Heat Density Trends in Data Processing - Term Paper Example

Hi I am sorry for the problems encountered. I am forwarding duly amended. Not including additional diagrams only two as it will require a lot of explanation which may be out of your paper length. Thanks. COMPUTING – ENVIRONMENTAL HAZARDS OF POWER AND HEAT Abstract The growth of information technology has resulted in rapid increase in number of computer-based systems and facilities, which have enhanced the requirements of power and consequently generation of heat. There is a general lack of awareness of the exact dimensions of the problems as its analysis in quantified terms is limited. Suffice to say that in just one year alone from 2000 to 2001 the increase in power density has been 100 watts/ft2. On the other hand there is differing perception in the manufacturers and the users in the heat density reduction parameters, which need rationalization. Thermal management also needs to be included while designing devices and the issue is likely to become more complex over the years. Reduction of power is also related to power aware computing, which is designing software and hardware to ensure optimum management of power. Introduction The growth of information technology has spawned a vast industry based around computers and information science. The impetus to economies and employment provided by computing ignores the silent yet alarming environmental hazard posed by systems compressed in small spaces requiring enormous amounts of power and generating large quantities of heat. Heat generated by computers is easily calculable, but one seldom takes into account the heat of hundreds of humans working in small spaces. There is increased awareness of environmental hazards of power and heat in computing, which are being addressed by the industry and academicians. The main issues to be considered are the nature and magnitude of the threat and measures that can be taken to minimize or overcome these. This is being carried out by a study of two prominent trends in computing environment, heat density and power aware computing. Section 1 - Heat Density Trends In this section we will examine the trends in power consumption and resulting heat dissipation in computers and data processing as well as storage systems and central office type telecommunications equipment. A White Paper, “Heat Density Trends in data Processing, Computer Systems and Telecommunications Equipment” is the main source of the study. (Uptime Institute: 2000). A number of other papers and presentations have also been considered to validate the trends in the Uptime Institute paper. There is a general feeling of smugness in the computing world generated by the optimistic growth prediction of Moore’s Law with semi conductor performance doubling every eighteen months. This has resulted in technology compaction, yet what has not been fully appreciated is that energy efficiency in computing equipment has not been correspondingly achieved. The vertical racking of servers to reduce floor space has resulted in heat dissipation of a magnitude of 10kW or higher within 2 foot by 2.5-foot area. (Uptime Institute 2000: 1,2). This has caused environmental problems as corresponding provisioning of air conditioning to limit the flow of heat from hot surfaces to the human environment and the need for a 24/7 cycle in the computer industry has created many unknown and little appreciated hazards today. This has received some attention in the industry as a collaborative effort within, to include major players as Amdahl, Cisco, Compaq and others to understand the trends is underway. (Uptime Institute 2000 : 2). The trends reveal that the annual rate of increase in power density over the decade from 1992 to 2002 in the case of servers, DASD and workstations would be to the tune of 15 %, while in the year 2000 to 2001 itself the increase has been 100 watts/ft2. The study also reveals that while sufficient power can be provided, there may be a need to provide water cooling or other methods to remove the heat from computers as the present air-cooled method involving fans and space differential between layers may not be adequate. (Uptime Institute 2000: 2). The standard means of calculating heat density value by the manufacturer is the footprint. (Uptime Institute 2000: 3). This is not a common standard used by the user, who prefers to calculate power density to include larger areas as aisles, white areas and so on thereby reducing watts per square feet. (Uptime Institute 2000: 4). There is however a need to rationalize between the two agencies. This problem has not been faced to a great extent so far as gross power densities have been low but will increase over a period. In addition to provide the customers in the IT industry ability to install high heat density computers and communication equipment with confidence and reliability and also to focus on the environmental aspect, a Thermal Management Consortium has been formed. (Uptime Institute 2000). The White Paper has identified four categories of gross space within a data center and calculated planning for space in relation to heat densities. These include, electrically active IT hardware product footprints, requiring 30 % of the space, service clearances another 30 %, site infra structure support equipment 20 % and electrically inactive areas another 20 %. (Uptime Institute 2000 : 5,6). Heat density trends are provided with predictive charts which are most useful for calculation of environmental effects in given facilities from power generation and heat densities. There are numerous other studies on power demands, which are supportive of the findings by the Uptime Institute. Research by a group of analysts of the University of Wisconsin’s, Madison Computer Sciences Center has carried out a detailed analysis of specific issues in thermal management to include the disk drive entitled, “ Disk Drive Roadmap from the Thermal Perspective A Case for Dynamic Thermal Management”. This study by a group of people including, Sudhanva Gurumurthi, Anand Sivasubramaniam, Vivek Natarajan and has highlighted the need for ecological management from the bottom up. (http://www.cs.wisc.edu). A study on server designs for handling high heat loads in Internet Data Centers, undertaken by an academic-industry group under David S De Lorenzo of Stanford University and Joseph Opdahl of Intel Corporation entitled, “Server Design challenges for the High-heat-load Internet Data Center” highlights the need for specific measures for microprocessor and system thermal management apart from facility cooling. The increase in density of servers would imply that power density is to be aligned to the thermal capacity created for facilities to restore the imbalance. However the focus of the paper is more on commercial criteria of reliability, cost effectiveness and so on rather than the environmental hazards, which are implied. (http://www.electronics-cooling.com) In a symposium on Next-Generation Thermal Management Materials and Systems the, “Challenges, Opportunities, and Technology Solutions for Microscale Cooling in Photonics and Electronics” was discussed. The discussion included a large number of devices including cell phones and MEMS, which require effective thermal management. The conclusion was for the need to understand technologies, “in the context of competing solutions and end-user needs, will provide key insight for R&D planning, product development, and commercialization efforts—for those who are driven to provide the next generation of electronic, photonic, and other integrated multifunctional products.” ( http://techventure.rti.org). The final research paper on the subject is entitled, “Dimensionless Parameters For Evaluation of Thermal Design and Performance of Large-Scale Data Centers” by Ratnesh K. Sharma, Cullen E. Bash, Chandrakant D. Patel of the Hewlett-Packard Laboratories, Palo Alto. The Paper confirms that large-scale data centers will be major energy consumers in the next generation and extended racking has led to very high power densities at the room level creating big heat loads and thus the need for effective design of data centers for creating energy balance. (http://www.hpl.hp.com/) Section 2 - Power Aware Computing The second trend being examined is that of Power Aware Computing. In their introduction to Power Aware Computing, in a series of papers published under the IEEE Computer Society, Mircea R Stan and Kevin Skadron of the University of Virginia call for understanding the functioning of computing systems which are deemed to be complex and require refinement to gain optimal advantage by designing appropriate power aware strategies both hardware and software. They feel that the application of power can be sub divided into various categories such as peak power, dynamic power, average power and so on and solutions evolved to optimize power in computing systems most effectively to enhance performance, reduce complexity, cost and power tradeoffs. Thus creating a new field, power aware computing. (http://csdl2.computer.org) Stan and Kevin have reiterated the trend in computer manufacturing and installation over the years, where the mantra they claim is to have faster, smaller and cheaper computers and only recently lower power has been added as a criteria. (http://csdl2.computer.org). The study reiterates that temperature is a byproduct of power dissipation. The main issues are discussed in a series of 25 papers. The Paper on, “Energy Management for Commercial Servers,” Charles Lefurgy and colleagues has underlined the need of ensuring that the power is kept under control by combining software and hardware power management techniques. In, “Dynamically Tuning Processor Resources with Adaptive Processing”, David H. Albonesi and colleagues focus on the processor. The authors have proposed that elements should be used at critical times rather than continuously, thereby saving power without performance losses, an approach, which is uniformly recommended by many analysts. Taking the example of peripheral devices, “Reducing Disk Power Consumption in Servers with DRPM”, Sudhanva Gurumurthi and co-authors have proposed a, “ dynamic rotations per minute scheme” for disk drives as a energy saving means. The thermal envelope in disk drives Slide 1 has been explained in the slide above which clearly demonstrates the rising graph beyond 1 minute and a plateau after 44 degrees Centigrade. The Thermal constrained design is being explained in another slide by the authors as given below and establishes the relationship between the data rate, linear density, RPM and diameter. SLIDE 2 In, “Leakage Current: Moore’s Law Meets Static Power” by Nam Sung Kim and colleagues recommend variations in power supplied to systems when active and idle thereby saving power. The modeling of Batteries for portable devices is being discussed in “Battery Modeling for Energy- Aware System Design,” by Ravishankar Rao and co-authors. This article also explores the specific issues related to energy aware designing for portable systems. (http://csdl2.computer.org) There are a number of other researchers working in the field of power aware computing who tend to support the analysis of the Power Aware Computer Team. A group of five teachers and a number of students in the Georgia Institute of Technology are working on the area of compiler optimization for power minimization where the main emphasis is on software based techniques for application based management of power aware computing. The researchers have examined the all-important issue of trading quality of service (QoS) parameters such as precision and performance in favor of power, and vice-versa. (http://codesign.ece.gatech.edu). Another study by Jose Gonzales of Intel Barcelona Research Center and Kevin Skadron, University of Virginia on, “Power-Aware Design for High-Performance Processors” emphasizes the need for power aware designing for low cost and mobile systems as well as for high performance systems. The issues studied include energy efficiency, dynamic and static power dissipation, and circuit and architectural techniques for reducing both dynamic and static power. (http://www.ac.uma.es). An International Workshop on Power-Aware Real-Time Computing (PARC 2005) held on September 22, 2005, has discussed power aware computing in systems from servers to embedded systems focusing mainly on software techniques. (http://www.cs.utsa.edu). A paper entitled, “ Profiles in Power: Optimizing Real-Time Systems for Power” by Graham R. Hellestrand, Mahdi Seddighnezhad and James E. Brogan calls for effective computation of power in terms of events in computing particularly related to modeling of software. The issue also highlights that, “advances in computer architecture and software have made it difficult or impossible to estimate or predict softwares execution time thus highlighting the difficulties envisaged in these processes. (http://www.cs.utsa.edu). A project on Power Aware Computing, Dynamic speed/voltage scaling for GALS processors by Shelley Chen, Anand Eswaran of Carnegie Mellon University has suggested that Dynamic voltage scaling (DVS) is a successful and scaleable solution to deal with growing power consumption associated with increased chip complexity. DVS is however not felt very effective when used with synchronous processors, performance being dependent upon clock speed. Thus higher clocking consumes greater power and lower clocking degrades performance. (http://www.ece.cmu.edu) Discussion An analysis of the main issues in Heat Density trends has revealed that technology compaction has not necessarily resulted in energy efficiency. Thus even smaller systems are consuming higher energy and thus releasing greater degree of heat causing environmental turbulence. While this issue has received some attention in the industry and collaborative efforts are underway these are purely from the commercial aspect of savings likely to accrued by reduction of heat densities thereby lowering cooling costs in systems as well as large spaces and facilities. The problems are likely to be increased with greater permeation of computers, an aspect also acknowledged in power aware computing. Heat density planning will be an essential facet of all devices, main frames and servers as well as facilities. Thermal management will have to be included in devise improvements as well as facilities planning for computers. The overall environmental effects of heat density however seem to have received limited attention. Power aware computing is seen as a relatively newer field and only recently lower power requirements have received attention in designing. Power aware strategies for computers have to be hardware as well as software driven. The symbiotic relationship between power and heat is a uniform trend emphasized in heat density reduction as well. Power saving can be achieved by critically dividing the activities being performed by computers and switching off power when certain functions are idle and providing it to those which are active. The need to support power aware computing as an environmentally desirable issue needs consideration. Conclusions Reduction of power and correspondingly heat densities in computing systems and facilities has not received adequate attention in international forums. While bodies as Thermal Management Consortium have been formed, these tend to see issues from a macro commercial and manufacturers perspective rather than a total systems or human environment perspective. There is also a need to establish a common standard between the manufacturers and users in promoting power aware computing and reducing heat densities. It is therefore considered essential that in the future, while creating a new facility, the gross average power and heat dissipation density will have to be calculated to ensure that site equipment is fully utilized and most efficiently contributes to the overall IT strategy of the company as well as the environmental conditions. It is also felt that environmental certificates from the point of view of optimization of power and heat densities should be made mandatory while establishing new facilities in computing. References: - 1. Uptime Institute Research Team. “Heat Density Trends in Data Processing, Computer Systems and Telecommunications Equipment”. Available. www.uptime institute.org./heatdensity.html. (4 March 2006). 2. Gurumurthi, Sudhanva, Sivasubramaniam, Anand and Natarajan Vivek. “ Disk Drive Roadmap from the Thermal Perspective A Case for Dynamic Thermal Management”. University of Wisconsin. Madison Computer Sciences Center. Available http://www.cs.wisc.edu/~isca2005/slides/02A-01.PPT. (5 March 2006). 3. Lorenzo, David S de. Opdahl, Joseph. “Server Design challenges for the High-heat-load Internet Data Center”. Available http://www.electronics-cooling.com/html/2005_feb_a1.html. (4 March 2006). 4. Symposium Next-Generation Thermal Management Materials and Systems. “Challenges, Opportunities, and Technology Solutions for Microscale Cooling in Photonics and Electronics” Available http://techventure.rti.org/summer200517/pdf/NextGen_Invite13.pdf. (5 March 2006). 5. Sharma, Ramesh K. Bash Cullen E. Patel, Chandrakant D, “Dimensionless Parameters For Evaluation of Thermal Design and Performance of Large-Scale Data Centers” Hewlett-Packard Laboratories, Palo Alto. Available http://www.hpl.hp.com/research/papers/2002/thermal_design.pdf. (5 March 2006). 6. Stan, Mircea R. Skadron, Kevin. “Introduction to Power Aware Computing”. IEEE Computer Society. Available http://csdl2.computer.org/comp/mags/co/2003/12/rz035.pdf . (5 March 2006). 7. Lefurgy, Charles. Et Al. “Energy Management for Commercial Servers,”. Available http://csdl2.computer.org/comp/mags/co/2003/12/rz035.pdf. (5 March 2006). 8. Albonesi, David H. Et Al. “Dynamically Tuning Processor Resources with Adaptive Processing”. Available http://csdl2.computer.org/comp/mags/co/2003/12/rz035.pdf. (5 March 2006). 9. Gurumurthi, Sudhanva, Et Al. “Reducing Disk Power Consumption in Servers with DRPM” Available http://csdl2.computer.org/comp/mags/co/2003/12/rz035.pdf . (5 March 2006). 10. Kim, Nam Sung. “Leakage Current: Moore’s Law Meets Static Power”. Available http://csdl2.computer.org/comp/mags/co/2003/12/rz035.pdf. ( 5 March 2006). 11. Rao, Ravishankar. Et al. “Battery Modeling for Energy- Aware System Design,” Available http://csdl2.computer.org/comp/mags/co/2003/12/rz035.pdf (5 March 2006). 12. Gonzales, Jose. Skadron, Kevin. “Power-Aware Design for High-Performance Processors” Available http://www.ac.uma.es/hpca10/tutorials.html#tut2.html. (5 March 2006). 13. Hellestrand, Graham R. Seddighnezhad, Mahdi. Brogan, James E. “ Profiles in Power: Optimizing Real-Time Systems for Power” Available http://www.cs.utsa.edu) (5 March 2006).. 14. Chen, Shelley, Eswaran, Anand. Article on GALS processors. Available http://www.ece.cmu.edu/~schen1/ece743/index.html. (5 March 2006). Bibliography Web Sites 1. http://codesign.ece.gatech.edu 2. http://csdl2.computer.org/ 3. http://www.cs.wisc.edu 4. http://www.electronics-cooling.com 5. http://techventure.rti.org 6. http://www.hpl.hp.com 7. http://www.ac.uma.es 8. http://www.cs.utsa.edu 9. http://www.ece.cmu.edu/~schen Papers 10. White Paper, “Heat Density Trends in data Processing, Computer Systems and Telecommunications Equipment” is the main source of the study. www.uptime institute.org. 11. Skadron, Kevin, Stan, Mircea. Power Aware Computing. Introduction to IEEE Computer Society papers. csdl2.computer.org. Read More