This is not a trivial question. Ethernet is relatively ubiquitous, in that most computers and peripherals speak ethernet. But whether ethernet (sic Fast Ethernet or Gigabit Ethernet) is the right fit for a business network infrastructure requires some deep consideration.
The answer may surprise you in this recent article .....
Quoting:
Like any network, latency at layer 2 is caused only by potential congestion when flows merge. In shared Ethernet of the past, it was somewhat more of an issue because flows always merged just to get access to the wire. And still, if the wire wasn't busy, Ethernet latency was if anything less than that of competing token passing schemes. But now with switched Ethernet, I don't see how anyone can make the case that latency is an issue, any more or less than it would be in any network with the potential for merging flows.
Not only that, but the 1500-byte frame length limit, which is the standard still, at the high speeds Ethernet is running now, allows for rapid interleaving of frames from different flows, much as the small cell size in ATM was designed to do (at much lower speeds). Which reduces latency for any individual flow, when flows merge.
As there isn't in ATM, or in SONET, or in most other layer 1 and layer
2 systems. The Ethernet layer does error *detection*, as do upper layers. Error correction is added in higher protocol layers, e.g. in TCP, or as forward error correction in UDP, or other error correction schemes one might gin up for UDP (including simple repetition).I can only think of three types of ethernet frames: the basic "type" and "length" formats, the 802.1Q extensions, and the new envelope frame descibed in 802.3as. There are far more options in, say, ATM, what with its multiple AALs, SVCs, PVCs, SPVCs. Extensions and options are added continuously, although they don't all apply to all products. If the various options in Ethernet aren't tested by vendors, it is not anything inherent to Ethernet. It is, if anything, because vendors try to keep costs down. Low cost versions of other schemes, if they were available anywhere close to Ethernet price points, would suffer from the same problems. And if these competing schemes were successful, they would also sprout a lot of extensions as well. Which would require testing.
Why offload the Ethernet checksum? And security is offloaded, because it's done at higher layers.
Cat-5 or fiber cables? No more fragile than used in other layer 2 schemes at similar speeds. And ruggedized variants of even the RJ-45 connector do exist.
Multicast support is inherent to Ethernet and to its MAC address format. With IP layered over Ethernet, one can make excellent use of multicast. In practice, this is limited mostly by what ISPs will allow over their nets. Not by the Ethernet layer. If Ethernet vendors fail to provide support for enough simultaneous multicast joins, you can expect that to be a cost-cutting measure; nothing at all to do with the design of Ethernet per se. So the same potential cost-cutting would occur with any other successful, mass-market layer 1 and 2 scheme, if any other actually existed.
It works perfectly well between two buildings, or even across the continent. The fact is, switched Ethernet becomes every bit as useful for long distances as any other layer 1 and layer 2 scheme, and its only real limitations would be its original lack of a built-in network maintenance features (now an OAM function has been added). Between buildings, this would not have been an issue anyway.
Maybe half-duplex Ethernet did. Since the days of shared Ethenet, the length limits of Ethernet are signal attenuation limits, which are purely a function of the physical layer and therefore apply to all schemes using the same physical media. When Ethernet uses the same phyical layer as SONET, e.g. in 10 or 40 Gb/s varieties, then the length limits are absolutely no different from those of SONET, which is a WAN technology.
I think it's misleading to attribute disadvantages to Ethernet that have nothing to do with Ethernet itself. For example, to imply that not well tested, cheap products are that way BECAUSE they are Ethernet. I also think that with the advent of switched Ethernet and higher speed Ethernet, many years ago, most of the other arguments vanish completely. If you have a strong case to make for isochronous service, maybe that's something to focus on.
Bert
Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.