[Horwitz02] Section 11.12 Planning for Future Network Capacity

来源:百度文库 编辑:神马文学网 时间:2024/04/28 15:06:15

Planning for Future Network Capacity

Network capacity isone the most limiting factors in any technology environment, due to thelarge amount of infrastructure needed to implement it. For example,just a single T1 from your office to the Internet requires a physicalcircuit (a PRI line) from the local telephone company, CSU/DSU units totranslate the T1 signal, a router to connect the CSU/DSU units to yournetwork, and an ISP to provide Internet service at the other end of thecircuit.

All of thiseffort and its associated cost gets you 1.544Mbps of bandwidth to theInternet. If you suddenly need more, you'll have to repeat this arduousprocess for another T1 or possibly higher bandwidth line. There areseveral measures you can take to make sure your organization has enoughflexibility in its bandwidth solutions as well as the foresight topredict problems before they arise.

Watch Long-Term Trends in Network Traffic

Trend analysisis an important part of system administration, and nowhere is this moreobvious than in network planning. A small company with 16 employeesmight start out with a single T1 to the Internet for Web and emailfunctionality, realizing that their average utilization will rarelyapproach the 1.544Mbps capacity of the line. The company's usage willgrow with the company, however, and eventually that same 1.544Mbps willbe a bottleneck for the company's communications. The astute systemadministrator monitors the bandwidth utilization over time andtherefore can predict the approximate point at which more bandwidth isnecessary.

Figure 11.4shows an example of one year's worth of bandwidth utilization on afractional 768Mbps T1 line. Since late January, utilization hassteadily increased (when read right to left), but is still nowhere nearthe capacity of the T1 line (98KBps). However, now that the systemadministrator knows that utilization is increasing, he or she can watchthe graph in the next few months to see if that trend continues andrecommend more bandwidth before the situation becomes critical. If thisincrease in utilization can be correlated with other factors—such asnew hires—the administrator can predict future bandwidth needs simply by knowing how many people are being hired.

Figure 11.4. An MRTG graph representing one year of bandwidth utilization measurements on a fractional T1 line.


Average Bandwidth Utilization

The average bandwidth utilization on a circuit should be no more than 30%, with peak usage not exceeding 70%. Beyond this, spikes in usage can bring a circuit dangerously close to capacity and degrade the performance of other network traffic on the line.


Use Variable-Bandwidth Circuits

With the increasinguse of fiber and other high-capacity media, many ISPs are able to offercustomers variable-bandwidth solutions, even to their offices.Variable-bandwidth circuits usually offer a set bandwidth to which youagree to limit the majority of your utilization, but the circuits arenot physically capped at that limit; you are allowed to “burst” up to amuch higher maximum rate for short periods of time. This leeway allowsfor spikes in traffic; the circuits are therefore called burstable.

Burstablebandwidth is very convenient because it offers you the capacity of ahigh-bandwidth circuit like a DS-3 for just a little more than the costof a lower-capacity circuit like a T1. For example, if your expectedaverage utilization on a new T1 is less than 25%, but you occasionallyneed to transfer files several hundred megabytes in size in less than 5minutes, you should probably consider burstable bandwidth from yourISP. Many ISPs will charge by bandwidth per interval. For example, on aburstable T1 you might be charged 1 dollar for every 5-minute intervalthat your utilization exceeds 1.544Mbps.