Roaming between WiNG switches

// Expert user has replied.
I Ian Jobson 3 years 4 months ago
0 4 0

All, I have a customer with WS5100s installed in store using L2 connected AP300s. They have MC9090s and are finding that when a terminal roams from an AP on one switch to an AP on another switch the terminal will do a DHCP request. They currently use version 3.0.2 but we have tested with 3.3.1 and we see the same issue. This has not been a problem before but they have recently deployed a new application which has not been written very well and is not handling the extended outages (sometimes 15-18 seconds) we are seeing on a roam. I would expect that when a terminal is roaming within a cluster of switches we wouldn't need to undertake a DHCP request when we go from one switch to another. We aren't moving subnets or networks, the only things that change are the AP and the switch. The real problem here is the application, however, I need to get an understanding if the DHCP behaviour is as expected. One other factor that may be causing the excessive delays is that they apparently have 2 centralised DHCP servers that are both active. But whilst this may cause issues or contention, surely this would only be the case when we do a request ... should we actually be doing the requests is a whole different question. Any one have any thoughts? Thanks IJ

Please Register or Login to post a reply

4 Replies

I Ian Jobson

Thanks Alona, That's kind of what I thought, my concern was whether the network in some way was not appearing as a continuous, single network and was therefore making the terminal believe it was roam between two networks and therefore would require the DHCP process. IJ

F Frank Barta

It is the expected behavior of the MC9090 to send a DHCP request on every roam. The reasoning is that we can not guarantee that a device which roams between two APs will be on the same IP subnet. The same is true for all of our Mobile Computers.

a art gabriellini

Furthermore to Franks point; once the MU has obtained a leased IP, the Microsoft stack caches this address so it may issue the renewal request @ roam. If the server responds favorably with a DHCP ACK, then there should be no further IP exchanges until the next roam where you should only see the IP request (renewal)... Obviously if the server issues a NAK to the clients' renewal request, then this forces the DHCP discovery stage, in which case if everything goes as expected, the client receives the DHCP offer, client issues the DHCP request (permission request to use the offered IP) where the closing return from the server is with a DHCP ACK...(RFC 2131 I believe) Any deviation from the above would warrant Netlog (not wireless) traces to expose root-cause, but if the application is forcing control over IP management from Microsoft (which I have seen quite often) then that is typically where issues arise. 

A Alona Gian

Ian, Switch can hardly cause MU to send DHCP request. That is something MU decides on it's own. In most of the cases - MU should not be renewing IP address on roam. It can renew - if it roams between WLANs or bands. I would suggest collecting a wireless trace and talking to MCD to understand why MC9090 behaves that way. Alona

CONTACT
Can’t find what you’re looking for?