Distributed System Structures

In spite of the modern-day ubiquity of the internet—in homes, workplaces, and mobile devices—the vast majority of users do not understand even the most basic of concepts concerning types of software-defined networks, their differences, their benefits, and their shortcomings. However, for the most part, that state of consumer understanding is just the way those who design and service networks would like things to be; indeed, contrary to the common use of the word “transparency,” it is well-known in such circles that the role of transparency here means in one sense to obscure from the user the perhaps intimidating or even frightening truth that single processes are now carried out over a diffuse network of devices; that instead of data transfers and operations being carried out on the user’s device alone, such functions are distributed over multiple devices, largely for the purpose of increasing speed. Much less still does the general public understand that there are different varieties of distributed system structures, though many may loosely be aware of the idea of a local-area network, a form of network operating system, due to the prevalence of such systems in the workplace. However, such a system, though it has its advantages, to be sure, is made to look almost archaic in capability when compared with the other form of distributed system structure known simply as a distributed operating system.

First turning to the idea of an efficient VLAN network operating system, it is obvious both that such a system has advantages in that it is simple to set up and maintain and that this selfsame simplicity greatly decreases its capacity to perform the more complex tasks that can be accomplished with a true distributed operating system. The set-ups employed in creating a network operating system are quite often symmetrical, such as the common “ring” or “star” systems—the former being a connection between each device (such as a computer) and a neighbor on either side, and the latter being a system in which each device is connected to every other device in the network (Tanenbaum & Van Steen, 2002). Both terms become more obviously descriptive when a map of the topology of the given network is drawn in visual form. As far as the transfer of data within such systems is concerned, the commands are often quite explicit, e.g. “get” commands, which makes it quite obvious to any system administrators just what exactly is going on at any given point. This makes identifying problem terminals easier than with the distributed operating system. However, the limitations on the network operating system dictate that it can only be used for a relatively small number of devices connected over relatively short distances (Tanenbaum & Van Steen, 2002). If the system is set up otherwise, latency plagues the data transfer and makes for a frustrating user experience. Where speed and large distances are concerned, by far the superior distributed system structure is the distributed operating system, which indeed is what comes first to mind when the phrase “distributed system” comes to mind, though of course network operating systems are distributed, too, in one sense.

Distributed operating systems function primarily to spread the workload of processing tasks over multiple computers or other devices, thus speeding up the process and decreasing the possibility that any one device will become overtaxed and perhaps freeze or crash. To do this, however, multiple layers of protocol are required, and indeed some amount of processing power is consumed in simply determining what portions of the task will go where (Tanenbaum & Van Steen, 2002). In addition, it can be very difficult to identify which link in the chain is broken when failures occur, though of course in some ways a chain is an inadequate analogy. The reality is that distributed operating systems lend themselves much less easily to the almost childlike diagrams often used to represent network operating systems. Like the social connections that hold people together in their lives, distributed operating systems feature a constant shift and flow of data, a continual re-making of who is helping whom in any given moment. There are “clusters,” like family units that present discrete groups for accomplishing particular tasks, and just as in life, load-balancing prevents any one entity from collapsing under the pressure.

When taken together, network operating systems and distributed operating systems offer two different structures that each have advantages and disadvantages, from the simplicity of local-area networks to the efficiency of parallel processing done on multiple devices, as in distributed operating systems. Understanding both is paramount to anyone seeking a deeper knowledge of the way computers can be connected in this day and age. In fact, in some ways, the study of networks can even be a source of awe, for in reflecting on the variety of system structures humans have created, we gain greater knowledge of the way our own interconnectedness functions.

Reference

Tanenbaum, A. S., & Van Steen, M. (2002). Distributed Systems: Principles and Paradigms (2nd ed.). Upper Saddle River, NJ: Prentice Hall.