802.11n good enough to replace ethernet?

[Geo]

http://arstechnica.com/news.ars/pos...gh-to-chase-ethernet-from-the-enterprise.html

Apparently some people are suggesting so.

Now, if actual good 802.11n routers would get released. . . .

/me goes back to waiting for Dlink DIR-855

 

[Tim Murray]

Theoretical results mean somewhere between jack and shit when you're dealing with wireless networking. The paper that was the basis of my robotics research last year said we should be able to support 50 cameras on a single access point at 15 FPS. Any more than 5 and we were getting 5 FPS. Thanks to all the interference in the lab (from servers, lights, cell phones, other 802.11x networks, etc), we had about 10Mbps to play with, not the 30 or 40 we should have had.

 

[digitalwanderer]

Ugh, I hate agreeing with Tim but it's hard to argue his point. It ain't no replacement for a hardwire until it can prove itself as good as a hardwire.

It hasn't yet. That doesn't mean that it won't, just it hasn't yet.

 

[Geo]

I think it's a bit premature as well, but the reality is the spec is 248Mb/s. . . .and the expected real-world throughput is ~75-100Mb/s. If it hits that, it's a contender for a lot of offices. That's not unreasonable based on the 802.11g experience re spec vs reality (about 1/3 to 1/2). 802.11g is rated at 54Mb/s but I've never gotten better than 25Mb/s sitting right next to the router.

 

[Anarchist4000]

Correct me if I'm wrong but wireless devices on the same network will also be interfering with themselves. So regardless of range, if you stuck 100 computers onto the same access point they will all interfere with each other. Also consider that wired ethernet is 100Mbps, if not more, on each wire so if you had a 10/100 switch with a gigabit uplink you could be moving a whole lot more data than wireless a lot more reliably.

Wireless definitely has it's uses but in an office I'd much rather run the wires to everything that I can and then use wireless for devices such as phones and laptops where it is more convenient and speed isn't as big of an issue.

 

[Rainbow Man]

Don't wifi networks use CSMA/CD like ancient coaxial ethernet and thus have performance go down the toilet wghen many devices are connected to the network?

An entire office floor full of wifi-attached workstations sounds like a goofd recipe for rotten performance..
Peace.

 

[Otto Dafe]

I just don't think wireless will ever cut it until TCP is either phased out or upgraded, and I'm starting to wonder if that's ever going to happen. TCP was great for its time, it's just unfortunate that it's terrible for the networks we want to build now.

 

[Davros]

Plus wireless never seems to be rock solid

 

[itsmydamnation]

I just don't think wireless will ever cut it until TCP is either phased out or upgraded, and I'm starting to wonder if that's ever going to happen. TCP was great for its time, it's just unfortunate that it's terrible for the networks we want to build now.

Just wondering what is so terrible in TCP/IP for current networks? Most of the worlds networks are MPLS so i fail to see how TCP/IP has any effect what so ever seeing that is layer 2 tagging. On the side of poor performance on a host to host basis its very rarely TCP/IP's fault but rather bad implementations of windowing (on the OS) or incorrect configuration of queing and shaping on a link, then there is just plan bad protocols like SMB .

the biggest issue for TCP/IP is the lack of NAT in V6.

what exactly is TCP/IP missing or has that makes it so bad for wireless, given everything that happens in wireless is at layer 1/2 ( even QOS) and TCP/IP is 3/4?

wireless uses CSMA/CA ( collisions avoidance) not CD ( CD requires to be able to listen to the "wire" while sending) which means as well as running in a half duplex environment you also take a extra performance hit. You can only ever get to 70% utilization before you start getting collisions. so you have what 248Mbit, half that because its half duplex, so 124. Now only 70% of that is good so 86.8MBit, now take away a few MBit because of the random delay nature of CSMA/CA so maybe 82Mbit in a perfect environment to share between multi users. Hardly a Ethernet replacement yet.

Then there is the whole issue of QOS. QOS on wireless requires you to delay "unimportant" traffic a random interval within a given range compared to Ethernet where you give priority to important traffic.

Lets not start on wireless "security" About the only thing that i would call secure on wireless is EAP-TLS but how many organisations would pay the money to implement it correctly

Yay a subject on B3D that i actually know something about.

cheers