Skip to content
Create an account for full access.

System levels in systems engineering. Example of computing technology.

In modern computing technology, it is easy to identify a dozen system levels. Here is an example of distinguishing these levels from the bottom to the top for classical computers (there are more of them; here is just an example of thinking about these levels, different manufacturers may distinguish them differently for different projects):

  1. Pure materials (silicon and carefully dosed "impurities" that give a semiconductor effect).
  2. Transistor parts (source-drain-gate).
  3. Field-effect transistor and inter-transistor connections.
  4. Logical elements (consist of several interconnected transistors).
  5. Arithmetic-logic unit/computational core, memory, and other parts of processors and various types of memory.
  6. Processors/chips: general-purpose (CPU), for graphic/matrix calculations (GPU), for network switching between processors, random access memory.
  7. Computer boards with processors and memory, external devices (e.g., solid-state drives).
  8. Computer rack with several computer boards, related input-output devices, power supply system, and cooling. Connecting cables.
  9. Server rack for multiple computer racks with power and cooling.
  10. Pods consisting of several interconnected server racks.
  11. Data center made up of several "pods," which may even include a telepresence robot with a camera for nighttime surveillance. A supercomputer at the data center level (as of June 2020, the seventh most powerful in the world) could have been assembled several years ago in just 3.5 weeks by only six engineers[1].

Pure materials at the scale of transistor parts have dimensions comparable to atomic sizes (the target for 2023 was 1 nanometer, to be reached by 2030, although it is not exactly about the physical sizes of the transistors[2]. Data centers with supercomputers today weigh tens of tons and have hundreds of cubic meters in volume. Engineers organize their thinking around system levels: systems at each of these levels consist of parts from lower levels and are themselves part of systems at higher levels. Systems at each of these system levels are designed by engineers who can solve problems at these levels. Together, all these levels provide the computing power that cannot be achieved using a reductionist approach, "a supercomputer straight from pure materials," and even "a supercomputer made of transistors."

It's absolutely unnecessary to call the systems at all these levels "system," "transistor system," "server rack system." The system is a type of object. We can "sentence" each word with its type, but it will sound strange. We don't say "the process the plane system is flying," we say "the plane is flying." We don't say "the central processor system in the computer system," we say "the central processor in the computer." This does not change the fact that the thinking of computer engineers is systemic, and their activities are organized around system levels in computing technology; there are many of these system levels, but in our oversimplified example, there are eleven.

When talking to people unfamiliar with systemic thinking, it is important to omit the type "system" at all these levels** (including the variety of system types — target, supra system, subsystem, system in the environment, and later there will also be a creating system). For a colleague familiar with system engineering and management standards, or at least having taken our course, it would be appropriate to say, "X is the target system," which is equivalent to X::"target system." For other people, it is better to say "X" or "X target" or even "it is important to consider X," but the word "system" as a type is better left unspoken if the interlocutor is unfamiliar with these types. But for your own thinking, it is essential to keep track of the type.

The success of today's computing technology is achieved precisely by the multi-level organization of computers—from the level of computer chips down to the transistor parts, as well as from the level of an assembled computer (rack for a data center) to the level of the entire data center. This is exactly the same increase in complexity as that which occurs in biological evolution and is governed by the same regularities[3]. Successes in electrical power engineering, rocketry, transportation machinery manufacturing, engineering as a whole—they are achieved in the same way, by the multi-level organization of their target systems, as well as by the multi-level organization of creator systems (for example, a factory for producing semiconductor chips, a factory for producing rockets, etc.).

The division of labor among people goes along the lines of different types of systems, but also along the lines of different system levels of the same type of systems. Systems at different levels are created/constructed/enabled by different "work methods"/"types of practices"/"content of work" of various creators/constructors/"enabling systems"/"creating systems" in various roles.

An engineer developing and producing pure materials for the semiconductor industry significantly differs in knowledge from an engineer manufacturing semiconductor chips using lithography. Moreover, their skills in development and production will also differ to the extent that they will be handled by different enterprises (some only develop, others only manufacture—and the specialization of the engineer personnel at these enterprises will be different as well). This is what the division of labor entails. The expertise of engineers producing pure materials, engineers who produce chips (integrated circuits) from them, and engineers designing computer racks and manufacturing these racks will significantly differ. However, all these different types of engineers and managers as engineers negotiate with each other about how systems of a lower system level are made into more complex systems of a higher level—including organization by enterprise managers who will handle all of this.

Systems thinking enables not losing focus in these complex collective negotiations on creating and modifying multi-level engineering systems and a complex network of enterprises that produce these systems. Not a single transistor channel, not a single fan in a rack, not a single cable in a data center, not a single department dealing with these systems, not a single type of equipment in these departments, not a single category of clients, not a single category of suppliers, not a single law of those societies where all this should work and bring irrevocable benefits to the most diverse agents-owners of various systems.


  1. Reference ↩︎

  2. Reference ↩︎

  3. Reminder of the sources where all this can be read in detail: first of all, you need to watch the film "Frustrated magnetism in solids" at https://en.wikipedia.org/wiki/Geometrical_frustration, then https://www.pnas.org/doi/abs/10.1073/pnas.1807890115, and https://www.pnas.org/doi/10.1073/pnas.2120037119. Taking into account that the genome is not contained within the target system itself, the system itself does not reproduce, but systems of creation/creators are made up of systems that have a memory of the target system. However, this changes little in the course of the reasoning. ↩︎