Advancement of modern physics was indispensable to calculate the cooling cost of a computer
A large computer company such as Microsoft · Google · Facebook operates a huge data center. In data centers with large numbers of computers and servers, the problem of "cooling" is important, and it is built in places where cooling costs are kept low even a little, such as the seabed, river, and Arctic region. We are commenting that Santa Fe Institute, a non-profit organization founded in the United States, is working on these thermal problems.
Home | Santa Fe Institute
https://www.santafe.edu/people/profile/artemy-kolchinsky
Microsoft announced that it built a data center in the sea off Scotland in June 2018. This data center is a server rack with 864 servers and 27.6 petabytes of storage placed in a sealed metal container and submerged in the ocean floor.
One of the reasons Microsoft built a data center on the ocean floor is that the data center at the bottom of the ocean can cool the server at a lower cost. It is said that approximately 5% of American energy consumption is used to run computers, but most of this energy will eventually be converted to heat. Therefore, when moving a large number of computers and servers like a data center, it is necessary to strive to prevent the machine from being broken by enormous heat energy, and its cooling cost is a magnitude that can not be neglected at all.
The problem of energy consumption by a computer is not limited to an artificial digital computer. For example, the human brain is also a type of living computer, it is said to use approximately 10 to 20% of calories consumed by humans. However, the human brain is much more thermodynamically efficient than an artificial computer, and as long as it is healthy, it will not cause thermal runaway. In the current evolution theory, Santa Fe Institute points out that there has been little explanation on the thermodynamic consequences of "how the human brain got excellent efficiency".
The Santa Fe Institute says that the theory to accurately calculate the thermodynamic cost of such a computer has been established relatively recently.
In order to accurately calculate the thermodynamic cost of a computer, research on a field called "information thermodynamics" that combines computational information science and thermodynamics is necessary. Information thermodynamics is said to be a scholar born to solve the so-called " Maxwell's devil " paradox "If the second law of thermodynamics fails, assuming existence that can observe the movement of the molecule" It is.
With the introduction of information thermodynamics, a researcher appeared to calculate the thermodynamic cost of information processing. For example, from the 1960's to the 1970's, studies were conducted on the calculation of basic thermodynamic costs associated with two-way disappearance communication paths . However, it was said that physics at that time could not clearly formulate and could not formally analyze the thermodynamic efficiency of the computer.
However, since then, the field called "non-equilibrium statistical mechanics" has greatly improved, and by incorporating the concept of "information" in the world of this non-equilibrium statistical mechanics, the paradox of "Maxwell's devil" saw a solution for a long time . According to the Santa Fe Institute, non-equilibrium statistical thermodynamics is still not completely established as a theory, but even by combining non-equilibrium statistical thermodynamics and conventional information science / computer science, It seems that it became possible to analyze the basic thermodynamic characteristics of computers.
In the intersection of science and engineering, pursuit of information thermodynamics is proceeding. With the development of information thermodynamics, it became possible to accurately analyze the cost of cooling the computer, which enabled us to quantify the merit of constructing a large data center in the ocean floor or the Arctic Circle thanks to that . In modern information thermodynamics, not only "resource-time trade-off" in conventional computer science but also the physical speed, noise level, and calculation system of calculation are thermodynamically adjusted for a specific environment The "thermodynamic trade-off" of whether there is a problem is also treated as a theme.
by by Fir0002
With the further development of information thermodynamics, Santa Fe Institute commented that the analysis of the thermodynamic cost in biological systems such as the human brain advances, further commenting that the next-generation computer design and development will be more advantageous.
Related Posts: