I'm taking my first networking class and I'm having trouble understanding which data link sublayer participates in encapsulation, adn which in framing. Right now from looking at sites found with Google and Wikipedia I believe the LLC sublayer is doing the framing, but some sites break down the frame as having some 802.2 sections and some 802.3 sections.
Can anyone suggest some reading that will help me gain a better understanding of this?
The anti-layer activists will perhaps complain, but ...
I'd say that framing, e.g. forming the Ethernet frame, belongs to the MAC sublayer. The lower half of the Link Layer. And that various encapsulations available at Layer 2 are provided by the upper half of the Link Layer, or LLC.
Should I be able to read and understand the 802.2 and 803 spec's to find this stuff? Is that how you guys find it? Those spec's make for some pretty dry reading.
Well, sort of. Some people believe, erroneously IMO, that the 7-layer model applied exclusively to the ISO/OSI protocol suite. Anyway, like Glen says, it's not strictly followed in designing hardware. But it is a useful conceptual tool anyway, again IMO.
I heard that said about the 53-byte ATM cell, i.e. but not about the 7- layer ISO model.
In the ATM case, the US wanted 64-byte payloads and the Europeans wanted 32-byte payloads. The average is 48 byte payload, + 5-byte header = 53 byte cells.
You expose the *source* of much confusion when you refer to "the hardware layer". *THERE IS NO HARDWARE LAYER!!* Similarly, there is no software layer.
The OSI model describes *functions*, and places those functions in a hierarchy delimited by a series of (somewhat arbitrary) layer boundaries. The model says absolutely nothing about implementation. The functions of any layer may be implemented in hardware or software; the choice is generally a tradeoff among cost, performance, ease of modification/maintenance, etc.
Historically, layers 1 and 2 were almost completely implemented in hardware, and the higher layers in software. For example, given the silicon technology of the 80s, it was impractical to implement the Transport layer functionality in silicon, due to cost and die size limitations. Today, that is no longer true, and we regularly implement the performance-critical portions of the Network, Transport, and higher layers (e.g., SSL encryption) in silicon.
By the same token, it is also feasible these days to implement much of the Physical layer in *software*; the receiver in a Gigabit Ethernet PHY chip typically includes multiple processor cores with sophisticated digital signaling (DSP) algorithms implemented in software.
The bottom line is: Don't confuse *architecture* with *implementation*. There is a complete discussion of this subject in Chapter 1 of "The Switch Book".
-- Rich Seifert Networks and Communications Consulting 21885 Bear Creek Way (408) 395-5700 Los Gatos, CA 95033 (408) 228-0803 FAX
Some might call it firmware. Is it in ROM (or EEPROM or FLASH), or RAM loaded by the driver on startup? It could be interesting to reload with a modified version of the DSP code.
(1) A standard is not a novel, I never read it sequentially. I'm not interested by all the descriptions in a standard, either because there are no applications, either it's obsolete, either it's too much mathematic, either it's not enough developped... all kind of reasons.
(2) When I'm working in a standard, I'm always in using the searching tool, trying to find all the occurences of a word. And often, I search words through multiple standards, it's easy with PDF.
(3) To do that I maintain my own database, and a synthesis card with hypertext links to the files which are classed in the directories.
(4) I work easier at home or... in the train, because at the office the phone often rings, people come and see me, the boss asks me boring things.
(5) And.. sorry... reading standards is incompatible with looking at the TV. But, as a consequence, it's more difficult to talk standard subjects to my wife or neighbours.
Thanks to everyone who posted for a very interesting discussion.
The instructor mentioned that companies are now transitioning to using IPV6 or both v4 and v6. Can I ask if anyone has been involved in this effort in their own company, and how smoothly it went?
Sorry, we don't believe here in a massive extension of IPv6 in the world. And our school is satisfied with IPv4.
Today, network operators have other cost-effective systems and fast equipments to number things or to deduce QoS, through T-MPLS, Carrier class Ethernet (EoT, ETY, PBT, Q-in-Q...), and even TDM.
You have to distinguish the service given to the client and the transport provided by the operator. The IP service is generally transparent to the operator which adds some layers below IP to transport the service without touching IP. And now Ethernet services are growing, so I let you imagine how the transport is in this case to keep e.g. VLAN and Ethernet addresses without touching them.
I meant that we don't believe in an important growth, development and spreading of IPv6 throughout the world. We see all-IPv4 inside entreprises, and all-Ethernet outside, even for peer-to-peer connexions.
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.