Hi all, with ATM, My isp says he can create a pvc through there core to create a virtual link between our 2 sites, can anyone give me a quick overview how he would achieve this? and with what equipment.
ATM is an evolution of Frame Relay.
Both are WAN technologies that go behind leased lines.
With ATM your router acts as an ATM DTE speaking light protocols with the ATM switch port.
shortly: packets to be carried inside an ATM network are placed inside an ATM AAL5 PDU.
That PDU is segmented in ATM cells.
Each ATM cell is a small fixed size frame made of 53 bytes with a 5 bytes ATM header.
ATM header contains VPI and VCI fields
a given PVC uses specific ATM VPI/VCI
the end-to-end PVC is made by associating at each hop incoming VPI/VCI values to outgoing interface and outgoing VPI/VCI
to be noted that VPI/VCI are very local they have a different meaning on each link and so they don't need to be unique as are LAN MAC addresses.
so ATM cells don't contain a L2 destination address but just local information that is used to send them hop by hop.
A PVC is set up manually and during the path both VPI/VCI can change
at the remote end the ATM DTE receive cells reassembly the AAL PDU and extract the IP packet inside and can route it out a LAN.
the path is predefined and allows for QoS features: a PVC is associated to a CIR committed information rate and other parameters that specify how it will be used.
Traffic in excess the CIR and less then PIR can be dropped
Traffic in excess of peak rate is always dropped.
The equipment are ATM switches: they have specialized hardware able to switch cells quickly and efficiently and to perform QoS controls.
Hope to help
Reading about ATM can be confusing, but thankfully the ATM PVC configuration is rather simple.
Imagine the following setup:
a. BPX is a type of ATM switch (with e.g. a BXM-155, a multiport ATM OC-3c/STM-1 interface card)
b. 7206VXR is a router (with e.g. a PA-A3-OC3SMI module, a single-port ATM OC-3c/STM-1 port adapter)
c. STM-1 port of the router is connected to a corresponding port of the switch (at points noted 1,3 above)
d. ATM switches are connected to each other via another set of their ports (using an intermediate telco circuit noted 2 above)
To setup an end-to-end PVC (provided that equipment and cabling/circuit work):
a. ISP issues a command (e.g. addcon) on one of the 2 ATM switches. The command causes the ATM switch ports at points 1, 3 above to be virtually connected end-to-end. ATM switch signaling handles the end-to-end PVC setup (which includes cross-connects at each hop e.g. connect port 1 to port 2 at BPX a, and port 2 to port 3 at BPX b with specific local VPI/VCI)
b. After ISP has the PVC in place, you configure the router port with the VPI/VCI of the connection (must match with the one configured by ISP on the local link), encapsulation typically aal5snap, and if you add proper IP addresses at both router endpoints, you can ping the other side of the connection. From the router's point of view the PVC is like a direct connection to the remote router.
p.s. ATM traffic management specifications typically use terms like PCR, SCR, MCR instead of the FR terms like PIR, CIR, MIR.
This is very helpful ,thanks for your info.
Now to add, Does ATM to have to have a specific physical inteface, or could it run over a normal ethernet port ?, how is this set up on the router ?
Also with the PVC,s can you explain at little more with regards to the local VPI/VCI setup, how come they dont need to match at each end? are the ATM switches also configured with a vpi/vci between them?
look forward to your reply
To use ATM you do need specific ATM hardware for the SAR (segmentation and reassembly) process. The specialized ATM switches understand only cells (as described earlier by Guiseppe) and not IP packets. The SAR capable interface breaks packets into cells (segmentation) before sending them towards the ATM switch and collects cells received from the ATM switch to create packets (reassembly). Adaption layers such as AAL5 are used to help the router understand which cells are part of the same packet.
Configuration of an adapter typically depends on the hardware (before the pvc setup you might need some hardware specific tricks to prepare the adapter for further use). A typical PVC configuration for the router with the adapter in my previous post could look like this:
interface ATM2/0.1 point-to-point
ip address ...
An end customer might not actually need to use ATM specific hardware in the case where the whole ATM story is being handled by the provider. In the example of my previous post I used ATM specific hardware. If the provider switch is a more versatile model, it could also have frame-relay or ethernet support (besides the ATM). In this case, ATM is used only between the provider switches, and the switches take care of passing other types of frames (e.g. frame-relay) towards the customer routers (ATM-FR network interworking). You might decide to contact your provider to explore available options.
The ATM switches have an internal ATM switch matrix. You can think of the matrix as an ATM specific routing table.
BPXa switch matrix
inport vpi/vci outport vpi/vci
(1) k/l -(2) m/n
(2) m/n -(1) k/l
You read the first line like this:
if cell received at port (1) with vpi/vci=k/l, send cell out port (2) with vpi/vci=m/n
You use similar logic for the second line (which handles traffic in the opposite direction).
BPXb switch matrix
inport vpi/vci outport vpi/vci
(2) m/n -(3) p/q
(3) p/q -(2) m/n
Now, you see that vpi/vci is used also between the switches (cells travel inside the ATM network and cells have vpi/vci). The operator does not typically configure the vpi/vci between the switches. The operator typically configures only the endpoints of the PVC.
You also see that vpi/vci has to match at both sides of a local link, so that the specific PVC "traffic flow" can be identified by each endpoint of the same link (else you get an unknown vpi/vci error). The vpi/vci can be different at each different network segment, and the switches handle the mapping between different ports with their matrixes as shown above. For example, for cells flowing from left to right:
a. router sends to BPXa cell with k/l
b. BPXa maps k/l from port (1) to port (2) with m/n
c. BPXb maps m/n from port (2) to port (3) with p/q
d. router receives from BPXb cell with p/q
(Note: the way this mapping works is very similar to how mpls forwarding works with a possibility of different labels at each segment of an LSP)
The vpi/vci assignment can be the same at each endpoint of the same PVC as part of a convention/policy in the network to make things easier.
One more thing:
ATM is also used in DSL setups. ATM was specified as the layer 2 above DSL physical layer to speed up DSL deployment. In the DSL case, the DSL CPE is most probably already ATM capable and the PVC configuration is normally not very complicated (usually it is GUI assisted).
And another thing, irrelevant to ATM, but relevant to how things in this forum work. People who regularly post answers here are volunteers. They gain satisfaction by helping others, and most of them expect their useful posts to be properly rated using the rating system, especially by the person who opens a conversation. Now, for me, it's irrelevant whether you rate me or not. I normally get enthusiasted by the questions themselves, but most people are not like that. Since I do not have enough knowledge to answer all questions that you might have, while other people here might, I would advise you to rate posts that help you solve your issues. People here notice who rates posts and who does not, and might not bother to respond in the future.
Hope this helps,
ATM adds it's own overhead to the IP packet. You can have a look at the following figure to see how an IP packet can be segmented into cells: http://www.cisco.com/en/US/tech/tk39/tk48/technologies_tech_note09186a0080129a87.shtml#topic1
The typical case is to have some routers at the endpoints as I described previously. In this case, the routers could have ethernet interfaces and the routers typically strip off the ethernet overhead and transport the only the IP packet over the ATM cloud.
In the general case, there are no hard limitations to what can be transported over the ATM part of the network. The capability to actually transport ethernet frames over the ATM cloud can be part of the capabilities of a DSLAM. For this to work end-to-end, the DSL CPE should have a matching capability. To be more specific, imagine the following setup:
The early DSLAMs had an ATM uplink (e.g. STM-1 or 155Mbps speed) that would collect all ATM/DSL user traffic from all its DSL ports and forward it further. Later, to support higher collective user bandwidth, the uplink typically moves to GigabitEthernet. So, if you look again above, what we basically have is this: ethernet<->ATM<->ethernet
It can be possible with certain DSLAMs and DSL CPEs, to simply bridge the ethernet frames over the ATM part (between the DSL CPE and the DSLAM, ATM is being used). This configuration effectively puts the traffic of all DSL clients of a single DSLAM in a single LAN.
Even with the plain ATM-uplink DSLAMs, it can be possible to bridge the ethernet frames further inside the provider network. You can have a look at the following figure for the general picture:
The format of the bridged information is documented in RFC1483 (http://www.ietf.org/rfc/rfc1483.txt , e.g. page 7 of the document).