Network Management

I reached out to my Aruba Sale Rep today for a Airwave demo. After looking at the data sheet and solution guide I have a few questions that I hope someone can fill in a few of my blanks...

 

A little back ground, I manage about 3,000 Cisco access points and 20+ Cisco WLCs. We have about 6,000 wifi devices concurrently on the network. Looking for an alternative to Prime and WCS.

 

After looking at both the deployment guide and the data sheet I was unable to determine exactly what the actual server limitations are. I then came across the hardware data sheet, where it states Pro Appliance 1500 devices / Enterprise 4000 devices. Since I have under 4,000 managed devices I would assume 1 Enterprise server should meet my needs. I am however concerned about the lag if the server is pushed to 4,000 devices. Any feedback ?

 

I need an explanation of part number AW-EXP1-2500. I understand from the p/n description this is a +1 device licenses. For example, if I had 2,600 devices, would this mean I would need 100 of p/n AW-EXP1-2500?

 

Im curious if there is any data to support how much traffic will be generated across the network to support 3,000 aps, 20 WLCs and 6,000 clients? 

 

HA is important to our organization. I don’t see anything in the data sheet or solution guide that talks about HA support specific to licensing or how it operates. Can someone point me to this info ? 

 

Does Airwave communicate with the APs directly in anyway to poll information or just the controllers ? 

 

How often are support updates released for Airwave ? 

 

Central vs Distributed Deployment .. Any suggestions on deployment strategies; 6 separate locations need to be monitored. Today each location has a WCS. Any feedback or advantages using Airwave in central vs distributed ?

These are great questions.  Let me see if I can answer them all.

 

You cited the APs and controller count, but will you also be adding in switches?  The scaling numbers are testing with 20% of the device count as switches (this was determined to be a median value for the number of switches).  The tests also account for continuous client traps and syslog messages per second (Table 3 of the 7.6 server sizing guide).  If you do include switches, they will also be counted in the device count.

 

Your concern for lag is valid, but keep in mind that as long as the APs are behaving as thin APs instead of autonomous APs, AirWave will primarily contact the WLCs and switches for data, so as long as the controllers are not overloaded and responsive, there shouldn't be much lag except for time of SNMP poll to time of SNMP response from controller.  You also have the option to override the polling cycles to stagger the data collection depending on your data priorities.  And SNMP traps will add additional data points to provide more realtime client tracking.

 

When it comes to licensing: for 2600 devices, you'd purchase a 2500 device license + a 100 device license, and then consolidate the 2 licenses on the licensing site (which should be fairly easy, but I've never tried).

 

I don't currently have direct data to describe the traffic over 3000 APs through 20 WLCs for 6000 clients, but a customer in the education space with a deluxe-sized single server AirWave has roughly 8000 devices, 10000 clients pushes an average of about 10-15 Mbps.

 

In regards to HA, we have a Failover solution currently, but it's a passive solution.  This will be changed in the future as there's research being done on new architecture that will improve the HA design to make it an active solution (instead of reactive).  The user guide should have some information pertaining to the Failover.  If there's anything unclear, you should be able to reach me over email (I'll send my email address in a private message).

 

AirWave will directly communicate with the controllers.  There's some light ping tests to all devices, but primary traffic will be SNMP, and then scheduled firmware upgrades are done over FTP/TFTP.

 

Major releases are twice a year.  Each major release is followed by patched releases that are typically in ranges of 2-6 weeks apart.  Patched releases are typically done based on customer needs.  To give you a better idea, 7.6 was released 3 months ago, with the current release AirWave 7.6.3 having followed 7.6.2 after 6 weeks.

 

Several existing customers do a central deployment (this works as long as the communication between locations doesn't incur too much additional latency).  Central deployments are advised if you're the solo super-IT.  If you've got a team, you may go the Master Console route where you have a Master Console watching smaller AirWave servers to yield a larger network view, while site admins take care of individual distrubuted AirWave servers.

 

Let me know if you have any additional questions.

Thank you for the quick response. My SE got back with me on the licenses and the size guide. Do you have a like for the OVF, we are going to run our demo on a VM. 

 

Thanks

The virtual image in OVA format.  For 7.6.3, it's located here:

http://support.arubanetworks.com/DOWNLOADSOFTWARE/tabid/75/DMXModule/510/Default.aspx?EntryId=10740

 

Direct link:

http://support.arubanetworks.com/DOWNLOADSOFTWARE/tabid/75/DMXModule/510/Command/Core_Download/Default.aspx?EntryId=10744

 

To view and download, a support account with access to 'download software' section of the support site is required.  If you don't already have an account, the SE or support should be able to set you up with one.

Excellent .. thank you 

Hello,

 

I would Like to Add Some Question.

 

What will be architecture for Airwave to manage 8000 to 9000 Devices?

 

How many Airwave DL360 Ent. Will appliance be there?

Will there be any Master Appliance which will manage Multiple Appliance? If Yes,  which is this Appliance?

 

Will wait for your responses.

 

Regards,

Prashant

@prashant

 

Have you discussed the same with a sales rep?  I'd suggest that'd be a good starting point, because in order to respond - there's additional questions that would need answering:

 

1) Are all 8k-9k devices in a single location or spread out?

2) What device types are we talking about here?  Switches, controllers, thin APs, Aruba Instant, all Aruba branded vs 3rd party device types?

3) What's the reporting requirements?

4) Timeline to deploy NMS?

 

Having this data, I can then know if I'm targetting trying to define a solution based on what's available now versus what might be available in a year.  I'd also have a better understanding of how I can split the device count.  If it's all Instant, I might even suggest looking into Aruba Central.  This is the exact conversation that the sales team deals with often - it makes them a great resource to leverage.

 

Dear Rob, Here are the Answer to your Question: - 1) Are all 8k-9k devices in a single location or spread out? ==> there are Install in Single Location. 2) What device types are we talking about here? Switches, controllers, thin APs, Aruba Instant, all Aruba branded vs 3rd party device types? ==> Switches, Controllers, Controllers Managed APs and Some Cisco Core and Distribution Switches. 3) What's the reporting requirements? ==> By Email and SMS. 4) Timeline to deploy NMS? ==> It will be deployed in 3 to 4 Months. Hope you'll have Answer. Regards, Prashant Patel

Given the timeline, I'd say you're probably looking at something like:

 

3 AMP Enterprise - I'd estimate a split of 3k per AMP, this also leaves some wiggle room if needed.  I wouldn't fully load 2 AMP Ent, and have an AMP Pro, because maxing out the setup doesn't allow for new features introduced in future releases - since new features have possibility of incurring additional resource hit.  As you begin - I'd start by dividing up the controller/AP count, then add the core/edge switches that are aligned with those devices - that way upstream/downstream port info is fully utilized.

 

1 Master Console or Glass - sales person can direct which is the better options (there's some more questions to be asked about high level views).  The difference is Master Console is solely a reporting tool versus Glass which is reporting and supports elastic search across the data of each AMP.

 

1 Failover Enterprise - failover should be sized the size of your largest server, if redundancy is not required - then this can be scratched.

 

 

Take the above estimate with a grain of salt.  The scenario could be that each server needs a 1 to 1 failover, in which the number of failover units would have to increase.  Or maybe having a network wide view of everything isn't as important, so you might skip the MC/Glass piece.  If you need to maintain historical data for say half a year, you might have to increase disk space.  Perhaps your current budget can only allocate enough to acquire 2 servers at the moment - the Sales Engineer would be able to help develop a deployment plan.  Keep in mind, my estimate is based mostly on the number of devices with some influence from the answers, the sales guys are the experts in generating the specs for a deployment - and probably have a lot more questions.  I'm somewhat removed from realistic scale testing to the point I have an Enterprise class server for 10 devices running functional tests.