RE: [WIRELESS-LAN] 'Clustering' and 'failover' in the context of Aruba

2007-05-23 Thread Russ Leathe
Hi John,

We just went through the same scenario.  We now have an Aruba/Alcatel
6000 with dual controllers...200 AP's as well.   

I concluded the following.

1.) Two controllers (same box) side A  B, A-side is active, B-side is
redundant.  A-side fails, B-side takes over
2.) Major/minor upgrade - Upgrade B-side, swap, everything looks good,
do a copy/synch to the A-Side
3.) Major/minor upgrade - Upgrade B-side, swap, upgrade fails, revert
back to A-side 
4.) Minor hardware failure - 1 controller fails, still operational
5.) Major hardware failure -  backplane failure = tons of phone calls

Item #5 is my weakest link.  However, because the 6000 has
dual-everything (except the chassis), I felt my exposure was minimal.
I could not justify the cost of an additional 6000 for 100% redundancy.
I looked at the smaller boxes with decentralized distribution...but the
smaller boxes couldn't handle all 200 AP's simultaneouslyPlus it
just adds complexity to our already overworked staff (troubleshooting
would be more complicated).

I hope this is helpful.  

Russ



-Original Message-
From: John Rodkey [mailto:[EMAIL PROTECTED] 
Sent: Tuesday, May 22, 2007 7:24 PM
To: WIRELESS-LAN@LISTSERV.EDUCAUSE.EDU
Subject: [WIRELESS-LAN] 'Clustering' and 'failover' in the context of
Aruba

We are currently considering expanding our existing wireless environment
to cover additional dorms.
By doing so, we will exceed the capacity of our current controller, and
can either add an additional controller
card or for a slight incremental cost, add another controller.  We
planned to add the additional controller, with
the idea that the controller would allow redundancy/failover/clustering
to  happen, so that if one controller
were to go down, for instance, the other would take over.

We were subsequently told that this was a faulty understanding of the
failover function.
So we thought we might be able to try another approach:  every other WAP
would be controlled by alternating controllers.
That way, if controller A, with waps 1,3,5,7,9... on it were to go down,
the coverage in any given building would be halved, because controller
B, with waps 2,4,6,8 ... would continue to run.
Nope, that is a bad idea, says the contact: each controller will
maintain its own heat map and routing info, etc. and as a result, there
would be nowhere to look for a unified picture of the wireless network.

So I'm confused: what is the exact nature of controller clustering or
failover under Aruba?
Given somewhere in the neighborhood of 200 APs, how should one configure
the controllers

John

**
Participation and subscription information for this EDUCAUSE Constituent
Group discussion list can be found at http://www.educause.edu/groups/.

**
Participation and subscription information for this EDUCAUSE Constituent Group 
discussion list can be found at http://www.educause.edu/groups/.


RE: [WIRELESS-LAN] 'Clustering' and 'failover' in the context of Aruba

2007-05-23 Thread Brooks, Stan
John,

Others on the list have responded and given some good answers to your question. 
 Let me add my experience and 2 cents.

At Emory we have 1500 APs running in two Aruba systems - one for the Academic 
side of the house and one for our Healthcare organization.  Needless to say, 
our Healthcare organization demands high availability :-)

Our architecture is similar on each side - one set of redundant master 
controllers and multiple local controllers.  In the Aruba architecture, the 
masters' function is to manage overall global issues - configuration, user and 
AP lists, heat maps, IDS correlation functions, etc.  Masters can also support 
AP connections, like the local controllers.

We don't have any APs homing to our master controllers (but we could if we 
wanted to).  Instead, we home APs to local controllers.  We also have a 
dedicated local controller as a back-up, i.e., if any of the local 
controllers fail, the APs would re-home to the backup.  We have one local 
backup/system, and can withstand ONE controller failure at a time.  Initially 
this a bit pricy, but as we've expanded, we find that a single backup 
controller works very well.  In the past two years, we've only lost one 
controller (bad sup card), so our backup controllers are idle virtually all of 
the time.

BTW - it is EASY (and necessary) to direct an AP or group of APs to specific 
controllers.  In the command line, the syntax (vers 2.5 and below) is ap 
location location code lms-ip specific controller IP address.  You can 
also set the backup controller using (again, ver 2.5 and below) ap location 
location code bkplms-ip backup controller IP address.

There are a number of ways to build redundancy with the Aruba system, with the 
best way dependent on your situation.

The method you mentioned with interleaving APs to different controllers WILL 
work because of Aruba's mobility/roaming capability.  The problem arises if you 
only have one master and one local.  Losing the master will prevent the global 
functions from happening (heat map, configuration, IDS correlation, etc) and 
the loss of servicing APs that are homed to it.  Losing the local results in 
loss of ability to service the APs homed to it.

Aruba licenses each controller to support a set number of APs.  If you lose a 
controller, those APs will home where you told them to go, but if that backup 
doesn't have capacity (based on it's licensing) to handle those APs, they are 
effectively down.  That's why we use an N+1 local controller model for our 
redundancy - a dedicated backup can handle all APs on any active controller - 
but sits idle most of the time.

I realize that my ramblings on this subject may not be quite clear - so if you 
need additional explanations, or just want to pick my brain, touch base with me 
off the list.  I've gone over a number of different redundancy scenarios as 
we've built our network, and may be able to offer some useful advice.

 - Stan Brooks - CWNA/CWSP
  Emory University
  Network Communications Division
  404.727.0226
  [EMAIL PROTECTED]
AIM: WLANstan  Yahoo!: WLANstan  MSN: [EMAIL PROTECTED]
-Original Message-
From: John Rodkey [mailto:[EMAIL PROTECTED]
Sent: Tuesday, May 22, 2007 7:24 PM
To: WIRELESS-LAN@LISTSERV.EDUCAUSE.EDU
Subject: [WIRELESS-LAN] 'Clustering' and 'failover' in the context of Aruba

We are currently considering expanding our existing wireless environment to 
cover additional dorms.
By doing so, we will exceed the capacity of our current controller, and can 
either add an additional controller card or for a slight incremental cost, add 
another controller.  We planned to add the additional controller, with the idea 
that the controller would allow redundancy/failover/clustering to  happen, so 
that if one controller were to go down, for instance, the other would take over.

We were subsequently told that this was a faulty understanding of the failover 
function.
So we thought we might be able to try another approach:  every other WAP would 
be controlled by alternating controllers.
That way, if controller A, with waps 1,3,5,7,9... on it were to go down, the 
coverage in any given building would be halved, because controller B, with waps 
2,4,6,8 ... would continue to run.
Nope, that is a bad idea, says the contact: each controller will maintain its 
own heat map and routing info, etc. and as a result, there would be nowhere to 
look for a unified picture of the wireless network.

So I'm confused: what is the exact nature of controller clustering or failover 
under Aruba?
Given somewhere in the neighborhood of 200 APs, how should one configure the 
controllers

John

**
Participation and subscription information for this EDUCAUSE Constituent Group 
discussion list can be found at http://www.educause.edu/groups/.

**
Participation and subscription information for this EDUCAUSE Constituent Group 
discussion list can be found at http://www.educause.edu/groups/.


RE: [WIRELESS-LAN] 'Clustering' and 'failover' in the context of Aruba

2007-05-22 Thread Brian J David
John,
We are an Aruba shop,
The way we have it setup here at BC is a Master (with all Vlans configured
on it) and 4 local controllers (sup2's around 600 AP70's) The master is our
failover, all 600 AP's are spread across the 4 locals and the master is our
backup (no AP's on it). I had a problem with a local 1 and when it failed
about 200 AP's failed over to the master and the users had no Idea. Now it
sounds like our topology is different than yours but as long as your master
can handle all the AP's that would fail over to it you should not have a
problem. 

Brian J David
Network Systems Engineer
Boston College


-Original Message-
From: John Rodkey [mailto:[EMAIL PROTECTED] 
Sent: Tuesday, May 22, 2007 7:24 PM
To: WIRELESS-LAN@LISTSERV.EDUCAUSE.EDU
Subject: [WIRELESS-LAN] 'Clustering' and 'failover' in the context of Aruba

We are currently considering expanding our existing wireless environment
to cover additional dorms.
By doing so, we will exceed the capacity of our current controller, and
can either add an additional controller
card or for a slight incremental cost, add another controller.  We
planned to add the additional controller, with
the idea that the controller would allow redundancy/failover/clustering
to  happen, so that if one controller
were to go down, for instance, the other would take over.

We were subsequently told that this was a faulty understanding of the
failover function.
So we thought we might be able to try another approach:  every other WAP
would be controlled by alternating controllers.
That way, if controller A, with waps 1,3,5,7,9... on it were to go down,
the coverage in any given building would be halved, because controller
B, with waps 2,4,6,8 ... would continue to run.
Nope, that is a bad idea, says the contact: each controller will
maintain its own heat map and routing info, etc. and as a result, there
would be nowhere to look for a unified picture of the wireless network.

So I'm confused: what is the exact nature of controller clustering or
failover under Aruba?
Given somewhere in the neighborhood of 200 APs, how should one configure
the controllers

John

**
Participation and subscription information for this EDUCAUSE Constituent
Group discussion list can be found at http://www.educause.edu/groups/.

**
Participation and subscription information for this EDUCAUSE Constituent Group 
discussion list can be found at http://www.educause.edu/groups/.