TO: Prof. Ellis

FROM: Remonda Mikhael

DATE: 3/3/2021

Summary of Article About Ternary Optical Computers Construction by wang ‘s, zang’s.

The following is a 500-Word summary of a peer-review article about “Ternary Optical”. The authors go into detail about the construction and theories of ternary computers by explaining how the theory and how they were tested and what were the results. 

Optical computers were gaining more attention because of the speed and power they operate at and the first ternary optical computer was proposed by Jin et al. Several theories and additions were added to the construction of the first ternary optical computer.  What had to be determined was how fast the quality of service would be which can only be tested once the construction was complete. This article will explain the four-stage design of the ternary optical computer. The remainder of this article will explain how the systems work, which algorithms used, and future potential direction it can take.

The ternary computer is broken up to three sections, each handling specific functions that also the computer to operate sufficiently. What makes this computer different from others is that it can process multiple inputs at once, reconfigure itself to understand the user’s request, and there is plenty of space in the processor to run complex algorithms. Queueing theory which is used to measure the speed of how fast the computer can operate which there are several different ways this is tested in different locations, which all of these tests are used to test this ternary optical computer. 

Two novel strategies will be presented: immediate scheduling and computing accomplished scheduling and how they are used. According to the authors, “Under the IS strategy, the data bits of the optical processor are equally divided into n parts and each part corresponds to a small optical processor that can be independently used” (Xianchao Wang et al., 9 February 2019, p.6243). It does this by sending data to unoccupied processors to so the operations can be done quickly. 

The other strategy is computing accomplished scheduling which happens after the request has been computed. This is done to make sure all requests are operated simultaneously. This is just by sending small bits of data to the processor first, then larger bits to the processor. The requests are received by the receiving server to transferred to preprocessing service to be process into a tri-value logic operator. The information received are done in a first-come-first-served basis which is then send to processor to translates the request into computer language and begins working on the requests.

The scheduling strategies and batch size has a direct result for various service processes. Using both strategies can increase the arrive time for results to be output because there are four stages each request must undergo before the request has been completed, but the increase in response time doesn’t increase dramatically because the system operates requests in parallel, so the delay is only affected minorly. Since all requests are not only operated in parallel, but are also broken into four smaller processors which is why the operating speed is quick for ternary optical computers.

Reference

Wang X., Wang X., Zhang S., Gao S., Zhang M., Zhang J. & Xu Z. (2019). Response time of a ternary optical computer that is based on queuing systems. The Journal of Supercomputing, 76, 6238–6257.