One of the most fancy and cutting-edge applications of computer technology is in the building of simulations that emulate the behavior of real-life or theoretical systems or situations. Such simulations can serve a variety of purposes such as training and education, research, and entertainment. To take an easy example, certain experiments in psychology or behavioral sciences are either unethical or impossible to conduct--for instance, the assessment of how a particular person would react given certain accident situations often cannot be known well using standard means because accident situations cannot be created at will. The military may wish to train its soldiers in war, but does not wish to needlessly expose them to the actual bloody horrors of battle. A pilot may need to be trained to fly a new type of aircraft, but there is little sense in risking the plane and her own life during the training process itself.
For all these and many other such applications, it is desirable to create a simulation of the actual scenario and to then train or test the human subject or user on the simulation. This method is commonly used for training crews for the space shuttle, advanced jet fighter aircraft, and the like, and the creation of simulation equipment that realistically emulates the real situation is a complex and involved task.
There are a great many challenges to the creation and use of simulations, but the major ones continue to concern traditional computer science issues. For a long time, the design of realistic simulations was limited for the reason that hardware required to operate the simulation in real time was either unavailable or prohibitively expensive. However, with the rapid growth in processor speeds in recent years, that is not a serious problem any more, and processor support for realistic simulations is inexpensively available. However, many other technical concerns remain that prevent simulations themselves from yet being commonplace and inexpensive.
If a simulation is to convey a sense of reality and emulate the real world, it must operate in real time from the perspective of the user. The CCITT G.114 Delay Recommendations (CCITT is the acronym for the old French name for the International Telecommunication Union, ITU--see <http://www.itu.int>; the acronym still survives in the literature and the standards prescribed under the former name) for systems that interact with human users are as follows:
0 to 150 milliseconds--"acceptable for most user applications"
150 to 400 ms--"may impact some applications"
above 400 ms--"unacceptable"
In addition, users in a simulation must be able to perceive the effects of their own actions within 100 ms, else dissatisfaction and disorientation result.
The available bandwidth can be a critical factor in deciding the speed of the response obtained, and thus in deciding the richness of a simulation, hence having a large bandwidth available would appear to be desirable. This is especially true since a simulation typically would be expected to carry some sort of image data, which in turn typically require enormous amounts of bandwidth.
However, larger bandwidth can come with its own assorted set of problems, especially so over larger networks. On local-area networks (LANs), bandwidth has not been a major issue because standard Ethernet (10 Mega bits per second) appears to work relatively well and is inexpensive; however, the demand for LAN-based simulations is rather limited. The far-greater demand is for wide-area network (WAN)- and Internet-based simulations, where the bandwidth currently is generally limited to T1 speeds (1.5 Mbps) and where increasing the bandwidth creates new problems.
Current protocols such as the Ethernet do not scale well to high bandwidth requirements, at least given current hardware capabilities. The Ethernet protocol requires each node to look out for a collision between its transmission and that of any other node on the network, and re-transmit if a collision occurs. In order to detect a collision, however, the network must have a minimum packet size, which is proportional to the desired bandwidth and the distance of transmission. For example, for a 10 megabit Ethernet running over 300 miles, the minimum packet size should be about 6400 bytes. Current architecture however cannot support such a large packet size. If one uses standard packets (40 to 60 bytes in size) and pads them with overhead bytes to reach the minimum size required for collision detection, one ends up wasting a large majority of the available network bandwidth, and no advantage is achieved for the simulation. Achieving high bandwidths is thus a critical bottleneck in building successful simulations, and continues to be one of the chief reasons why simulations are generally too expensive for widespread use.
For these reasons, simulations continue to remain restricted for the most part to military and highly specialized applications, and much research interest at present focuses on making them widely and inexpensively available to the general public.
This is the complete article, containing 789 words
(approx. 3 pages at 300 words per page).
