PfSense network topology question
-
Ok, so there are a lot of replies so I hope I will not forget anything. First of all, I am a computer science student, but very new to IT. Only been doing this for like 6 months or so and nothing serious so far. I do have a list of MAC addresses. There are no guests allowed, unless they are registered in the table. We have approximately 340 devices registered currently. So I have and exact number. The problem with testing some AP to see how it would fit our need is that I don't have a budget for it. Either we buy everything once or we don't buy anything. Stupid public sourcing thing… About the heath map. I can do it with the existing infrastructure, and it would give me some better understanding of the situation. The thing with POE is if I stick with Unifi they ship their modules with POE (i think injection maybe) modules, so i got that covered. The problem with cables is that i don't want to pull a cable to every room, or even to any room, because there are 3-4 students/room and each room would need a switch for everyone to be able to use cable. I want to go all wireless, plan for the future. Things are little different here sadly :( They want people to create castles from dirt, and they expect us to do this network planning and construction with only 2-3 people. No technicians, no nothing, just 2-3 guys doing everything. So this is sad. I want ac because it will be good even after a few years, and we will not get budget for a new network only in maybe 2025... The current architecture is built in, prepare for it, 2001-2003! So... I will try to dive deeper into networking architecture, but there are too much variables and a too money-grubbing leadership.
-
And sorry if i came off a little rough, I don't want you to solve my problems, I only want you to help me a little along the way, and I also want to thank you everything you did so far, because without you I would still be thinking about routers as APs and woud not know there are waaaay better solutions than that :)
-
I want ac because it will be good even after a few years, and we will not get budget for a new network only in maybe 2025…
The physics of the bands are what they are. Time is not going to change them. I would focus on delivering solid N in 2.4 and 5 and you'll be ahead of most. All AC does is stack more channels together to increase throughput. If you want to reduce your available channels on 5GHz to make it like 2.4, 11ac is for you.
There are some density improvements in 802.11ac (MU-MIMO) in the second wave of AC products. I don't know who's shipping wave 2 yet, if anyone.
If you want to build for AC, that's your decision. Know that what you buy today will be obsolete (the old AC) in 6 months.
You might call Ubiquiti and see if they'll send you a 30-day loaner. Don't know if your project is large enough to get their attention, but worth a try.
-
Thank you for your information. If ac is not really worth is, does the 5 GHz ban worth it? And how long do you guys think N will hold on? couple of years? Will it be old in say 5 years. Because if I am not mistaken G was released in 2003 and N in 2009 and we are still supporting them. So yeah, meybe we should get a decent AP on the N channel. But 2.4 is getting crowded. At N rate in 2.4 or 5 GHz standard APs can do around 300 mbps. Some claim 450 mbps. They need AC in 5 GHz for 1300 mbps if I am not mistaken. We have a gigabit connection. My opinion now is that we should go for the 2.4 / 5 GHz N capable APs. Am I right? 5 GHz to get some stress of the 2.4 band with the clients supporting 5 GHz.
-
Also, If money is not a question, then I would only gain with buying an AP supporting AC, or not?
-
You are going to have to do 2.4 and 5 regardless. AC is 5 only. N is not going anywhere.
You have lots of decisions to make. Who cares if N or AC can to 1300Mbit/s if your internet pipe is only, say, 1000Mbits? Remember that everyone has to share that. A benefit to having faster wi-fi is reduced air time to get the same data onto the wire.
Are your clients going to be isolated from each other or be able to communicate? Do you want someone to be able to see everyone's network shares if they're both connected to the Wi-Fi or not?
And people aren't going to get that if they're not in the same room with the AP. And they won't get it anyway. It's marketing hype.
-
WiFi is half duplex …. 1300mbit/s will "only" result in around 500-600mbit actual throughput in ideal circumstances.
also the 1300mbit/s will only be achieved by using 80Mhz channels.
in europe only 5 non-overlapping-80Mhz-channels are available for use on the 5ghz band .... Only 1 of those can be used all the time, the other 4 are DFS channels and have to make way for other radio/radar interferences when detected. (= loss of connection for clients, until the reconnect).imho the current 802.11ac is never gonna be a huge success in a corporate environment, when dealing with multiple AP's .... it might be OK if you just use one at home and don't have many neighbours.
-
Yup. Build for N. 20Mhz channels on 2.4 and 40Mhz channels on 5. If you get that solid your users will be happy.
If you really want them to be happy, budget in a small, low-power ESXi or XenServer platform (C2758?) so you can host things like the Ubiquiti controller, an additional caching DNS server, maybe a documentation web site or wiki, a dorm forum, or whatever, and, of course, your pfSense (which pretty much eliminates XenServer at the moment).
Your biggest problem is going to be getting cat to where you should be mounting the APs.
Figure $1000 for the switch, $1000 for the C2758, and 10 UAP-Pro APs at $220 each. Assorted racks, jacks, cables, etc and you're looking at about $5k + labor to pull the cat.
-
Also, If money is not a question, then I would only gain with buying an AP supporting AC, or not?
You will only get full AC speed in close proximity to the AP (5-10m) with free line-of-sight and no obstacles/walls in-between. It is more likely to end up with N speed for most of your users anyway.
And AC needs a hell lot of bandwidth in the pipeline - which is air. You cannot reserve it for you exclusively. So everything using these channels as well will have a negative effect on your throughput. And you will be amazed of what's in the air competing with you!
BTW, you would need to connect two 1Gb/s cables bonded to each AP to push the full 1300Mbit/s between switchports and AP.
Forget about AC, it's neither worth it nor doable in close density installs.Having said that, have a look at the Ruckus Wireless APs.
Their ZoneFlex series uses beam-forming extensively to "shoot" to hosts, not crowding unnecessary areas.
Look at the 7372 or R300 series devices.
Have a look and some good reading at their TechTalk and White/Black Paper sections. Lots of knowledge there!Finally, you will want to "look" into your wireless domain for planning and solving problems.
I found Metageek quite useful. If you cannot afford the Wi-spy spectrum analyzer then go with their WiFi-card only option, which is inSSIDer scan. -
7372s are amazing. But they're about double the cost and with 10 I'd suggest a ZD1100 to go along with it. Yes, it would work great. A little spendy for this install. The $5k quoted above will quickly become $10k plus annual maintenance.
I'm sort of surprised OP's school doesn't have an overall wireless policy to follow.
-
Thank you for all the suggestions. I will reply to them as soon as i will have time. Probably sometime in the evening (in my time zone).
-
7372s are amazing. But they're about double the cost …
That's why I mentioned the R300 as well. It is listed as 3/5 of the 7372 at my distributor (approx. net 300,- Eur vs. 500,- Eur ) which should put it in the ballpark.
-
Dear friends!
Today was a sad day. They told me, that I have one hour to submit the list of required things for the network. I gathered all I learned from you and some pre-existing things and submitted the following list :
Server
Motherboard - GIGABYTE GA-970A-UD3P
CPU - AMD FX-8320 3.5 GHz AM3+
Memory - 4 x 2 GB DDR3 1333-1600 MHz
NIC - PCIe Gigabit Ethernet NICCables - Cat6
Main switch - 1x D-Link DGS-1100-24
Spare switches - 8x D-Link GO-SW-8G
Wireless APs - 10x Unifi UAP-PROThis will be the setup if they accept the plan. We will see soon. I went with Unifi because they offer a great all-in-one package hardware and software wise. I ditched the AC standard, as you have said, not worth it.
I hope this thing won't be a complete failure, because my neck is on the line, kind of. Please leave your opinions below, thank you :) -
My motto is always ask for what you want and have them tell you no, if they didn't give you a budget to work with then you get the best of everything and let them tell you "NO". I try to have my boss tell me NO at least once a week, in other words if you don't ask, then you will never know what you could have gotten. I don't think your list is bad I probably would have gotten a core i processor but should be fine for what you want to do.
-
Looks to me your building your own server? That might be fine in your house, but not in production setup. Why would you not buy an off the shelf server from a major player get support, and be done with it? Who you going to call if that DIY box crashes - how fast can you get a replacement on site?
-
Yes I am building my own server. I chose this option because I don't need the things off-the-shelf products offer like virtualisation, ECC memory management and things like that. I chose AMD over Intel, because the socket types list was very long and confusing for me. Plus worse money/value ration imho. Also, I have always worked with AMD in the past, and no problems at all so far. Again, I don't need the fancy computational power Intel provides with his quality core i series and things like that. I did not choose the Xeon series, because again, don't need it. We are living in a dormitory, this is not a company. People don't really care about uptime, 24/7 availability. They don't mind to have even a couple of hours downtime sometimes, because what they use the internet for is mostly Facebook, Youtube and school. Also if we have a problem, we detach a few cables, if not solved, we do a restart, if still not good, we revert to a previous state. If we have a HW problem, then no worries, we will have the current server as backup server. Back to the processors. I like AMDs raw power. 8 core. I know Intel has 8 threads, but still, not the same thing. also we have 3.5 GHz and we are not calculating floating point numbers. Off-the-shelf products are expensive, and if I spent the money on the things I don't need I would not have money for the things I do. I live in the dorm where the server is, so maintenance is not a problem. Mikeisfly, I did exactly that. We will see if they say NO. Lets hope for the best. (For the record, sent this from a core i7 laptop xD)
-
The biggest problem I see is that your processor is burning at 125w, which will make it HOT and quick to fail should you fail to clean it very regularly. Dust will be your enemy. To mitigate that, I'd put a huge heat sink on it the size of a boat anchor. That should make it reliable even with a bit of dust (There will be dust)
Other than that, should be fine assuming you don't run into any compatibility issues.
That being said, I'd have gone with a more purpose built server board. it wouldn't have really added much to the price.
However, you should do well with this if you take steps to keep that CPU frigid.
-
We did order a huge ass fan and heat sink for the CPU-and also a few coolers for the box and finally, we will have cooling on the "tower" itself.
-
And good quality thermal paste
-
Sounds like you should be good to go for a few years.
