For the higher bandwidth pipes like OC3 and above....there are very few business situations where bandwidth increases have a significant impact on speed. Bandwidth plays but one part in the issue of IT performance, and in particular "speed" as perceived by the users. It is important to understand that transmission control protocol, the TCP in TCP/IP, plays a much bigger role. The job of TCP is to provide reliable data delivery. It was developed back when circuits were predominately analog and errors were common. Of course, the data flow was significantly different as well and consisted mostly of small strings of text. TCP regulates the flow of data between end points using an aggressive correction algorithm that is very sensitive to errors and latency. The impact of TCP on speed begins with just a 10 msec latency. An OC-3 bandwidth pipe (155 Mbps) has a theoretical throughput of about 19 MBps, assuming maximum packet sizes and minimal IP and routing overheads. That OC-3 pipe with 10 msec round trip delay will have a maximum throughput of about 6 MBps. Add an error rate of 1%, and that drops to 2 MBps. That 1% is between nodes, so it includes any errors from premise wiring as well as any errors across the WAN. If the latency is 100 msec, the maximum throughput on that OC-3 pipe is about 500 KBps. This is all TCP related and excludes any processing time on either node. A DS-3 bandwidth pipe (45 Mbps) can almost achieve that same rate of performance. Now that is all assuming only two devices communicating for the duration of the data transfer. As the device count increases, the impact of congestion increases. Still, circuit congestion is rarely the problem with business IT performance. Regardless, most applications never come close to achieving maximum performance. In most cases with 780 Kbps connectivity or higher, the communications delay represents less than 10 percent of the total time for any IT task. The remainder of the time is with processing at either end. Likewise, few applications optimize their data content to take the best advantage of the available bandwidth. The payload in most packets is full of application overhead from poorly optimized processing. Before leasing more circuit capacity, know what you are using (percentage of capacity), how you are using it (usage by application), and what speed issue you are fighting (task analysis). With this information, you can determine if you need more bandwidth, less latency, application optimization, or maybe a jump off of TCP for some specific requirements. To navigate through the complex assessment and decision process for determining the appropriate bandwidth configuration for your business IT application(s)....I strongly encourage the aid of an unbiased independent consultant. The technical advisior team at DS3-Bandwidth.com will provide this service to you at no cost. Just one less issue for you to worry about. |
