Active users per AP before performance drops
05-12-2015 12:28 PM
I am curious gang, what are you guys seeing as far as the number of ACTIVE users that it takes to really start killing performance on Aruba APs. When I say ACTIVE, I mean just that, actively passing lots of data/traffic like web or streaming or whatever. Here is what I have found through first hand and reliable users experiences:
AP60/61 - 2.5 G - funky shark fin shape not withstanding, these little guys performed well and my heaviest use cases for this series was maybe 25 or 26 users and in reality only 10 to 15 of them being ACTIVE users. As we have grown, all of these have been replaced and are no longer in action.
AP93 - 2.4 N - Several years ago on my first wave of wireless campus expansion, these little guys were an excellent and affordable way to do so. I still have about 50 or 60 of them deployed and they are rocking right along. I do have several in some labs that get hammered with about 30 to 35 active users during certain classes. I personally went and checked it out during one of these periods and noticed only a little performance difference as compared to the other AP93s on the same floor. It was a little slower but being one of our network admins here, I tend to notice that more than the average user as I am sure you guys do too. Overall I am very happy with them even though I am starting to phase them out for more current models.
AP105 - 2.4 & 5.8 N - These have been my wireless work horses through out two of our campuses. They make up the vast majority of my APs and have been fantastic. In our last new building (2 years ago) I installed nothing but these guys and have not been disappointed. However, there have developed several very high density areas in this new building. I had three or four locations that were consistently hitting anywhere from 40 to 50 active users for most of the day, Monday through Friday, every week. Now for this community college, that's quite a bit! I started to get wind of a few complaints here and there so I went and checked it out. As it turns out, the performance wasn't really degraded but there were so many active users that people were having a hard time getting our portal page to come up so that they could sign in. After several tries, they got through and were happy as clams (provided clams are ever really happy, I'll just go with it for now). I imagine if you didn't have a portal page for authentication (say open for guest or maybe 802.1x) that the 105s could easily handle 60 active users.
AP205 - 2.4 & 5.8 AC - I just got my first shipment of 40 of them for a new building that will be complete this summer and go live this fall. If I can I'll try and update towards the end of Fall semester with my experiences with them.
AP225 - 2.4 & 5.8 AC - I have recently deployed about 30 of these high density APs. I am running them on 802.3at PoE injectors until such time as we upgrade out PoE switches. 10 of them were specifically for high density area replacement for AP105s like mentioned above. Continuing the specific example from above, I replaced that AP105 with a 225 and all portal page issues disappeared. I still see the same numbers of active users (Its actually a Port City Java coffee shop in our Union Station building so it is always full of students, staff and faculty) If anything I have actually seen a slow increase int he active user base there and have even seen it as high as 67 ACTIVE users (about 80 total users associated to it) and its performance was still great. And mind you, this is only using ONE of its gig Ethernet ports.
So these are my experiences, I am interested in hearing yours!! So please share!
Network+ | CWNA | CWTS | ACSP | ACMP | ACMA | BREC
Re: Active users per AP before performance drops
05-13-2015 07:33 AM
So did I do RF scans, additional spectral analyses and try to plow through various waterfall and waveform views?
Did I break out iPerf and set up several clients and a base and run multiple throughput tests based off of artificial traffic generation?
Did I make gratuitous use of websites such as http://www.speedtest.net/ and http://www.speakeasy.net/speedtest/ to test download/upload throughput even though we all know that these tests are subjective based upon what server you choose and thus do not supply you with reliable results? (but you know we ALL still use them anyway haha)
Negative Ghost Rider.
When I got into the wireless game about 10 years ago my mentor said that one of the best things you can do is to try out what the user experience is like for yourself. Find out what your clients surfing habits are like (or a close educated guess will generally work too) and then do it. Then you have to be honest with yourself and ask “Am I happy with this experience? Did it meet my expectations as an end user? Did I accomplish what I set out to accomplish while online in a timely manner?” I took his advice to heart and it has served me well.
Now don’t get me wrong, when designing a wireless network I don’t just run around asking the wireless fairy where I should place APs and hope that they work. I do the legwork, I do spectral analyses when I feel it is needed (i.e. I would in a Downtown location but obviously not at a remote campus located in a rural area that time forgot and still has dinosaurs roaming about). I read building plans and blueprints taking notice of construction materials and their associated attenuation factors. I also work with the various departments to find out where their people will be located, what they want out of the system, where student concentrations will be, yadda yadda yadda…
So network design and post-performance evaluations for me are two different animals. And before the number crunchers who reside on a higher plane of geek-dom than I do rip me apart, I am stating for the record that I am not, nor do I claim to be an RF Scientist. I am however, a network administrator who has a passion for the wireless side of the house and I enjoy working with it.
So back to the original question of how I tested the APs and what did I use. Well, there is not a simple answer to that. After a while you develop an intimate knowledge of your environment in relation to your AP locations. This is especially easy for smaller shops like us with a couple of hundred Aps on two or three campuses and a few small satellite locations. I don’t know how the guys at Ohio State do it with their billions of APs. Anyway, I see the high density areas every day as I walk through campus. I know where the students and staff are and where they gather other than in classrooms. Also, the controllers have some nifty tools on their dashboards that help as well. In specific the “Usage” and “AppRF” pages on the Dashboard tab. So knowing my environment, knowing what my peak usage times are, and using these dashboard tools I am able to discern quite a bit. I look at the dashboards every day (I only have 5 controllers so it’s not really a big deal, again I don’t see a shop like Ohio State using this method LOL) and also at the combined statistics on my AirWave server. Admittedly, I am a little OCD about it and constantly check them throughout the day. So when an AP peaks my interest I check it out and first look at how many clients are associated vs. how many of those clients are authenticated vs. how many are actually passing data or setting idle.
As an example, while I am writing this staff and instructors are preparing for graduation ceremonies in our gym. One of my AP105s in there is reporting 57 associated clients, of those only 42 are authenticated and of those only 32 or 33 are actively passing data. By my reasoning then there are currently 32 ACTIVE users on this AP. So that is how I determine what I call active users.
Now on to performance. Like I said earlier, my methodology is all about replicating the user experience. So before I go out to see how performance is in a certain area, I check out the “AppRF” section on the related controller and also on my AirWave server. I take note of the tops applications and services which is usually Facebook, Instagram, believe it or not – Windows Updates, iTunes, Pandora/Spotify, Netflix (we tried blocking Netflix once and almost had an insurrection as the instructors rose up in rebellion swearing up and down that they needed Netflix for instructional purposes) and the beloved YouTube, and then I go web surfing after confirming that I am associated to the target AP. Then I take an honest look at how its performing from an end user standpoint of someone who doesn’t give a rats-@$$ about network infrastructure and how it all works and blah blah blah… but just wants to get their Ogre Mage leveled up so they can go on that Guild Raid to WunderStone during the two hour break they have between classes. Being an ex-gamer (the wife and kids kind of stomped it out of me) this is an easy viewpoint for me to take.
And then I judge, and I judge harshly. And then I work to correct what I see as an issue even if most end users don’t see it.
So MemphisBrothers and Andy from Pellissippi State (who emailed me), I don’t have a magic bullet testing tool that will give me any info that I want or need. What I do have is some experience, some nice built in tools on the controllers, but I mostly have the ability to communicate with my end users and understand their experience on my wireless network. And believe me; if these kids think it sucks, they will not hold back LOL
Network+ | CWNA | CWTS | ACSP | ACMP | ACMA | BREC