I have been interested in changing the design of my current topology from a 6 daisy-chained layer 2 switches, to 6 layer 3 stackable switches (Netgear GSM7352S). With stackability i will be getting a poor mans chassis design essentially, be able to maintain a stack of switches through a single IP, and redundant design (slave switch fails another takes over, master fails then copy back tables to another switch). Now with this ability to create an internal ring topology and use layer 3 protocols for inter VLAN communication why would i encounter the following issue from my provider :
In discussing this with some of our maintenance folks and backbone
engineers a couple things have come up.
In Regards to VLANS:
It is our policy to not allow units to route our IP space, all
routing needs to be handled by UNIV. - what that means in this
scenarios, is any VLANs you would want to create we would need to
route and thus would need to traverse your link to the building
equipment and back. This would mean all traffic between VLANS would
be limited to the 1 Gig you currently can achieve.
Besides this point, from looking over the documentation sheets and
looking into stacking technology, if you were allowed to route
between VLANs on your equipment you should be able to traverse
between VLANs all internal to the L3 stacked boxes. Stacking
together devices would effectively make them like a chasis device.
In Regards to Stacking:
The reasons we don't do stacking on campus is because it doesn't
scale well in our environment and management is not generally agreed
to be easy. There are also worries about how a stack handles a case
when a unit fails - does the stack redirect traffic around the ring,
or do the units below that failure suddenly lose connectivity.
I'm waiting to hear back from our maintenance team on a definitive
word on how our support agreement for your subnet or the rest of
** might be compromised by the addition of stacked
switches that are not UNIV approved. If they were to say this could
cause issues they are not willing to support we'd have to ask you all
to not implement the stack.
As I said on Friday, a 10G aggregation switch with a tree topology
would be the recommended structure. In light of the "No Routing
Policy" these could all be L2 devices as this is what they will
operate at. If you wanted to implement multiple VLANs we can look
into a UNIV. managed L3 device which would have connections to each
of your L2 switches.
That said, baring any issues from the maintenance team, if you were
to implement a L2 stack you are likely to see "better" transfer rates
between each rack, however without this netgear switch being a device
I've tested and worked with I can't say definitively that it will
give you any performance gains.
OK OK, now am i just not getting the whole VLAN concept or does this guy not know what he's talking about. Since i'm using VLANS on my internal network and just routing VLAN to VLAN internally, how would that affect the UNIV?