For laymen bandwidth is the only element that is related to the speed of the network. However, latency is another element that contributes to the speed of network. Corporations and companies now do need to insure low latency on their network to ensure high quality of service. Latency can be termed as the several kinds of delays incurred in processing of network data. A low latency network connection has minor delay, making businesses dealing in VoIP very efficient. On the other hand, high latency connections suffer from long time delays. The theoretical peak bandwidth remains fixed due to the technology used but the actual speed you obtain varies over time due to high latencies.
High reliability on network is another important thing that needs to be ensured by companies. High reliability can also be termed as high-throughput computing, in which company has to ensure the reliability and robustness over a long period of time scale. High reliability guarantees a perfectly reliable system even with unreliable components. The cost of having both high reliability and low latency on networks is very high, small to medium level companies have to consider a tradeoff between the two criteria. High level or multinational corporations can afford both things and do not need trade off. To easily understand the low latency and high reliability or throughput, we can take the example of a bus and car travelling parallel on the same road. Now which one will have low latency and which will have high throughput? Obviously, car will have low latency while bus will give a high throughput.
The decision between the high reliability and low latency really depends on the type of business and niche of the company. Every business should provide a reliable environment for users to operate, but at the same time they must not compromise the integrity of the execute node. Companies needing large amounts of computing power for short periods of time should go for low latency, these companies are concerned about the operations per second or how fast an individual job can be completed. However, the companies that are in need of large computing amounts for longer times are more interested in the number of jobs completed over a long period. These companies should go for throughput computing or high reliability.