Wireless Access

Hi,

I want to switch from single 7005 to VA and there are a few questions about seizing and licensing Aruba MC-VA-50

- I cann still use my existing AP and PFE and other per AP licensed licenses remain, right?

- Can I start with KVM and switch over to a HyperV based solution when it will appear later? (We do use massive HyperV on premise) and with the same license?

- Is there any seizing-tool for estimated traffic vs. hardware recomendations?

- Any hardwae I should avoid (Intel/Broadcom/Qlogic/Emulex/LevelOne...?) (I would prefer to place it in a existing IBM Bladecenter-H with an HS23...)

- Is it possible to buy one license for MC-VA-50 and have an active/passiv cluster together with a mobility master? (Like it is licensed with Sophos UTM, e.g.) or do we have to plan a full license for a passive MC-VA-50?

Thank you in advance - F.One

• Yes you can migrate your old licenses from your hardware controller to a virtual via MNP
• You can, but there's no direct migration path, so while you cannot take a backup from KVM and restore on HyperV, you can more or less either use the MM to just add the two (KVM and HyperV VMC) to the same node to inherit the config from the MM, or you can manually copy/paste the relevant config elements. If you need more details, please engage with your local Aruba account team and we can get you more info.
• The data sheet lists AP counts and user counts. Assume max throughput for any virtual platform is HIGHLY dependent on CPU and RAM and the switch fabric of the virtual servers and hypervisors. However, the max throughput while defined by CPU, the most we see is between 4-6Gbps. If you want/need more thorughput, nothing will beat hardware controllers with dedicated crypto processors.
• In terms of platforms, you want to focus on dedicated virtuazlied platform (don't plan on cloud virtualized serivces like Nutanix or AWS to work, even if they use VMWare or KVM). The network stack for virtualizing network switches (like the VMC) have some specific requirements in the vNetwork stack that don't work well/at all when moved to the cloud where those network fabrics are highly modified and optimized for end-device/host based network connectivity. Intel processors are what we QA to. The bladecenter SHOULD work, but again, if the IBM backplane has a highly modified network fabric, you may fine issues or it may require some tuning/playing around with the vSwitch settings or dedicating some network interface to just the VMC.
• With an MM, when you load an MC-VA-XX license on the MM, that defines how many APs are allowed to terminate on a virtual controller. It *does not* define the number of virtual controllers you can deploy. So with an MM and an MC-VA-50 license, you could have a single VMC with 50 APs, you could deploy two VMCs in a cluster with up to 50 APs terminating on them, or you could even deploy 50 VMC each with one AP. WITHOUT an MM, the MC-VA-50 is applied to the platform and you only get one controller.

Note if your 7005 is licensed directly and you are moving to MM, you would migrate your 7005 AP/PEF/RFP licenses to your MM, not the VMC. You are limited in how many times you can migrate your own licenses, after that, it will require some significant work to move them again.

Hi Jerrod,

thank you for these good informations.

You said "the max throughput while defined by CPU, the most we see is between 4-6Gbps." (This is the max that I would ever expect from a two socket Xeon system with an 10Gbit/s interface) but as far as I understood your reply I could deploy MM and e.g. 4 (KVM/hyperV)-Hosts with a MC-VA-50 license and loadbalance users between virtual MCs.

If this is possible I could hit two birds with one stone:

- troughput

- redundancy

Otherwise I have the choice between 7030 and 7205 single device with no redundancy at all. 

Thank you in advance - F.ONE

Sure, pretned you want max performance and you want 10Gbps for all 50AP combined. That's 200mbps per AP. So what you could do is have an MM, 5x MC-VA-50 platforms deployed (where each one does say 2Gbps), and you make one controller cluster and ostensibly each cluster member ends up with 10 APs each for a 2Gbps thorughput x 5 VMCs = 10Gbps. It just "costs" you MORE VM infrastructure, resources, etc. 

Hi Jerrod,

thank you very much, again.

Is there any kind of table outside where I could look for estimated performance of network througput for different processors.

Of course I know this comparison chart but it gives me no idea wether a system is suitable for at least 2Gpbs traffic or not.

Or do you can tell a minimum xeon processor famaliy?

Thank you in advance - F.One

Any Corei7 or higher above 2Ghz, assuming it meets the core count requirement, should be enough. The number of cores more than the clock speed matter (generally speaking). We don't characterize CPUMark or anything like that, that's the cost of using 3rd party hardware/software to virtualize, there's just too many factors to have any reasonably accurate numbers.

OK, thank you.

Unfortunately I cannot use the function "accept as solution" - it ends up in "authentication error" every time I try...