Hi all,

I've got a question I'm hoping someone might be able to shed some light
onto for me about the B1600 chassis. We're looking at Blades as as a
possibility to replace a raft of Netras and Intel boxes which are taking
up too much space & power.

Basically, in our current network topology, each server has 4 interfaces,
and is in two networks; Each server having a redundant connection
to another switch for the network in case we lose a switch or NIC. This
is all managed using the usual interface failover script.

Onto the question. I understand that the B1600 can support two internal
switches, and that these can operate in a redundant fashion; my question
is whether or not we can run two fully redundant VLANs (for each
network) on those switches and still fully populate the chassis. (I
appreciate this wouldn't help us if we lose a 'NIC' on a blade, but I'm
far more worried about losing a switch)

Putting it another way, are there enough internal 'ports' on the switches
to support more than 8 blades in a dual-homed, redundant configuration?

Apologies if the question sounds confused; but I'm new to blades as a
technology and I've been thinking about this too hard.

Many thanks,

Steph

In article <pan.2004.10.21.14.23.06.562013@lineone.ten>,
Steph <stephanos@lineone.ten> wrote:
>Hi all,

>Putting it another way, are there enough internal 'ports' on the switches
>to support more than 8 blades in a dual-homed, redundant configuration?

Each blade has two ports, one connected to each switch. So the answer is yes.
I think your confusion stems from the 8 ports on the back. Internally they
are 24 port switches. 24-16=8 (this is what is left at the back).

If you have a chassie with only one switch you get no redundancy.
Even if you only have 8 blades.

>Apologies if the question sounds confused; but I'm new to blades as a
>technology and I've been thinking about this too hard.

If I misunderstood your question please fell free
to post a followup.

/regards
Fredrik



--
Fredrik Lundholm
dol @ ce.chalmers.se

On Thu, 21 Oct 2004 15:33:55 +0000, Fredrik Lundholm wrote:

> In article <pan.2004.10.21.14.23.06.562013@lineone.ten>,
> Steph <stephanos@lineone.ten> wrote:
>>Hi all,
>
>>Putting it another way, are there enough internal 'ports' on the switches
>>to support more than 8 blades in a dual-homed, redundant configuration?

Hi Fredrik, and thankyou for your response; I have a couple more questions
for you

> Each blade has two ports, one connected to each switch. So the answer is yes.
> I think your confusion stems from the 8 ports on the back. Internally they
> are 24 port switches. 24-16=8 (this is what is left at the back).

As I understand this then, the two switches do not share the same 'fabric'
of connections to the blades. i.e. I can have my blades in one IP network
that is redundant across both switches (using both blade interfaces), or
have two IP networks (one on each switch) with no redundancy; What I need
to know is whether or not I can have a switch set up with VLANs of two IP
networks, and the other switch monitoring it and doing some sort of
availability checking (e.g. heartbeat, or something) and 'takes over' the
connections if the first one fails. As you say each blade is connected to
each switch on each of it's interfaces, I suspect this isn't possible...

Which does lead me onto another question -- the B200X blades have 4 ports
IIRC; what do they plug into?

Again, I'm new to the whole tech so please excuse my ignorance.

Many thanks,

Steph

In article <pan.2004.10.22.10.11.08.872708@lineone.ten>,
Steph <stephanos@lineone.ten> wrote:

>What I need
>to know is whether or not I can have a switch set up with VLANs of two IP
>networks, and the other switch monitoring it and doing some sort of
>availability checking (e.g. heartbeat, or something) and 'takes over' the
>connections if the first one fails. As you say each blade is connected to
>each switch on each of it's interfaces, I suspect this isn't possible...

Well, using a dual star toplogy with the blade server and IPMP
in Solaris I think it is possible. (And nothing prevents the two
switches to be cabled up together)

>Which does lead me onto another question -- the B200X blades have 4 ports
>IIRC; what do they plug into?

Aha interesting, never bothered with the B200x, but with four ports
they would connect two of them to each switch.

>Again, I'm new to the whole tech so please excuse my ignorance.

No problemo..

Fredrik



--
Fredrik Lundholm
dol @ ce.chalmers.se

Steph <stephanos@lineone.ten> wrote:
> As I understand this then, the two switches do not share the same 'fabric'
> of connections to the blades. i.e. I can have my blades in one IP network
> that is redundant across both switches (using both blade interfaces), or
> have two IP networks (one on each switch) with no redundancy; What I need
> to know is whether or not I can have a switch set up with VLANs of two IP
> networks, and the other switch monitoring it and doing some sort of
> availability checking (e.g. heartbeat, or something) and 'takes over' the
> connections if the first one fails. As you say each blade is connected to
> each switch on each of it's interfaces, I suspect this isn't possible...

Not only CAN you do that (although at the host level, not the switch level),
but you can do even better than that...

Firstly, failover. Simply use IPMP on the host between the ce0 and ce1
interfaces, one of which goes to each switch. So if one switch fails, the
OS will flip over to using the other port.

Secondly, multiple network. Unless I'm mistaken, the ce interfaces on the
blades support 802.1Q VLANs, as do the switches. So you can run multiple
VLANS, with IPMP running on top of each, to give as many redundant
networks to each blade as you want!

Scott