Cluster+SMLT vs. Stack+D/MLT 2

[maxyz]

for a small lan (ca. 100 devices) based on a 55xx core and 45xx access the local nortel reps recommend the implementation of a cluster+smlt core based on two 55xx. i'd prefer a stack+d/mlt with the same two 55xx.

nortel argumentation:
- smlt for high availability, <1s failover times
- vrrp for L* redundancy/failover

my argumentation:
- stack and/or d/mlt failover times should be <1s aswell
- but i have a single router on L3 which is much less complex in administration

do i miss anything?

design attached

 

[HungryHouse]

Gotta a agree with Nortel on this one friend. SMLT is superior as you have far better redundancy to two different 5500 stacks. The configuration is not really complex. THE IST link between core switches is proven and reliable. There is ONLY one big advantage to YOUR design over Nortel...cost!
THE SMLT ability of the 5500 actually is only available with the Advanced Routing license which is NOT cheap. $5495 list price per stack! So...if you dont already have this, and money is tight, then the DMLT is a fine alternative.
I hope this is helpful
-HH
NNCSS- 8600
NNCDS- Ethernet Switching

 

[maxyz]

thanks for your answer hungryhouse.

SMLT is superior as you have far better redundancy to two different 5500 stacks.

i totally unerstand were smlt is superior over a spt based solution. but i want to compare the use of only two 55xx switches in the core either in an smlt or stack configuration. nortel keeps telling that "somehow" (they do not go into detail how, thats what i try to find out) an smlt based solution is faster in failover then an "in stack" failover, which i can hardly believe?

i think the reason they want to sell smlt in this project is as you mentioned the license pricing.

 

[007barney]

First of all let me state I work for Nortel... here is my take on the pros/cons...

5500 SMLT Core
+ Less disruption when upgrading your core. With 5.1 code you can load a new version of code on to the first 5500 and reboot, during reboot your core outage is minimal (same sort of outage as a link failure). you then upgrade the second 5500 and reboot. upgrading a DMLT pair is several minutes of outage while the stack reboots.
+ You can create lots of SLT's (single port SMLT, where the edge device connects to your core with only 2 links - 1 to each core) into the core from the edge.. I think the limit is the number of physical ports on the 5500. With DMLT you can only have 32 MLT's. This is a big factor for most people.

5500 DMLT Core
+ Simpler config
+ Simpler troubleshooting

If you choose the DMLT core option I would highly recommend using at least 3 5500's. This is because when stacking only 2 5500's if you lose a switch the remaining live switch will no longer detect a stack (since it is isolated on its own). This means that replacing the faulty switch with a spare is not an automated tasks (our process of upgrading the new unit to the same code and transferring the appropriate config does not work in this case). If you have 3 switches in a stack and one fails the remaining 2 switches will still detect a stack and automatic switch replacement process works fine.

All in all I think this stacks up to SMLT core as the best way to go but it really depends on:
- Availability requirements (e.g consider the outages for upgrades and hardware replacements)
- Cost
- Scale (number of SMLT/SLT links into the core)

Hope this helps?

 

[maxyz]

thanks a lot 007barney!! the advantages for slt/smlt you bring into the equation make a lot more sense to me than the up to now to me mentioned "somehow faster failover".

and i learned that what we really are looking into is slt (1 link to each core) - the local nortel people always talked about smlt (ok, conceptual/functional difference is minimal, but still...)

it looks like we will use stacking/dmlt as we weigh 'simpler config' and 'simpler troubleshooting', at least currently, higher than shorter outages in case of hw replacement and scalablity (i doubt this installation will ever come near the 32 mlt limit).