Article Preview
Top1. Introduction
The explosive growth of data from information digitization has been identified as the key driver the increasing requirements of data centers. A total of 161EB data were created and copied in 2006 (F. Ganz et al., 2007), and the amount of data increased to 1,800EB in 2011 (F. Ganz et al., 2008). The annual growth of data has reached 60%, and all data are distributed across global data centers. Over the past few years, many research efforts have been invested in designing highly scalable and fault-tolerant data center network structures because of the explosive growth of data(Al-Fares, Loukissas, & Vahdat, 2008; Costa, Donnelly, Shea, & Rowstron, 2010; Greenberg et al., 2009; Guo et al., 2008; Li et al., 2009; Mysore et al., 2009; Xie, Deng, & Zhou, 2013). The temperature in data centers significantly affects the failure ratio of high-speed network devices. Therefore, understanding the cooling efficiency of the thermal environment in data centers is important in implement highly fault-tolerant data center network structures.
Energy efficiency has become one of the most important challenges in designing data centers. For example, in 2007, ICT used approximately 8% of the electricity generated worldwide, of which approximately 4% was consumed by data centers (Borzycki, 2011). Moreover, new data centers in the United States are projected to demand 5GW of power (which is approximately 10% of the current generating capacity of California) and cost $4 billion/year to power in 2005 (Chase & Doyle, 2001). One average data center in the United Kingdom consumes more power than the city of Leicester in a year. The energy cost of a single data center in the United Kingdom is approximately £7.4 million a year until 2010 (Fildes, 2006). Moreover, network power consumption is expected to occupy 20% of the total ICT power consumption in Japan by 2025 (Imai, Leibnitz, & Murata, 2013).
A large number of network equipment exist in data centers and emit abundant heat. Network equipment fails if the temperature is high. Moreover, ICT is a large energy consumer (Vizziello & Favalli, 2013) and the cooling cost in data centers dominates the overall energy cost. Cooling cost reportedly constitutes a large portion in the energy cost of data centers (The Green Grid, 2011; Lee & Zomaya, 2012). Cooling cost can reach up to 50% of the overall energy cost (Sawyer, 2004). Even with the advanced cooling technologies used in Blue Gene/L, cooling cost still remains to be a significant portion of the total energy cost (Li, 2012). Figure 1 shows the cost factors of a typical data center. The figure demonstrates that cooling cost would exceed server cost n data centers (The Green Grid, 2011). Furthermore, Gartner (The Green Grid, 2006) claimed that traditional data centers waste more than 60% of the energy used to cool equipment. Therefore, improving the efficiency of cooling systems in data centers is important.
Figure 1. Cost factors of a typical data center