I am trying to study the CWAP book and understand the concepts and have a quick question:
How does the larger data frame size in 802.11 affect going through the AP to an Ethernet network?
If I was FTPing an 2000 byte packet across an AP to an FTP server on a wired network, who/what would do the fragmenting to make this an acceptable Ethernet?
Would it be the Ethernet switch connected to the AP? That would be my guess, but I am not sure.
Or would the AP be smart enough to recognize that this packet would need to be fragmented into smaller fragments and then sent out onto the Ethernet wire?
The maximum frame size for Ethernet (ignoring Jumbo frames) is 1518, so any frame larger than this size must be fragmented where the frame goes from the wireless to wired network - that is at the AP which has the Ethernet interface.
For an 'official' explanation see P295 of the CWNA study guide as corrected by the published Errata document.
HA! Nowhere in the 802.3 specification does it talk about fragmentation. Common misconception. If you did as you suggested and sent a 2000 byte frame from a WLAN node to a wired node, the AP would drop the frame. The AP, upstream switches, and the wired nodes would have to be capable of supporting Jumbo Ethernet frames (up to 9000 bytes if memory serves) in order to pass this frame.
but Devinator, re the quote I made from P295 of the CWNA book. OK the AP will drop (not fragment) any frames > 1500 bytes unless jumbo frames are supported. But the book goes on to state that IP would fragment any packet >1500 bytes - presumably it does this on the wireless as well as the wired side, so the frame size question is moot (more or less) i.e you cant send a frame of 2000 bytes?
...Or maybe I am in for another HA! :-)
the IP layer can fragment, but Ethernet doesn't. The default MTU of IP is typically 1500 bytes. This makes the whole conversation moot....however, if, under some very weird circumstances, you had an IP packet that was >1500 bytes but <2346 bytes, then it would still fit into the 802.11 MAC frame but not be passable by the AP onto the Ethernet segment. It would be dropped by the AP.
Another way of saying this is that 802.1D bridges between 802.3 Ethernet and 802.11 WiFi can not fragment packets at layer two, while IP routers between Ethernet and WiFi can fragment packets at layer three but at a significant processing cost.
Many access points contain both a bridge between Wi-Fi and Ethernet, and a router between the bridge group and a "WAN" Ethernet port.
To facilitate bridging with Ethernet and routing with Ethernet, WiFi clients typically use an IP MTU of 1500 octets, much smaller than the 2304 octet maximum MSDU of WiFi.
Yep, that's another way of saying it alright. ;-P