SUMMARY: What's the point of switches in Arbitrated Loop?

From: NetComrade <netcomrade_at_bookexchange.net>
Date: Mon Nov 29 2004 - 15:33:48 EST
Thanks to all who replied.. Based on the following, perhaps will research
Brocades, althought I thought the Qlogic switches had better support for
Sun hardware, given that they make fibre cards for Sun.

Here are the replies I got, original email below:

"Buy the Quickloop license from Brocade (or whatever your switch
vendor's equivalent) and it sets up "F-CAL alleys" in your switched
fabric.

Personally I bought used Sun 3510s and ditched all my A5200 - the
annual maintenance was too much and 1Gbit devices are too old and
slow."


"
1.) Managed switch, hubs often don't have any management.
2.) TL mode. Most switches will append the proper field so that devices
speaking fcal can talk to fcsw devices if so desired.
3.) Transceiver quality. A5200s only support older gbics. With an sc-lc
cable, you can use higher quality and newer sfps that can both hear a
lower db signal, but also clean it up and boost it to the next device.
4.) Diagnostics. While switch info is limited due to the time they have
to ship a frame, they can give you some interesting info on what is
going on.

Basically it gives you a more controllable, cleaner environment when you
have many more configuration options.

Point-2-point leaves you to the fate provided by the hba and the gbic in
the A5200.

I would stick to the switch.

And hold on to the A5200, it's one of the greatest jbods around. Sun
last good array some would say ;-)
"

"
a switch can give you masking.  You can expose specific LUNs to specific
Hosts.

let me expand on that last mail I sent you.  If you have a disk array that
is capable of creating LUNs,  IIRC the 5x00 series are simple JBOD.  The
reason you are likely seeing confusion on the hosts, is when you put vxvm
on
the Linux machine and do the vxdctl enable it goes out an probes the disks
it can see.  This probe involves a bit of writing to each disk.  Since
these
disks were already present on the SUN, the Linux vxdctl enable will
overwrite the bits written by the SUNs vxdctl enable.

To avoid this, you need to be able to carve LUNs and mask.

As a side note I think that the 5200s are nearing the end of their support
cycle.  The 5000s went unsupported a couple of months back.
"

"
You need to step up to a more current switch technology, look at some of
the low
end Brocade switches and they will provide the functionality that you
need.
"

"
I'm not quite sure what you're asking.  What do you consider a
"drawback" and what did you expect the switch to provide?  The only
real advantage a switch (versus a hub) gets you with an FCAL topology
is greater bandwidth.  You'll still have to deal with LIPs, since
there's no fabric name-server or the like.

BTW, I've been informed (by a Qlogic engineer) that the newer SANbox2
and similar product firmware revisions don't support private loops,
due to some sort of software licensing issue.  To get FCAL private
loop functionality, I had to put the older StorEdge T3s and QLA2100
HBAs onto a Brocade switch to them to see each other.

I've never heard of "stealth mode" for LIP before.  That may be some
Vixel-specific FC enhancement.  Or, I may just have lead a sheltered
life.

The real advantage I see with switches involve fabric mode.  With a
fabric, no interruptions occur when you add/remove a device, nor do
device IDs change when you need to move them from one port/switch to
another.  Gotta love WWNs.
"

"
for many-to-one connections.

1 HBA attaches to 1 switch port. the switches can route the FCP traffic
from
multiple device ports to that 1 port
"



On Wed, 24 Nov 2004, NetComrade wrote:

> We are trying to preserive the life of A5200's.
> 
> We've setup a test environment which involves 2 Sun hosts, 1 Linux host
> and 1 A5200 and 2 SanBox 8's. While setting it up (I was not the person
> doing it), I discovered that the switches don't hide any of the drawbacks
> of Arbitrated Loop. I thought that each port would become it's own
> loop. Kinda of like described here:
> 
> http://data.fibrechannel-europe.com/technology/whitepapers/wp28.html
> 
> However, it's not the case.. So what's the point of switches? The disks
> still seem to cause a LIP, maybe the switch is only smart enough to direct
> IO to a particular port, where it knows the disk is.
> 
> Also, once we attached the Linux box and installed Veritas on it, VxVM on
> Sun boxes lost touch with about half the disks.. I highly doubt anyone has
> attempted this craziness, but if you have any ideas on how to preserve
> A5200's w/ more or less painless migration (to Linux boxes), I would
> appreciate your feedback.
> _______________________________________________
> sunmanagers mailing list
> sunmanagers@sunmanagers.org
> http://www.sunmanagers.org/mailman/listinfo/sunmanagers
_______________________________________________
sunmanagers mailing list
sunmanagers@sunmanagers.org
http://www.sunmanagers.org/mailman/listinfo/sunmanagers
Received on Mon Nov 29 15:02:19 2004

This archive was generated by hypermail 2.1.8 : Thu Mar 03 2016 - 06:43:40 EST