It’s been almost 2 weeks since BarCamp and the final figures and bits and bobs of data are finally trickling in.

There were roughly 170 BarCamp attendees.
There were over 275 unique mac addresses at the event.

That means each person on average had approximately 1.6 devices on the wireless network. The busiest time was between lunch and dinner on Saturday.

Heres the bandwidth chart for the weekend:

It’s interesting to note that you can directly correlate the peaks and valleys in the bandwidth chart with the happenings at barcamp.

A little background: I’ve been doing networking since roughly 1993 – where my first network was two 14.4k modems hooked directly together, and ‘ata’ was run on both at the same time creating a 14.4k baud serial connection. We played descent that way! (it was awesome). Next was coaxial networking cards (token ring!) to play doom 1. Then shortly after that cablemodems came around and people were using ethernet. I’ve spent a long time troubleshooting networks and doing so has taught me well. A quote from the matrix comes to mind:

I don’t even see the code anymore.. I just see … blonde… burnette… redhead..

Except for me replace ‘blonde, burnette, redhead’ with ‘Streaming the game, uploading video, opening a torrent application’ or ‘people go to sleep, people wake up, people have lunch’ etc.

This time at barcamp we used two different wireless hardware vendors. Aruba and Xirrus. The hardware aruba gave us had multiple access points, and the hardware xirrus gave us had one huge one. We used the aruba hardware on saturday and the xirrus hardware on sunday.

Saturday we had the majority of everyone who came to barcamp in attendance, as is the norm. Because of udp/gaming/filtering issues last season I completely opened up the firewall, disabling all protocol filtering and traffic shaping. This was good for performance, but ended up hurting performance as the teensy pentium 3 firewall we currently have herding all the traffic started to strain under the load.

Add to that the dilapidated hardware which would often reboot if you touched it wrong, and you end up with a mostly-stable network “so long as you don’t breathe on it”. The firewall being the most fragile.

Here are some interesting things I was able to point out:

I could tell when people started streaming the football game:

It was clear EVERYONE was online:

I quickly found out when my packet filtering rules began working:

I could tell who was in what talk! :D

As for the wireless performance results that everyone was looking for, I can’t in good conscience say that the test was fair at all. Aruba gave us 3 ap125’s, which we used two of (each ap125 has 3 radios in it: A, B/G and N), and Xirrus gave us one giant radio(which contains I think 12 radios, 4 A, 4 b/g and 4 N).

Here’s the short and skinny:

The aruba radios were able to successfully cover all of the grounds with no trouble, just like last season. The Xirrus array had some trouble penetrating the multiple rooms and walls. Every ’sharp corner’ in that venue is made with a thin bit of metal. That being the case, if one were to take an xray of the venue, one would see lots of little lines floating around in the air – interfering with radio transmissions. So for this venue more radios would do better than less radios, simply because of the width of the place and the saturation required.

The Xirrus array though, was quite cool and people gathered around it to take pictures. I got reports that “the internet is faster” on Sunday, but that was easily because only 1/3rd of the people were there on Sunday compared to the day before – the bottleneck this time was the firewall being overwhelmed. Next season we’ll have a shiny new netbook as our firewall and we’ll be able to greatly increase our capacity (and likely not have to deal with scaling for quite some time).

As far as a ‘comparison’, it was unfair in Arubas favor because we had more radios from them than we did for Xirrus, but looking at the map after the fact, its quite clear that one Xirrus array should be able to out perform two Aruba ap125’s without breaking a sweat.

Overall:

• Aruba: The hardware is great, provides awesome coverage. Each ap125 has 3 omni antennas in it, each attached to a radio – A, B/G and N. They’re quite buff! The software, I’m afraid to say, sucks. A lot. The web gui used to admin the mc200 we have has code problems and doesn’t work in Chrome or Firefox – I have to use safari or (god forbid) Internet Explorer to even be able to navigate its UI. There is no way to change the IP address of the thing without ssh’ing into it, and changing the SSID, encryption type, captive portal or anything like that is abysmal – page elements do not load, or do not refresh properly, or contain the wrong data. The hardware I have is in my opinion “Set it up once and NEVER TOUCH IT AGAIN – for risk of having to scrap the entire config and start over”. Also, there is limited functionality as the ‘good stuff’, as it were falls underneath a more expensive license than we were not given with the hardware.
• Xirrus: The hardware contains four quadrants – each with a certain number of radios – I’m still not quite sure how many, or if Xirrus counts an ‘A’, ‘B/G’ or ‘N’ as separate radios. If they do, the total is something like 12 – each of the four quadrants supports A, B/G and N, and there’s an omni in there as well for monitoring. These are all directional antennas as opposed to the omnis that are in the Aruba gear.The xirrus array right out of the box provides everything – all the potential functionality that the hardware and software are capable of. The web UI works in every browser, and one of the cooler perks is that the 802.11 triangulation works with zero config – you naviage in the UI to the right page, and you can see a 2d map of where clients are in relation to the array. Aruba won’t let you do this unless you have a floorplan – which I DO have (see above), however I didn’t want to risk bringing down the wlan for the sake of a pretty screenshot. Another thing that is clearly evident is that the xirrus directional antennas penetrate farther than the aruba omnis in any given direction – nearly 2x as far.

Conclusion: Despite the frustration in dealing with arubas UI, the multiple-radio solution works better for BarCamp since the venue is essentially a long rectangle. However – if I were to get my hands on another identical Xirrus array and lay them out the same way I laid out the Aruba gear – I have a very good feeling that the xirrus gear would outright spank the Aruba gear in performance.

As far as the rest of the event went – I heard nothing but good things! During the opening ceremonies that DivX was happy to record I’d say easily 1/3 or half of the room said that this was their first barcamp! Our community is growing A LOT. I heard nothing but good reviews about the food and the event – Intuit gave us the thumbs up that from their perspective everything went well. Stay tuned for an announcement for when we’ll be doing it again!

There a TON of pics on flickr of the event as well. Consider signing up for the mailing list to stay on top of all the future announcements and chatter during the off seasons!