Ethernet and IP

 

Within a protocol stack, fibre is really layer 1 - the physical layer. There are several mechanisms used for the next layer or layers up - eg, ethernet, MPLS, ATM, SDH - this web page is an overview of some of them.

In most commercial LAN`s, within buildings, the physical layer is usually UTP, but increasingly, fibre is used as the backbone medium. It is also used extensively for WAN backbones. Whether it is single mode or multi-mode, uses wave division multiplexing, LED`s, lasers - that`s all part of the physical layer, and not covered in this web page. This page is more about the data link layer and the network layer.

 

Ethernet

The internet, and most current commercial LAN`s are based on ethernet, and on IP - most commercial LAN`s are now based on copper UTP, however ethernet is increasingly being offered as a fibre backbone mechanism as well.

Ethernet and IP are two of the layers in the whole IP protocol stack, the whole stack has the following layers - as far as the carriage of data is concerned :-

The IP stack is not based on the ISO OSI 7 layer model, as the original creators of the IP stack felt that the ISO OSI 7 layer stack was too rigid a design. So they went their own way.

There are some other protocols used instead of TCP and UDP, and occasionally, IP - however these other protocols are mostly used for various kinds of network management and system messaging, rather than for the carriage of data.

In the IP stack shown above, the original data link layer was a single layer called Ethernet II. However it looks as if through the years, the data link layer has evolved into two separate layers, which are specified in the IEEE 802 standard.

Most Ethernet networks are based on CSMA/CD - carrier sense multiple access / collision detection. It is a sort of unmanaged network anarchy - but has the means to deal with contention built in. It was developed by the University of Hawaii as a way of interconnecting various campuses on different islands. CSMA/CD is part of the MAC layer.

In CSMA/CD, each node connected to the physical network has an equal right to transmit data to another node - when a node wants to transmit data to another node, the sending node listens to the network, if it hears some kind of traffic, it waits until that traffic is finished. Once it hears nothing on the network, the sending node starts to transmit the data onto the network. At the same time as it is transmitting, the sending node also listens - if it starts to hear traffic which is different from what it is sending, then it knows that some other node has started transmitting at the same time, and it ceases the transmission, waits a while, and then tries again.

Ethernet works well and can provide fast connections, as there is no management system giving each node permission to transmit data.

However as traffic builds up, there is inevitably more collisions and resends, so contention becomes an issue, and latency can increase considerably. Various techniques have evolved in order to try and reduce latency - such as priority tagging and VLAN`s., along with switches instead of hubs, and layer 3 switches.

VLAN`s`s - virtual LAN`s - these are a way of dividing up a physical network into seperate virtual sub-networks by adding a VLAN tag to the ethernet frame. It works at layer 2, and in a normal UTP corporate LAN, they can reduce traffic considerably in any particular VLAN.

The existence of a VLAN structure is indicated by the addition of a 4 byte tag added to the MAC frame, after the Preamble / Start Frame Delimiter / Destination MAC Address / Source MAC address. This additional tag is recognised by VLAN aware switches and routers, and is stripped off by the edge switches, before the frame is sent to the final destination computer.

This tagging is often referred to as IEEE 802.1q, as that is the standard that defines it, as an extension to the MAC frame.

An additional technique is to map IP subnets to the VLAN structure, and then the dividing up can be done at the layer 3 level, which can be a lot faster.

For priority tagging, a 3 bit word is used to set a priority - this gives 8 levels of priority, with level 0 being the least important, and level 7 being the highest priority. Level 7 is used for network control.

These 3 bits are contained within the same 4 byte tag that is used for VLAN specification. CSMA/CD itself doesn`t know about it, but routers and switches will recognise the priority setting.

This priority tagging is also referred to as IEEE 802.1p, as again, that is the IEEE standard that defines it.

There is a problem associated with the VLAN or priority tagging - because the header of the MAC frame that contains the tagging extension has increased in length by the 4 bytes of the tag, the maximum permissible length of the MAC frame has increased from the normal ethernet length of 1518 bytes, up to 1522 bytes. If a MAC frame of length 1522 bytes is sent to a network device that doesn`t know about the tagging extension, the device will see it as a mal-formed frame, and drop it. So VLAN or priority tagging can only be used within a network segment in which all the network devices are IEEE 802.1q and IEEE 802.1p aware. At the edge of this network segment, the tagging must be removed, before onward transmission.

Finally, although as said earlier, most ethernet is based on CSMA/CD, 10Gb/s ethernet doesn`t use CSMA/CD - it is based soley on full duplex, so there is no problen with contention, as the node at each end of the fibre has exclusive use of the fiber to send data.

 

IEEE 802

The IEEE has a whole suite of standards connected with specifications for ethernet LAN`s and WAN`s - the IEEE 802 standards.

IEEE 802.2 is the specification for the Logical Link layer - it`s only 253 pages.

IEEE 802.3 is the specification for the Media Access Control layer - it comes in 5 sections, and section 1 alone is 671 pages.

IEEE 802.3 contains several different sub-standards, ie, IEEE 802.3a, IEEE 802.3b ........ IEEE 802.3z, IEEE 802.3aa, IEEE 802.3ab, ....... IEEE 802.3as, all of which cover numerous parameters of the MAC layer, from 10Mb/s up to 10Gb/s.

Historically, IEEE 802.3 hasn`t always been part of the IP stack. When the IP stack was being designed, there was an IEEE standard that was still under development, which was IEEE 802.3. The IEEE was trying to create an open standard for a data link layer that didn`t care about what protocol was running above it. However the IP designers reckoned that they wanted a type field, to identify the protocol in the next layer up, so they went with an original Xerox designed datalink layer, called Ethernet II. Eventually the IEEE took it on board, and built a type field into IEEE 802.3.

So the IP / ethernet stack has two type fields -

So at this point, the data link layer for ethernet networks is very well specified for different provider inter-operability.

 

MPLS

MPLS - Multiprotocol Label Switching - is a mechanism for forwarding packets through a network of routers, as an alternative to ethernet.

MPLS uses a set of labels which it attaches to the incoming ethernet frames, and then uses these labels to control the routing of the frames. Each hop has its own label, which is generated by the router that is about to launch the frame onto the next hop.

An MPLS label is 32 bits in length -

Where the label is placed depends on what protocol the MPLS network is carrying.

MPLS routers are configured with routing tables, which they use to choose the routing for the next hop.

So a bit like VLAN tagging in ethernet, an MPLS network is a specific network segment where all the routers must be MPLS aware. An MPLS network segment is called a domain.

Ethernet is regarded as a connectionless-orientated routing system, as each frame is sent on its way along whichever path the router thinks is the best path at that time - two successive frames can follow different paths, with different transit times, and IP has to be able to deal with this.

However MPLS is regarded as a connection-orientated routing system, as the frames are sent on pre-configured paths, with all the frames in a sequence following the same path, and so arriving at the other end of the MPLS network in the correct sequence. This makes the routing algorithms for MPLS much simpler than those for ethernet, and routing can be faster, reducing transit times across the MPLS network segment.

MPLS allows for traffic engineering - the route that frames will take is specified to the routers, so that the particular route can be configured to the maximum benefit of the type of data in the frames.

It also means that Quality of Service is easier to maintain, so signals like VoIP, audio, video, and SCADA all benefit from MPLS. Another type of data traffic that can benefit from Quality of Service is remote or virtual desktop operation - it can be very frustrating for a user to have to wait for responses to their mouse and keyboard actions.

In some respects, an MPLS route through an MPLS domain is like a tunnel, and is equivalent to a VPN through the network. MPLS is in fact used to create VPN`s, and although an MPLS VPN does provide traffic isolation, and hence some intrinsic security, MPLS doesn`t allow for any kind of encryption, so can`t provide secure VPN`s. However the VPN data could be encrypted before it reaches the ingress router of the MPLS domain.

MPLS packets can be any size, unlike, for example, ATM, so an incoming ethernet frame can be forwarded through an MPLS domain without being chopped up and re-assembled at the MPLS network edges, which further aids transit times.

MPLS works at layer 2 and a bit above, so it is sometimes referred to as layer 2.5.

MPLS - as the name suggests - will carry a wide range of network protocols, such as IP, ATM, and frame relay.

However MPLS has to be aware of what network protocol it is carrying, as the way the label is added depends on the network protocol being carried. In addition, MPLS has to interact with the network protocol in certain ways. Firstly, of course the ingress MPLS router has to be able to read the source and destination addresses.

Secondly, MPLS has to decrement the TTL counter for each router hop, in just the same way that, for example, ethernet routing does. In IP over ethernet, every router decrements the TTL counter by one - if the counter gets down to zero, the frame is regarded as too old, and is discarded. MPLS routers have to do the same thing.

It depends on the network protocol being carried by MPLS as to how, or if, this is done. If the MPLS domain is carrying ethernet / IP frames, then the MPLS label contains a TTL field - when the ethernet frame arrives at the ingress router, the router reads the value of the TTL field inside the ethernet frame, and puts that value into the TTL field in the MPLS label. Every MPLS router in the domain decrements the TTL counter by one. When the frame reaches the egress MPLS domain router, the egress router puts the value in the MPLS label TTL field back into the ethernet frame TTL field. So once it is through the MPLS domain, the ethernet TTL field contains the same value it would have done if the ethernet frame had gone through the same number of non-MPLS routers.

Some fibre backbone providers advertise that they offer a service called VPLS - virtual private LAN service. This is really a way of using MPLS to preserve VLAN information across an MPLS domain. It is one of the areas of MPLS that is still under development, there are different ways of achieving VPLS.

There are two sets of routing information required for an MPLS domain.

 

IP Transit

IP Transit isn`t a technology like ethernet or MPLS, it is a kind of service offered by some carriers in order to transport IP traffic across the internet. However since it is strongly associated with fibre WAN backbones, it is worth including on this page.

Because it is a service, rather than a technology, there doesn`t seem to be too much technical information about it on the internet - searching on Google reveals a huge number of companies offering IP Transit, and you can pick up various bits of information from some of them.

Before the carriers can offer IP Transit services, they have to have -

This set-up is known as an "autonomous system", and there are three types of autonomous system :-

The "autonomous system number" uniquely identifies each routing domain on the internet, and they are used by, and are essential to, the internet routing protocol, BGP. Without an ASN, BGP would not be able to find the domain.

ASN`s were originally 16 bit numbers, allowing for 65536 different assignments. Since about 2008, 32 bit numbers have been issued, expressed in the form of x.y - where x and y are 16 bit numbers.

Within the 16 bit format, various ranges of numbers are used for specific purposes :-

It appears that at least some of the carriers offer two levels of IP Transit :-

Before a customer can connect to an IP Transit provider, they have to have one or other of :-

Some of the carriers offering IP Transit can provide IP addresses out of their own allocation.

Lastly, I guess that the best IP Transit provider for any particular customer who is looking to connect to the internet will be the provider who can connect to the Tier 1 providers with the least number of onward hops. So the IP Transit provider should either be a Tier 1 provider themselves, or they should have at least one direct connection to a Tier 1 provider. Multi-homing to two or more Tier 1 providers will add to resilience.

 

 

 

 

 

website design by ron-t

 

website hosting by freevirtualservers.com

 

© 2024   Ron Turner

 

+                                   +  

 

Link to the W3C website.   Link to the W3C website.