DWDM w/FCP->IP converters.

We are starting to plan out our DWDM core, one of the major apps driving
this will be our HDS 9980 and our net design group is thinking about using
FCP-IP converters with 10GigabitE backbone for our brocade switches which
will be located at different datacenters, one of the issues they brough up
which sounds valid is that we only had 32 channels/lambda and it would be
very wastefull to provide all these channels and we should look into doing
ISL trunking on the brocade switches and they will do the protocol
encapsulation and send it all over IP i'm still fairly new to the SAN world,
can anyone give me pros and cons to their proposal ?

Our DWDM core isnt dedicated for the HDS please keep that in the back of
your mind.

[D1] [D2]
[b] =[b] =[b]=[b]
\\ //
[b]=[b]
[D3]

D = Datacenter
B = Brocade

So my question is if we use ISL trunking on the brocade switches will it
still require additional lambda channels ?

--
Rodrick R. Brown - Sr. Unix Systems Admin
rbrown[(at)]rodrickbrown.com
rbrown[(at)]nyc.gov

"Rodrick Brown" wrote in
:

We are starting to plan out our DWDM core, one of the major apps
driving this will be our HDS 9980 and our net design group is thinking
about using FCP-IP converters with 10GigabitE backbone for our
brocade switches which will be located at different datacenters, one
of the issues they brough up which sounds valid is that we only had 32
channels/lambda and it would be very wastefull to provide all these
channels and we should look into doing ISL trunking on the brocade
switches and they will do the protocol encapsulation and send it all
over IP i'm still fairly new to the SAN world, can anyone give me pros
and cons to their proposal ?

So my question is if we use ISL trunking on the brocade switches will
it still require additional lambda channels ?

If you use ISLs, you need a seperate channel for each link, two or more
links to an ISL.

I would probably steer clear of FC over IP solutions from a complexity and
performance point of view - every time you encapsulate something, it'll
cost you precious milliseconds of latency on your storage systems.

On the up side, it'll save you money for the DWDM gear, but i don't know if
that will be cheaper when you go and buy the FC-over-IP gear...

--
/Jesper Monsted

Jesper Monsted wrote in
4.163:
If you use ISLs, you need a seperate channel for each link, two or
more links to an ISL.

That would be two or more links to a trunk (which most people should be
using for redundancy).

Damn that whisky...

--
/Jesper Monsted

Jesper Monsted writes:

If you use ISLs, you need a seperate channel for each link, two or more
links to an ISL.

I would probably steer clear of FC over IP solutions from a complexity and
performance point of view - every time you encapsulate something, it'll
cost you precious milliseconds of latency on your storage systems.

On the up side, it'll save you money for the DWDM gear, but i don't know if
that will be cheaper when you go and buy the FC-over-IP gear...


I worked on a project last summer which originally was going to use two
diversely routed DWDM links for networking and FC links. The budget was
cut and so we ended up with a pair of GigaMAN links (plugged into gigabit
switches we had/needed anyway) and four San Valley FCIP switches. Very
simple design, gigabit on one side, FC on the other. Configuration is
dead simple, took me about 10 minutes to learn from the San Valley reps
who flew in to help us get up and running. Also supports throttling to
less than 100% of the gigabit link if you want to share it with other IP
traffic.

The cost of the FCIP switches is quickly dwarfed when you look at even a
year's worth of paying for DWDM or high speed MAN links, unless you
already own the fiber in the ground and you aren't assuming its
depreciation expense. Whether DWDM or high speed MAN makes more sense
depends more on how much total bandwidth you need and your local telecomm
options than anything else.

BTW, the latency that the FCIP conversion adds, at least in the San Valley
SL700 I'm familiar with, is in the low hundreds of microseconds. The
latency of the link itself will usually from couple times that to a couple
dozen times that depending on how far it has to go and the path it actually
takes. I've seen a 50km link travel 300km through a half dozen exchanges,
you need to get this information from the telco up front before you sign
anything

--
Douglas Siebert

"I feel sorry for people who don't drink. When they wake up in the morning,
that's as good as they're going to feel all day" -- Frank Sinatra