CS1000e split across a LES circuit (ELAN/TLAN isues)

[ansellrk]

Afternoon,

This is a fairly complex question so I will try and lay it out as best as possible!

We have a totally Nortel network and voice system split across 150 sites. The four main nodes all have passport 8600 using OSPF and are joined together in a diamond fashion for resilience.

We wanted to split our CS1000e across two sites but we have been told (and it appears true) that the ELAN & TLAN are not routable which means that we have had to enable spanning tree across these sites.

The main issue though is that the default route for the TLAN sits on a single site which means if that goes down the redundant portion won't work anyway!

Does anyone else run a CS1000e across two sites like this (using OSPF on Nortel platform)? If you do could I ask for some configuration information on how you have achieved this please? We believe there MUST be a better way but we don't seem to be getting anywhere with our support company!

Any advice, documentation or confirmation would be welcome!

Thanks

Rob

 

[allenmac]

Who told you the ELAN & TLAN were not routable?

 

[dulfo666]

Elan and Tlan are totally routable. I get to the elan of our New York switches quite happily from over here in London. Tlan is more than happy to be routable as our multiple TLan Vlans show. Also having Softphones over broadband.

Splitting you CS1000e across sites is more than sensible.

 

[allenmac]

I agree with dulfo666. the CS1000e is designed to be installed in a distributed environment, where everything communicates via IP.

It would not make sense to make the ELAN and TLAN non-routable.

I currently have 9 CS1000's and all of the TLANs and ELANs are routable - that is what the default gateway is for.

 

[ansellrk]

I had a feeling I was being given duff information! OK. So can you explain how you do this? How does the ELAN route? We have even read Nortel documentation that supports this! How did you configure the High speed link between the two halves of the CS1000?

I know one of you has said that the ELAN is in New York and London... but this can be achieved without having seperate subnets and making it a single VLAN. Not best practise I know.. but it can be done!

When we had seperate ELAN/TLAN the CS1000e wouldn't communicate properly and we were told by our support compnay that this because the ELAN must exist on a single network.

I also want to make sure we are not talking about MG1000's here on remote sites... that is a totally different situation... I am talking about the core CS1000e. I also want to ensure that we are talking about the TLAN as Nortel sees it (between the CS1000E) not the CLAN where the IP phones exist.

I am talking about the two halves of a CS1000e existing on seperate sites.

Sorry to go on but I want to make sure I am on the same wavelength as everyone else! Cos our support company are adamant about the ELAN!

Thnaks for your input.

 

[allenmac]

This is the definiation of Campus redundancy - installing the 2 cores in diverse locations.

See the Software Input Output Administration NTP NN43001-611 (Release 5) and look in the LD117 section.

There are commands such as:

CHG L3ELAN x Active <address>
CHG L3GW x <Address>
CHG L3HSP <side> <address>
CHG L3HSPGW <side> <address>
CHG L3HSPMASK <side> <mask>
CHG L3MASK <side> <mask>
CHG L3MASK <side> <mask>

These are the commands to configure the HSP IP networking values. You will need to read the section in the book for detailed descriptions for each command.

 

[ansellrk]

OK... from NTP NN43041-310 is says...

The CS 1000E redundant architecture allows for the separation of
Call Server 0 and Call Server 1. The two processors are connected by either
a direct 100BaseT crossover cable or a carefully engineered Layer 2/VLAN
infrastructure.
Campus Redundancy provides the ability to separate the CS 1000E Call
Servers in a campus environment for “campus mirroring”. This feature
enables two CP PM or two CP PIV Call Servers, one active and one
redundant, to be connected through an Ethernet network interface. Campus
Redundancy can operate using any vendor’s Layer 2 switching products, in
addition to the BayStack 470. The distance depends upon network parameter
limitations specified in Communication Server 1000: System Redundancy
(NN43001-507).
To separate the redundant Call Servers, the ELAN subnet and the subnet of
the High Speed Pipe (HSP) can be extended between the two processors with
an Ethernet switch, using Layer 2 protocol.

It mentions the LAYER 2 protocol every time... no mention of Layer 3?!

 

[tnphoneman]

The elans need to be on the same network segment. Remember that the offline cpu will take on the ip address of the active elnk host when it becomes active there for both cpu's must be physically on the same vlan/subnet.

As for the HSP that part is very true. This link MUST be a layer 2 tunnel from one cpu to the other and is NOT routable.



Signature===========================================
Artificial Intelligence Is No Match for Natural Stupidity.

The latest survey shows that 3 out of 4 people make up 75% of the population.

The original point and click interface was a Smith & Wesson.

Red meat is not bad for you, it is the green fuzzy meat that is bad.

 

[ansellrk]

This confirms the information from my supplier and the Nortel NTP's... so the ELAN is certainly not routable due to the fact it has to be on a Layer 2 protocol.

This probably means that we will bring the CS1000e back into our main site and put an MG1000 where the other halve of the CS1000e currently is!

Thanks for the input. It does also make sense!

 

[ansellrk]

So... can I ask. For you people with CS1000E... how many have it deployed on a single site using MG1000's for major remote sites and SRG50/200 for smaller remote sites?

This is the way we will probably end up doing it and it seems to be the only solution!?

Thanks

 

[dulfo666]

The elan IS routable. The HSP is NOT.

 

[ansellrk]

How can the ELAN be routable! The document CLEARLY states that the ELAN has to use Layer 2?!

 

[dulfo666]

Ye. And if you go back to the early NTPs, Nortel stated that the Elan had to be physically isololated from the rest of the network. The latest recomendation is that the Elan is routeable so that you can have your devices sitting in different places.

That is why applications like Contact Centre 6, now sit in the Nortel server lan, although they must link to the elan side of the Nortel PBX (via routing).

The HSP (High Speed Pipe) is the only link that is not routable. This is the link between the processors. It must be on its own 'logical' network.

 

[ansellrk]

Can you send me a link for the document with that information in? I am looking at the CS1000e 5.5 NTP's (because this is the version we have) and it states that the ELAN should be seperate and on a layer2 network. To my knowledge these are the very latest NTP's and the very latest version of software?!

 

[ansellrk]

Ah Ha! Found a document with it in! It is NTP NN43001-260 (Page 149 onwards) titled "How to create a routable ELAN subnet"

Time for me to read and digest and then put it in front of my support company!!!

Cheers for the advice I will update you all with the final result.

 

[ansellrk]

mmmm... not really a great deal of information. But there is a statement in the NTP that says...

ATTENTION IMPORTANT!
The primary and secondary TLAN network interfaces must be in the same subnet and broadcast domain.
The primary and secondary ELAN network interfaces must be in the same subnet and broadcast domain.

Surely this means it has to be a layer2 network?

I am sooooooooo confused!!! Nortel say one thing in one paragraph and then counteract it with another.

 

[tnphoneman]

Here is the important thing. You have an active ELAN IP address that all ancillary applications such as CallPilot, Contact Center, Signaling servers use to communicate with the Call Servers. You then have a secondary ELAN IP which in inactive until such a time that the command is issued to split the cores and the offline core will activate the secondary ip address.

When the offline cpu becomes the active cpu it activates using the primary ELAN ip address but it has it's own connection to the ELAN network. Basically it is a hot standby. They don't have to be in the same room or even in the same building but they do need to be in the same vlan as the callpilot and other applications will need to be able to communicate with either CPU but the CPU's will always use the same IP address they are just connected to separate ethernet ports.

Now the second connection on each CPU is what is called the High Speed Link(HSP) This is the link that must have a layer 2 network between them and is not routable.

I think the routable issue with the ELAN is not properly explained in the manuals. It used to be that Nortel recommended that the ELAN subnet not be routed to any other part of the customer's network because of broadcast storms rebooting the PBX. What they mean by routable is that you can route to and from the elan subnet.

CPU example: CPU 0 is active and the elan ip address is 10.1.1.1. Now for some reason CPU 0 goes off line, now CPU 1 becomes active but the IP address for CPU 1 is now 10.1.1.1. This is why they need to be on the same vlan. As I said earlier the other secondary IP address only comes into play when the CPU's are split and the offline cpu uses that one.

I hope I didn't confuse you any more and that this cleared some things up.

Signature===========================================
Artificial Intelligence Is No Match for Natural Stupidity.

The latest survey shows that 3 out of 4 people make up 75% of the population.

The original point and click interface was a Smith & Wesson.

Red meat is not bad for you, it is the green fuzzy meat that is bad.

 

[ansellrk]

I have read what you have written three times and I still am not sure I really get it! If both processors need to be in the same subnet (10.1.1.x) then you can not have a routeable network between them. They need to be on one logical network. I see what you mean about the routable part on the NTP, Nortel are talking about routing OFF the ELAN, not within the ELAN.

The real problem is that we use OSPF and have a totally layer 3 network. If we try to put layer2 traffic on the network then I have to enable spanning tree. If I try to close a spanning tree loop on the network all hell breaks loose!!

I know the easiest option for me is to put the CS1000e into a single site and use dark fibre to connect both the HSP and the ELAN and this may well be the solution I opt for as the complexity of the network is such that messing around with it could be more hassle than it is worth. We have already had loads of issues with IST's, SMLT's etc!

I think my mind is clearer on this now which is all I really wanted. Thanks for all the advice. I actually have a meeting on Monday with one of the Nortel technical guys so armed with what I know I will tackle him on the subject.

If anything more comes out of this I will update this thread.

 

[tnphoneman]

You are correct the Elan needs to be on one logical network because depending on which one is active that is the one that controls that IP address.

Good luck.

Signature===========================================
Artificial Intelligence Is No Match for Natural Stupidity.

The latest survey shows that 3 out of 4 people make up 75% of the population.

The original point and click interface was a Smith & Wesson.

Red meat is not bad for you, it is the green fuzzy meat that is bad.