120 NIA 3
Selektel 2008 St. Petersburg NCA +69.5 mln. Rub 1500 700 200 2 2
YUTK 2008 320 mln. Rub 1000 300 NIA 1,5 NIA
M1, Stack 2007 Moscow $15 mln 2500 NIA NIA NIA NIA
Tehnogorod 2007 Moscow NCA + 10 mln. Rub 1500 NIA NIA 1 3
Karavan 2008 Moscow $7 mln 1000 NIA NIA 2 NIA
Ixcellerate 2008 NIA 15000 6200 NIA NIA NIA
Zelenograd 2008 Zelenograd 3 bln. Rub 16000 14000 1215 21 3
Capital costs (CAPEX) across all projects in Russia are determined by the following correlations:
1. CAPEX = -2856583 + 22136-S (in which case R2 = 0.72; P-value for the coefficient of variable S is 4.7E-11. P-value of free constant is 0.76). Low P-value for the coefficient at variable S makes it possible to predictably say that the construction cost of one square meter (with a root-mean-square error of 2 339 dollars) is in the range between 19 797 and 24 475 dollars. This agrees with the expert assessments of 15-25,000 dollars [10].
2. CAPEX = -3375063 + 78751- N (in which case R2 = 0.8; P-value for the coefficient of variable N is 2.03E-13, P-value of free constant is 0.67). With a root-mean-square error of the coefficient at N=, the construction cost in terms of a rack is in the range from 71 994 to 85 508 dollars. Therefore, the obtained construction cost of one rack is approximately 3.5 times higher than the construction cost of one square meter of DPC. This more or less equals the correlation obtained from marketing research.
To refine the cost, separate regression models depending on DPC location can be constructed:
♦ Moscow: CAPEX_Moscow = -2651754 + 22612- S or CAPEX_Moscow = -3315038 + 73616- N;
♦ Regions: CAPEX_Regions = -8077885 + 26586-S or CAPEX_Regions = -6171183 + 95935- N.
Summarizing the construction of regressional relationships and comparing the calculation results by a model with real data, one may conclude that the model rather suitably describes real data. Deviations of the estimated data from the averaged data for all selection of values are shown in Table 4.
Тable 4.
Geographical segmentation DPC construction cost
DPC location Moscow Regions Russia
Average cost of building 1 sq.m (aggregate CAPEX of sample / aggregate S) 19 686 22 890 22 291
Average cost of building 1 rack (aggregate CAPEX of sample / aggregate N) 62 080 85 400 80 407
Unit cost of building 1 sq.m (regression) 22 612 26 586 22 136
Unit cost of building 1 rack (regression) 73 616 95 935 78 751
Confidence interval - cost of building 1 sq.m (regression) 17388 - 27836 24 153 - 29 019 19 797 - 24 475
Confidence interval - cost of building 1 rack (regression) 59 201 - 88 031 93 366 - 98 502 71 994 - 85 508
Average surface area of DPC 1 583 1041 1258
Average number of DPC racks 509 242 349
Deviation of construction cost of 1 sq.m 13% 14% -1%
Deviation of construction cost of 1 rack 16% 11% -2%
In some projects, the abbreviation NCA is seen in the column "project cost". This means that the project cost has not been revealed by the company, but when analyzing the data from the SPARK-Interfax system, an increase in value of the noncurrent assets (NCA) to include the specified amount can be found when constructing the data processing centers.
After collecting data on DPC construction projects, a procedure of adjusting them to a single currency (in our case the US dollar was selected) and prices of one year (2013 was selected) was carried out. This has been done using the price index for engines and equipment used in construction.
Unfortunately, in some cases data on projects was incomplete: for example, with the known cost of the construction and area of engineering sites, the number of racks was unknown. In such cases correlations identified in market research [8] and presented in Table 3were used for data recovery.
Table 3.
DPC market dynamics in 2011-2016
2011 2012 2013 2014 2015 2016
Racks, '000 units 15.9 18.7 23.1 28.7 34.5 42.2
Area, '000 sq. m. 52.8 62.6 84.6 103.7 121.8 146.8
Sq. m. / rack 3.3 3.3 3.7 3.6 3.5 3.5
After processing, the original sample regression models were constructed with breakdown across DPC construction projects in Moscow and in the regions.
A planned DPC surface area (S) and planned number of racks (N) were selected as independent variables.
A very interesting pattern can be derived from this table: the cost of DPC construction per 1 sq.m (or 1 rack) in Moscow is lower than in the regions by approximately 20%. This can be explained by the fact that the dimensions of a statistically average Moscow DPC exceed the dimensions of regional DPC by 80%, and with the project scaling-up the unit cost significantly goes down for each new rack or square meter.
These correlations can be used in evaluating the cost in the initial construction phase, and in the cost estimation for further development of the DPC, if it is foreseen.
Due to insufficient representativeness of the sample, it turned out to be impossible to include such parameters as Tier level and build time for the center in the regression model.
The impact of Tier level on the price of 1 sq.m can be taken into account by multiplying the CAPEX value on correction factor Kt, the values for which were obtained based on Table 1:
Kt = 0.8 for level Tier 2; Kt = 1 for level Tier 3; Kt = 1.8 for level Tier 4.
To take into account the price dynamics over time, the research results presented in paper [11] can be used. These show that the construction cost of 1 sq.m increases by approximately 30% per year.
Thus, the DPC construction cost in year G is determined by the ratio:
CAPEXg = CAPEXKt1.3(G-2013).
3. Breakdown of DPC maintenance costs
Operating costs for DPC maintenance can be divided into five main groups:
1. Payment for power consumption. In calculating this parameter, one should rely not only on the value of kWh consumed by racks, but take into account the power consumption structure.
2. Rent of premises. This parameter depends heavily on the geographical location of the DPC and vary with time.
3. The payroll budget can depend heavily on the processing centers, irrespective of the level of reliability, on the basis of the operation continuity requirements.
4. Maintenance. The maintenance cost is determined by the composition of the systems used.
5. Other costs: appreciation of equipment, processing center insurance, etc.
Considering the research results of various companies, significant differences in the structure of operating costs
of Russian and foreign processing centers should be noted. In particular, the Russian market is characterized by distribution of costs [1, 12, 13] presented in Table 5.
Table 5.
Structure of cost of Russian DPC composition
Krok Datadom CNews Radius Group
Payment for electrical power 42% 25% 25% 42%
Rent of premises 9% 24% 20% 16%
Payroll budget 36% 40% 40% 35%
Maintenance 5% 11% - 7%
Other costs 8% - 15% 7%
American companies use a different structure of operating costs [9]. The difference is due to a different approach to clustering of costs subgroups among all operating costs, as well as the specifics of the Russian economy, in particular the wage gap, power cost and so on. Nevertheless, in all research the DPC maintenance cost includes expenses involved in electrical power (average 30%) and rent of premises. Typically, DPC maintenance cost also includes personnel costs and maintenance costs. Further articles of expense items for the processing center operation can differ widely.
The operating costs can be derived from the capital costs at the expense of such a key indicator as the DPC power, and subsequently based on the DPC power and power consumption costs.
From analysis of the power consumption structure in various centers [9, 12], it is apparent that the IT equipment used consumes about half of the power used by the data processing center. This means that all DPC racks consume power, and their cost is approximately 15% of all operating costs. At the moment, most DPC suse 5 kW racks for 42U, but the cost of 1 kWh of power in different regions differs widely.
Thus, let us assume that the operating costs are divided into five groups, each of which contributes to the overall cost:
1. Payment for power consumption (~ 30-35%).
2. Rent of premises (~ 15-20%).
3. Payroll budget (~ 25-30%).
4. Maintenance (~ 10-15%).
5. Other costs: appreciation of equipment, insurance, etc. (~ 10-15%).
To estimate the value of the operating costs, you can use a regression model. Before we start developing it, we have to highlight the main principles used to estimate
these costs. Let us consider DPC power as a central variable for calculation, because:
♦ it can be quite accurately determined from the initially claimed technical characteristics (in particular, number of racks and DPC surface area)
♦ costs associated with the payment for electric power are the most notable group of the operating costs.
Thus, let us introduce new variables to generate regression enabling us to estimate the structure of the operating costs:
OPEX — operating costs within a year, dollar
M - DPC power, mW
e — electrical power cost, doll./kW/h (different for each region of Russia).
Let us consider the relationship between the power of the processing center and its characteristics, having constructed appropriate regressions:
• M = —0.17797 + 0.01192- N (R2 = 0.93, P-value for the coefficient at N is 6.61E-16);
• M = 0.24135 + 0.002671-S (R2 = 0.66, P-value for the coefficient at S is 2.28E-07).
It is interesting that the specific power of each additional rack in DPC became 11.9 kW. Considering the fact that in the power consumption structure in-house equipment uses about 50% of all power, a generic rack in
42U has a power of 5 kW. This confirms the suitability of the data obtained.
As is clear from the regressions obtained, it is better to use the number of racks to assess the power. In assessing capital costs, it was proposed to build separate regressions for Moscow, the regions and Russia as a whole. In this case, it is inexpedient because such a key index (except for DPC power), as the cost of electrical power is significantly different for each Russian region and should be chosen separately.
In summary, proceeding from the previous analysis of the operating cost structure (about 30% OPEX is accounted for by electric power), the assessment can be conducted by the following formula: OPEX = M-[365 days]- [24 hours]- e/0.3 or OPEX = (—0.17797 + 0.01192- N)-29200- e/0,3 OPEX = (0.24135 + 0.002671- S) •29200• e/0.3.
Conclusion
Obviously, all data processing centers differ from each other. Thus, there is no multipurpose tool which could exactly calculate money flows. The proposed procedure makes it possible with appropriate accuracy to estimate the characteristics of data processing center development projects. ■
References
1. Radius Group (2012) Ekonomika TsOD [Economics of data centers]. Available at: http://www.rvip.ru/1065/document1071.shtml#5 (accessed 1 June 2016) (in Russian).
2. Patel C.D., Shah A.J. (2005) Cost model for planning, development and operation of a data center. Internet Systems and Storage Laboratory — HP Laboratories Palo Alto. Available at: http://www.hpl.hp.com/techreports/2005/HPL-2005-107R1.pdf (accessed 1 June 2016).
3. Hamilton J. (2010) Overall data center costs. Perspectives. James Hamilton's Blog. Available at: http://perspectives.mvdirona.com/2010/09/18/ OverallDataCenterCosts.aspx (accessed 1 June 2016).
4. Lebedev P. Kommercheskie TsOD vRossii: smena biznes-modeli [Commercial data centers in Russia: Shift of business model]. CNews. Available at: http://www.cnews.ru/reviews/free/cloud/articles/articles13.shtml (accessed 1 June 2016) (in Russian).
5. Comprice (2012) Obzor rynka datatsentrov Sankt-Peterburga v I polugodii 2012 g. [The review of the market data center of St. Petersburg in the first half of 2012]. Available at: http://www.comprice.ru/articles/detail.php?ID=475362&print=true (accessed 1 June 2016) (in Russian).
6. TelecomBloger (2013) Osobennostipostroeniya nebol'shogo datatsentra v Moskve [Features build a small datacenter in Moscow]. Available at: http://telecombloger.ru/14535 (accessed 1 June 2016) (in Russian).
7. Sukhov R. (2008) Kak rasschitat' zatraty na TsOD? [How to calculate the cost of the data center?]. Zhurnal setevykh resheniy LAN (electronic journal), no. 04. Available at: http://www.osp.ru/nets/2008/04/4944787/ (accessed 1 June 2016) (in Russian).
8. Pavlov A., Basistyy D., Kusakin D. (2012) Biznes-plan i finansovaya otsenka sozdaniya TsODa vysokogo urovnya nadezhnosti [Business plan and financial evaluation of establishing a data center a high level of reliability]. iKS-Media. Available at: http://www.iks-media.ru/is-sue/2012/7/4430815.html (accessed 1 June 2016) (in Russian).
9. AboutDC (2011) Statistika TsOD [Statistics of data centers]. Available at: http://aboutdc.ru/page/64.php AboutDC.ru (accessed 1 June 2016) (in Russian).
10. IT Weekly (2013) Data-tsentry rastutkakgriby... [Data centers are growing like mushrooms...]. Available at: http://www.it-weekly.ru/analyt-ics/tech/45540.html (accessed 1 June 2016) (in Russian).
11. Orlov S. (2008) TsOD: rossiyskaya spetsifika [Data center: Russian specificity]. Zhurnal setevykh resheniy LAN (electronic journal), no. 09. Available at: http://www.osp.ru/lan/2008/09/5535759/ (accessed 1 June 2016) (in Russian).
12. Pavlov A. (2008) Inzhenernye resheniya, ispol'zuemye dlya snizheniya zatratpri ekspluatatsii TsOD [Engineering solutions used to reduce costs of a data center]. Open Systems. Available at: http://www.osp.ru/data/670/942/1238/10.pdf (accessed 1 June 2016) (in Russian).
13. Gabrielyan V. (2009) Kakpostroit'sobstvennyy TsOD [How to build your own data center]. Byte Russia. Available at: http://www.bytemag. ru/articles/detail.php?ID=15495 (accessed 1 June 2016) (in Russian).
Оценка совокупной стоимости владения центром обработки данных
Л.А. Пирогова
менеджер проектов LEAN Компания Lamoda
Адрес: 115114, Москва, Летниковская ул., д. 10, стр. 5 E-mail: lbaydalina@gmail.com
В.И. Грекул
кандидат технических наук,
профессор кафедры управления информационными системами и цифровой инфраструктурой Национальный исследовательский университет «Высшая школа экономики» Адрес: 101000, г. Москва, ул. Мясницкая, д. 20 E-mail: grekoul@hse.ru
Б.Е. Поклонов
кандидат технических наук,
доцент кафедры управления информационными системами и цифровой инфраструктурой Национальный исследовательский университет «Высшая школа экономики» Адрес: 101000, г. Москва, ул. Мясницкая, д. 20 E-mail: bpoklonov@hse.ru
Аннотация
В современных условиях наблюдается рост спроса на услуги ИТ-аутсорсинга, что влечет за собой активизацию процессов проектирования и строительства центров обработки данных (ЦОД). Поскольку ЦОД представляет собой сложную и дорогостоящую систему, возникает задача обоснованного выбора будущего проекта на основе показателей оценки затрат, которые могут возникнуть на этапе проектирования и эксплуатации центров обработки данных.
В работе анализируется один из возможных комплексов показателей для оценки затрат на создание и эксплуатацию центров обработки данных. В процессе анализа выявлены основные группы капитальных затрат при создании ЦОД, которые не в полной мере учитывались при оценке суммарного объема капитальных вложений по ранее предлагаемым методикам. В статье предложены регрессионные модели оценки проекта строительства центра обработки по двум показателям. Предложено оценивать капитальные затраты в зависимости от проектируемой площади технических площадок и от проектируемого количества стоек серверов. На основе разработанных моделей проведен анализ строительных площадок центров обработки данных, который показал адекватность модели реальным данным. Были установлены основные группы операционных затрат на содержание ЦОД и предложена регрессионная модель их оценки. На основе регрессионного уравнения предлагается рассчитывать мощность центра обработки в зависимости от площади технической площадки или количества стоек серверов. Стоимость эксплуатации центра обработки данных определяется, исходя из величины мощности. Анализ информации о стоимости эксплуатации различных центров обработки данных достаточно хорошо согласуется с расчетами, полученными на основе разработанной модели.
Предложенные модели позволяют с приемлемой точностью оценить характеристики проекта создания и последующей эксплуатации центра обработки данных.
Ключевые слова: центр обработки данных, регрессионная модель, строительство, репрезентативности выборки, структура затрат, операционные затраты, капитальные затраты.
Цитирование: Pirogova L.A., Grekoul V.I., Poklonov B.E. Estimated aggregate cost of ownership of a data processing center // Business Informatics. 2016. No. 2 (36). P. 32-40. DOI: 10.17323/1998-0663.2016.2.32.40.
Литература
1. Экономика ЦОД || Radius Group. [Электронный ресурс]: http:||www.rvip.ru|l065|documentl07l.shtml#5 (дата обращения 0l.06.20l6).
2. Patel C.D., Shah A.J. Cost model for planning, development and operation of a data center || Internet Systems and Storage Laboratory — HP Laboratories Palo Alto. [Электронный ресурс]: http:||www.hpl.hp.com|techreports|2005|HPL-2005-l07Rl.pdf (дата обращения 0l.06.20l6).
3. Hamilton J. Overall data center costs || Perspectives. James Hamilton's Blog. [Электронный ресурс]: http:||perspectives.mvdirona. com|20l0|09|l8|OverallDataCenterCosts.aspx (дата обращения 01.06.2016).
4. Лебедев П. Коммерческие ЦОД в России: смена бизнес-модели || CNews. [Электронный ресурс]: http:||www.cnews.ru|reviews| free|cloud|articles|articlesl3.shtml (дата обращения 01.06.2016).
5. Обзор рынка датацентров Санкт-Петербурга в I полугодии 2012 г. || Comprice.ru. [Электронный ресурс]: http:||www.comprice.ru| articles|detail.php?ID=475362&print=true (дата обращения 01.06.2016).
6. Особенности построения небольшого датацентра в Москве || telecombloger.ru. [Электронный ресурс]: http:||telecombloger. ru/14535 (дата обращения 01.06.2016).
7. Сухов Р. Как рассчитать затраты на ЦОД? || Журнал сетевых решений LAN. 2008. № 04. [Электронный ресурс]: http://www.osp. ru|nets|2008|04|4944787| (дата обращения 01.06.2016).
8. Павлов А., Басистый Д., Кусакин Д. Бизнес-план и финансовая оценка создания ЦОДа высокого уровня надежности || iKS-Media. [Электронный ресурс]: http:||www.iks-media.ru|issue|20l2|7|44308l5.html (дата обращения 01.06.2016).
9. Статистика ЦОД || AboutDC.ru. [Электронный ресурс]: http:||aboutdc.ru|page|64.php (дата обращения 01.06.2016).
10. Data-центры растут как грибы... || it-weekly.ru. [Электронный ресурс]: http:||www.it-weekly.ru|analytics|tech|45540.html (дата обращения 01.06.2016).
11. Орлов С. ЦОД: российская специфика || Журнал сетевых решений LAN. 2008. № 09. [Электронный ресурс]: http://www.osp.ru/ lan|2008|09|5535759| (дата обращения 01.06.2016).
12. Павлов А. Инженерные решения, используемые для снижения затрат при эксплуатации ЦОД || Открытые системы [Электронный ресурс]: http:||www.osp.ru|data|670|942|l238|l0.pdf (дата обращения 01.06.2016).
13. Габриелян В. Как построить собственный ЦОД || Byte Россия. [Электронный ресурс]: http:||www.bytemag.ru|articles|detail. php?ID=l5495 (дата обращения 01.06.2016).
БИЗНЕС-ИНФОРМАТИКА № 2(36) - 2016
