Ive been spending a lot of time over the last couple of weeks staging/testing various pieces of equipment for a certain network configuration. The primary objective was to identify IOS versions, hardware and software configurations that will allow us to support certain MTUs required for MPLS and MPLS based pseudowires/x-connects. IOS is annoyingly inconsistent...
Heres a couple of photos of the little lab Ive been using for this testing.
Full lab overview
This is a shot of the full lab setup. We do a lot of work with radio links, and the testing being done is specifically related to the upgrade of an existing microwave network, and deployment of a new segment. Theres a few different Cisco switches (left rack: 3500XL, 3550-24; middle rack: 3560G, 3560, 2950-24T, 3550-12G), Cisco routers (middle rack: 1801, 2 x 2811, 2801; right rack: 3745), the radios (left rack) are Alcatel 9400 LX/UX's, and theres a couple of E1 modems sitting ontop of the 3745.
The other equipment is unrelated to what I was testing.
The radio rack and ODUs. We will be re-using several pairs of Alcatel 9400 LX/UX radios in the new network segment, so we have a pair linked up to allow us to test our hardware and software configurations as they will appear in the real world.
E3 and E1 tribs
The trib types involved in the new network segment are E1 and E3. The E3 trib (coax cable) is the primary trib for data, while the E1 trib is for a separate OoB/DCN network running along side it. Forgive the ghetto wiring of the E1 trib, that was done purely for testing purposes.
Closer look at the ODUs for the radio links. Ordinarily they would be direct mounted to an antenna or vice versa depending on the size of the antenna, or might have a flexible waveguide to connect to the antenna, but for our lab environment a pair of transitions and an attenuator in the middle will do.
Most of the Cisco gear tested
Most of the Cisco gear being used as part of the test. The final design will probably call for the use of 3550-24 series switches (bottom switch in the radio rack) and a combination of 2811 and 3745 routers. In the lower 2811 in this picture is a NM-1A-E3 module which is connected to one of the radios, and a WIC-1T connected to one of the E1 modems in the next picture...
...here. This 3745 also has an NM-1A-E3 installed, connected to the other radio, and thus linked to the 2811. It is also linked via ethernet through the 3550-24 to the 2801. The two E1 modems are connected to the two 2811's, and using the E1 trib on the radio are thus connected together. Our deployment will most likely use VWIC-1MFT-G703's installed directly into DCN routers to reduce the amount of active equipment.
Topology of the lab
And this is the topology of the network. Each of the routers runs OSPF and LDP, as MPLS was a big part of the testing performed in this lab.
The primary purpose of it was to confirm the maximum supported MTUs on each piece of equipment and whether they will be big enough to support:
* A 1500 byte IP packet encapsulated in MPLS: requires a 1508 byte MTU (1500 byte user payload + 2 x 4 byte MPLS labels), or 1526 byte raw frame size inc outer ethernet headers
* A 1500 byte IP packet transported through an MPLS based psuedowire: requires a 1530 byte MTU (1500 byte packet + 18 bytes of user ethernet headers + 4 byte pseudowire control word + 2 x 4 byte MPLS labels), or 1548 byte raw frame size inc outer ethernet headers
The results are positive in that all of the equipment we tested is capable of (at the least) 1546 bytes MTU or 1564 byte raw frame size on ethernet. NM-1A-E3 supports 4470 so is not a worry, and the E1 wont be carrying any MPLS traffic so was not tested.
Awesome. that's a lot of sexy gear.
have you had a chance to run a ber tester across the pseudowire network with that config to see what your delay and jitter is once the network is saturated vs idle?
You said the PW is for the Eth circuits, can you explain? I was under the impression PW's are for building TDM circuits across Ethernet transport mechanisms.
PW can be used to transport many things over MPLS. At least Cisco (maybe other vendors too) refer to it as AToM, or Any Transport over MPLS. It can even be used to transport POS/SDH over Ethernet instead of the usual other way around.
Essentially we stick an xconnect command on an interface at one end of the PW, then on a similar interface on the other we stick an opposing xconnect command. For ethernet you have to take into account VLAN headers, so both interfaces at each end need to have the same configuration, i.e. if you use a sub-int at one end you need to use a sub-int with the same VLAN ID at the other end aswell.
Its pretty simple the way it works. A received frame is wrapped up in some MPLS headers, then switched through the network to the other end as per a normal MPLS packet, unwrapped and transmitted as per the original frame.
Hence the reason why we have to support at least 1530 bytes MTU when transporting ethernet through a PW if we want to carry a full 1500 byte IP packet for the user.
* 1500 bytes for the user packet
* optional 4 bytes for user VLAN tag
* 14 bytes for user ethernet headers
* 4 bytes for the PW control word
* 2 x 4 bytes for a minimum of 2 MPLS labels
* 4 byte carrier VLAN tag if we are transporting through a VLAN (we do a lot of work with VLANs)
* 14 bytes carrier ethernet headers
* Total: 1530 MTU required to be supported by underlying network elements, 1548 bytes for the raw frame
Our equipment will be configured to support a maximum MTU fo 1546 bytes, as that is the maximum MTU supported by Cisco 3550-24 switches. This effectively leaves us 16 bytes to play with. More MPLS labels (carrier-in-carrier for example), Q-in-Q, etc are all possibilities.
My only experience with PW is TDM emulation... so what you say makes sense based on my very extremely limited knowlege of MPLS.
|reply to TomS_ |
Why use mpls on a routed net with a PTP. What is the benefit? Is this to avoid complex routing issues or is the mpls traversing further down the net? Just curious. It seems mpls has become the latest craze and sometimes used for the sake of using it. ie. Higher up buzz word(kill)
|reply to TomS_ |
IP VPN products for customers primarily. But we also use an MPLS VPN to house the management side of our WiMAX gear. Keeps it all nice and safe.
Those two among others.
But definitely not so we can claim another buzz word, its purely for functional reasons!
|reply to TomS_ |
What's a trib?
Tributary. Basically its a data channel transported by the radio.