For those outside looking in. ... it seems that is the "Broadband Nation" in the economy a little crazy all over Gigabit Ethernet capabilities. Careful .... There's more,) the story as the eye (or bandwidth meter .... "wink".
You can summarize it as such .... Users often than not, both inexperienced and power users, seem under the impression that the bandwidth associated with the end of the speed with which all the data end. Situation in the fact that only half the equation. You must be sent to the actual volume of data and views .... To determine whether there are any concerns, bottlenecks or other performance problems are.
Now take a step back and look really closely at your use. For example, you have probably often asked to check if the links to various servers were saturated, ie, mail server hosting hundreds of mailboxes.
One would expect it to have very much to do, when in fact the most basic lines hardly moved past, a> 25% light or in this case 25Mbits/sec (100Mbps link) and for the record that combines the most frequently both directions on a full duplex connection.
The moral of the story is do not need or even want 1000Mbits/sec or even 10000Mbits/sec. It is, if the need or want is 101Mbits/sec (in whole numbers here) for argument's sake that the next phase flow and / or warrants 1001Mbits/sec quantify the major considerations for 10000Mbits/sec.
The price per port Gigabit Ethernet offerWarrants certainly do not need to stay away violently. However, careful planning will continue to ensure recycling to save the ever-growing copper infrastructures and on the now outdated necessity of dietary fiber-for-performance (in many cases) that were not too many years not heard by most. Now you can get a gig switch from Linksys or D-Link, for a few pennies. The bang for the buck is at a great price point. Costs elsewhere in high-end components do not have other values really important at this time, butMore good reason.
I would suggest first steps while looking at 100Mbits/sec infrastructures and how to possibly teaming or aggregation of bandwidth used to add links such as Fast EtherChannel or Switch Assisted Load Balancing (eg with respect to Cisco and HP respectively). These simple steps create by using network cards / ports one at a time, often to replace close to gigabit speeds without the switching element devices. This approach may cost moreand an effective and proven technology rarely used. However, he offers a lot more than just speed and redundancy. Of course, these are aggregation methods have been developed in the Gigabit market. But in general they hold value in the super high-end throughput volume requirements ..... and they are not usually used in the access layer (also known as the desktop layer).
A few years ago led the Cisco 40xx series switches. The first model had several slots, each of whom had 6 Gigabitthroughput. Two cards can plug into this slot. They had six 1-Gigabit slot, intended primarily for inter-switch trunking. The other had eighteen 1-Gigabit slots, which seem to be at first glance a bad idea.
The rationale for the 18-port card, however, was that a typical Windows-based servers of the time, with the typical processors was only 300 Mbit / s drive, or so the throughput. There was a net benefit to inserting such a server to a gigabit port, as would the individual bits in and Clockin a nanosecond, but as 10 nanoseconds on a Fast Ethernet interface. Without a lot of queuing theory, that is, statistically, a good thing.
100-MB cards in parallel still holds down the bit rate, but also adds a non-trivial costs of additional cables and physical ports.
Traffic statistics can be misleading if you just look at the average transfer rate and conclude the link is little used. Especially when you compare it with delay-sensitiveApplications such as VoIP, you need to look to the peak, not average utilization, because it is at a peak when you are most likely to experience delays. For routine transaction processing, an average load can be so good, but not for audio and video, which is the cumulative effect of the delay.
Now I am a big fan of Gigabit Ethernet. But the driver is QoS in a very broad sense, not only traditional traffic analysis, but more how closely related video / multimedia can be supported worldwideconnected computer to connect to server clusters.
Review the existing tele-traffic distribution in most companies and most do not need big GE, but some do already, and some really dynamic organizations will be shouting yes, please.
If we change the perspective, there really are organizations out there who want to be on those services that could not previously provide by means of LAN congestion or overuse of resources are available on the server cluster.
Remember thatoffers no overall best solution for all companies and users.
The prices for Gigabit Ethernet infrastructure is a major factor in the implementation of it in every organization. As more and more server vendors to equip their servers with 10/100/1000 NICs (powered by Intel and Broadcom chipsets), (network-side of the house, if not purchased and installed / in the last 2-3 years), most likely 10 / 100.
It makes sense that all new servers to 1000 Mb / s capable, butserving on the counter next to the things, if it is not a widespread need for 1000 Mbit / s, can not be retrofitted with it your entire architecture. What is a blade or two on the core switches, and switch a smaller distribution or two of the few points that aggregation need it.
Since the price per port arrives for GigE, cash in on the supply and trade-ins, to update your devices. In addition, you will see your bandwidth. It has always been a historical trend, where, whenever there is additional capacityPeople really lax on the management of resources.
640K used to force a memory barrier, the people with their code to be efficient, now you have gigs to concerts, where people are careless with their code, database indexes, etc. and finally, the more memory. The same applies to the network traffic. But I'm not sure that you are stingy with your bits, just sensible.
10GE is now a different story, and I can only see the need, not in linking universities and buildings with asignificant proportion of users with high bandwidth. The price of a single distribution is 2-3 XENPAK 10GE switch, or even a bandwidth shaping device.
The final practical message ..... Choose wisely.