Cisco Wireless Best Practices Flexible Radio Assignment (FRA)

what we recommend or what I've been recommending to customers is you know unless most seasoned administrators are not going to let something loose without having at least a reasonable understanding of what it's going to do to their established infrastructure right so the question is how do I evaluate on my infrastructure how f ra is going to impact it now that I have 38 hundreds up the answer is slide three okay you want me to go to spy three or you want to go through all the best practice recommendations well I know through the best practice recommendations first okay so first thing is put an FR a in static mode on the aps and globally enabled what this is going to produce is it's going to give you the coverage overlap factor for each thirty eight hundred twenty eight hundred that you have deployed and it's going to do that passively without changing the role of the AP so in other words if you're deploying to point four and five gigahertz today you put in thirty eight hundred now you want to you know start playing with this FR a do it safely first take the aps out of auto mode there is a command at the command line to do this and then you can look at the results of what FR a thinks of your infrastructure before you commit to making changes highly recommend that FR a sensitivity high ninety percent why it's been very conservative FR a is extremely conservative by nature and we've had people that leave it set to low which is a hundred percent coverage and then they get you know a bunch of ninety to ninety eight in our experience there's use cases where even seventy five percent coverage may be well enough for instance an AP on an outside wall that you're only using 50% of the circle you can probably flip that to dual five gigahertz if it's seventy five percent covered FR e interval this one gets a little confusing for folks sometimes it has to be equal to or greater than the dca interval the minimum is one hour so if they're using a DC a timer for every eight hours the minimum FR a interval is going to be every eight hours now it's not incredibly important to run FR a every single hour the reason for that is because FR e is going to establish what interfaces are redundant we're going to put them into service right and then after that fre really shouldn't change that much so a couple of conditions that can come up for you are one you see a coverage hole in 2.4 on one of the aps that is contributing to the redundancy of a dual five gigahertz cell meaning I've got three ApS that I counted on for covering my two point four of that cell one of them experiences a coverage hole that will revert the AP back to 2.4 gigahertz to fill in that coverage hole got it got it the other the other condition for fr a running is if for instance you switch to 80 megahertz bandwidth and we've seen this in DBS it may be that you were okay running dual five gigahertz with 40 megahertz channels but when it flips to 80 megahertz bandwidth on one of the other interfaces there is no longer room to run dual five gigahertz interfaces because of that hundred megahertz separation requirement in that case EAP is going to yet and that case DA P is going to be flipped over to a monitor mode and then it will be evaluated at each fr a run for how long it's been in monitor mode and whether or not there is now room to move it back to five gigahertz so if your f ra interval is set to eight hours it's going to be eight hours before we evaluate that make sense so the recommendation is the modern are or shorter yeah an hour and I would leave DCA at ten minutes unless there is a very good reason why-y-y you've changed that got it alright client network preference at default what this says is that it prevents an fr a radio an ex or radio from changing roles if there's three or more clients already on the 2.4 gigahertz radio now this is good in a production environment but you may want to override it in a lab environment - the connectivity profile the connectivity profile will allow an immediate change regardless of the number of clients on 2.4 so we recommend default neighbor discovery packet factor this is a very very important piece alright neighbor relations is extremely important to Fr a ok and there was some some changes done to the default pruning timer or retention value for MVP messages that we here that was done in 8.0 in 8.1 we have the ability to change that to increase the timer there you go in 8.0 mr4 which is not a factor for for this build but what we have found is changing the NDP factor to 20 increases the saturation and the stability of neighbor messages the default being five now says we're only holding on to those messages for 15 minutes highly advise anybody who's going to turn on fr a2 prior to that at least an hour before that changing that NDP factor to the default from the default to a factor of 20 and that will ensure stable neighbor relations and a good fr a run and lastly very very very prudent DCA of course over time can settle in with some less than optimized changes highly recommend running a DC a recalibration on the first run of Fr a what that's going to do is it's going to run fr a every 10 minutes followed immediately by DC a running every 10 minutes DCA is going to optimize channel assignments and fr a is going to get better so is there any configuration for this yeah that's in the slide okay let's go to the rest yep so go to the first slide with the detail after this all right here's the config advanced fre revert all the static command what that does is it puts all the ApS in manual mode F are enabled so that you can calculate the coverage overlap factor then under monitor dual band radios you can see the coverage overlap factor that's being assigned and you can identify individual radios the full command is at the bottom of the slide here configured vance FR a revert all implies all xor radios auto implies only the ones that are renato then the second input is whether you want those radios to be static or ATO okay next slide this is showing what the radio role assignment just to clarify that is an under radio role assignment you can have Auto meaning that it is available and go ahead and click your I think this is a build-out slide you might want to get rid of the builds yeah Auto default makes the radio available to fr a manual takes the radio out of global fr a assignment meaning frh still running it just doesn't have the right to assign this radio being inclined serving 2.4 gigahertz puts that interface in the fr a calculation if you put that five gigahertz there is nothing it's no longer part of the fr a calculation for we're done the radios make sense yes stop sooner because you are talking about mainly managing the 2.4 gig only right so precisely for right how much of already 2.4 coverage you want and are there and then calculate on the basis of it that do we need enough of radios on 2.4 can we turn them off so absolutely the client serving has to be 2.4 because the whole algorithm is to see how much of 2.4 energy is there and and make sure that we can turn off certain 2.4 radios etc right absolutely correct once it's been assigned a role it's the domain of DCA to decide whether it's good in that role or whether it should be in a monitor mode and the priority is always going to make dual 5 gigahertz first and then monitor mode if we can't fit that perfect okay perfect next slide all right go ahead and this is a build-out slide - but your controls of the configuration and you might want to just skip to this last slide if you're not doing the builds but this is showing flexible radio assignment at the controller on the GUI first enabled sensitivity by default is low we're saying make that high and leave the interval one hour so the default is going to come as low but you're recommending it to make it high as the best practices right I am you get for this a page for the evaluation point unless you're making active changes it can be set to whatever it is in production we're recommending hi we found that that high gives us high gives us a good saturation we have good coverage at 2.4 we validated that at Georgia Tech and other locations in some cases you may want to go back to low or even override some of the assignments because the high density requirements but that's that's an advanced subject and you really have to know what you're doing there sorry okay alright well these cover here yeah this covers the client Network preference which is the new command in 8.2 and this basically is the personality update for DCA the default is no change from current DCA behavior from current TBC behavior from current DBS behavior or DFS EDR and behavior however FR a defaults to the connectivity profile the connectivity profile puts a bias on keeping clients connected doing less changes being less sensitive you follow me so this automatically puts DPA and to a higher bias on current channel meaning we tend to want to stick to the channel rather than move for temporary things definitely prefer current channel when voice and video are involved and lower the sensitivity for for DCA now you can override this in DCA just setting this command if you only want DCA to still run a medium you can go to the GUI on DCA and select medium and it will override that part of the connectivity profile DT PC should not reduce power when weak signal clients detected so it's it focuses on continuing to keep everybody connected DVS again less volatility there it wants to put lower bias on the client type meaning it's not as prone to changing flex DFS CD DRM priority are maintaining current channel voice video client increased contribution and then no band role switching where more than three clients exist that's the production mode saying if we determined that that interface is redundant then F RA or DCA would wait until that interface had less than three clients to make that change on a DCA interval that makes sense yes it makes sense so this is one of the other best practices what would you recommend in a long with the fre with the previous configurations which you have shown right absolutely this is the default profile is the best for an unknown environment in other words a production environment until you get a real handle on what f ra is going to do if you go to the throughput if you go to the throughput profile note that regardless of how many users you have on 2.4 gigahertz it's going to flip the 5 gigahertz which is not the end of the world by the way because most of the 2.4 gigahertz clients in most environments are going to be dual band capable and they're going to reattach to the same 5 gigahertz cell it's just not going to be without disruption but it should be a very very low disruption and you know administrators have to evaluate that all right you want to go to the next slide this covers an neighbor discovery protocol resolution and the NDP timer new command introduced in 8.1 is the NDP or the neighbor timeout factor by default up through 8.0 this was one hour the reason for this is that NDP is very much a probability argument we don't have free rein to transmit DFS channels so we tend to have less resolution there what what tends to happen in a dizzy environment is that NDP gets missed or it gets collided or it's not received due to RF conditions in that case you've got no option but to extend the buffer or for how long we consider a neighbor valid right the downside to that is particularly where you'll see it is in prime if you have a predictive heat map time is going to hold that predictive heat map for that ap until it times out of the NDP pruning timer so that's why it was moved back to 15 minutes interesting to note in 8.2 mr 4 and beyond this is going to be returned to the default value of one hour it has actually caused quite a bit of instability to go to 15 minutes so a factor of 20 restores the legacy function that's our recommendation to run and it was fine for everybody all the way through eight dot oh by the way okay okay eight dot in 8.4 8.2 ml for you expect this to revert back to one hour right I'm assuming they're going to put it into the EMR I know for a fact it's the decision in 8.4 okay got it okay and then DC a recalibration this is what we used to call restart however I'm calling it a calibration now because that's a nicer name see a start-up mode enables an aggressive channel search removes customer configured hysterics as' in other words ignore sensitivity we have another buffering factor called NCHS F check that insures that a change is not disruptive for a group of neighborhood ApS in other words ApS that can hear one another if you make a channel change they should all get stay the same or get better that checks removed lower internal dampening there are several things that we do to make us not twitchy that's removed while we do a start-up mode and it's much more than simply changing sensitivity to high on DC 87.6 and before DC a entered startup mode on a reboot of the RF group leader so anytime you rebooted that RF group leader you automatically did a DCA restart above that they have changed behavior 8.0 and forward it's only invoked on a reboot if there's been a code upgrade so this is important to get out there because a lot of people assume when they reboot you know if they have an 85 40 or 55 20 as their single controller there's no concern about who the RF group leader is right yes they assume that if they rebooted the controller they're doing the dca restart and that's not the case by the way this is also the deck that I showed to Gregg what we're talking about automating the dca restart yep that's got some traction we're going to be working that into a proposal and getting it into 8.4 because it's just this is a lot of stuff to explain Sujit it should just be done when it needs to be done so this relies come on is you are suggesting to off that you have done called other tests if you explain who are with this command or or end your configurations with this command yeah I'm going to get click your body anytime you're adding or subtracting channels any time you're changing band let's go ahead and kick back a minute previous slide there you go so you see the box that popped up to the right anytime you're adding or subtracting channels changing bandwidth adding or subtracting a piece from the environment change AP models or you just have long time and steady-state meaning you know for a fact it's been a year since this has been done pick a weekend kick it off it won't be it won't be that disruptive but it will optimize the channels and it will ignore all the dampening perfect I think this is something we need to start communicating with a lot of customers and feet so that because this is something you have been seeing a lot not happening and it makes a difference when we do this the first thing I do the first thing I do in and I'm I tell you what I'm going to do another deck about RRM tune up like before you freak out about RRM do these things right and this is a big one okay now a little bit about F RA and coverage overlap factor 2.4 gigahertz radios that are members of the same ap group are going to be calculated together now this this is important because if they're not in the same ap group even though they may be adjacent to each other in placement they're not going to be calculated together and this is a little bit different because a lot of people are using ap groups for management of SS IDs and other things more and more this is becoming critical to our F with our profiles and the micro networks that we're creating within larger networks so very important to remember that and it's easy to overlook coverage overlap is the percentage of a given cell that's going to be covered by ApS at - 67 DB or greater so what that says is if I have 90% coverage the 90% of that cell is being covered by other ApS at - 67 DB or greater neighbor cutoff for fr a consideration is minus 70 DBM so we're only going to consider radios that are neighbors of ours that we are seeing at above minus 70 DBM and then fr a is only going to use the four loudest neighbors when evaluating any single ap for redundancy so if you have eight or nine neighbors that are above minus seventy other than being an incredibly dense network we're still going to look at the four loudest of that list as being the coverage candidates for that ap another thing that's important to remember is that all ap models are considered in coverage calculation it doesn't matter whether it's a thirty eight hundred twenty eight hundred if it's part of the same RF network and in the same AP group we consider every 2.4 gigahertz radio as a potential candidate for covering the fr a radio so the recommendations around this that we've had has been to say at least one aap group or floor and there's a lot of good reasons to do that other than just fr a in fact we we typically have floors now even in Cisco infrastructure where we might have two or three aap groups that are basically the repeat of the AAP group for the floor except in a in a micro environment like a like a lecture room where we need to have tighter RF configurations because of the density so make sure that you have at least one per floor and this makes it easy to adjust things like transmit power and things like that where you can go into the AP our profile settings on the AP group and make a change to a whole group RF profiles are used we can set a lot of these factors in an RF profile and then the final thing to remember is only 228 3800 can be marked as redundant and this is just a primer on the hardware and the and the restrictions right so yeah this is the question we get all the time that how is our separation done as well as will there be any interference between the slot zero and slot one when we kind of put both of them in the five gig and how we have taken care of it we usually still get this question from a lot of customers absolutely this really talks about the differences between the I model and the e model and the sameness the AI model is intended for you know make Roma cooperation right maker radio is is slot zero same limitations as the AI Model T X power and channel our custom for our M hundred megahertz channel separation is enforced by software right now same SSID B's will be broadcast out of both we don't really have a configuration ability to separate Justice IDs on to radios not in the same band because right now we can say one SSID only for 2.4 and one for 5 gigahertz different problem in the configuration dialog if we have 2 5 gigahertz that's a detailed slot on fr a reverse you

Channel: CiscoWLAN

Views: 6,436

Keywords:

Id: cb1rV16opj4

Length: 24min 32sec (1472 seconds)

Published: Fri Dec 02 2016