Re: XO 1.5 - CONTENTION WINDOW
On Tue, Dec 1, 2009 at 5:43 PM, Franco Miceli fmic...@plan.ceibal.edu.uy wrote: Can someone explain to me why these registers have been set to such values? Is this a mistake or it is something done for some reason in particular? Hi Franco, What values are you expecting, and why? cheers, m -- martin.langh...@gmail.com mar...@laptop.org -- School Server Architect - ask interesting questions - don't get distracted with shiny stuff - working code first - http://wiki.laptop.org/go/User:Martinlanghoff ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: XO 1.5 - CONTENTION WINDOW
What I was expecting was to have CWmax closer to 1024. What I think happens when you set CWmax to 31 is that -as you have more and more machines trying to compete for the medium- you could be having more colisions -since they would wait less time slots to tx and collisions would be more likely to happen. When you set CWmax to a higher value you prevent this situation. Of course this is for a higher number of machines using the medium. Maybe I'm wrong but this might be optimized for a certain number of machines. If so, could you tell me for what number? (not exactly just an estimate). Thanks Martin. Franco 2009/12/2 Martin Langhoff martin.langh...@gmail.com On Tue, Dec 1, 2009 at 5:43 PM, Franco Miceli fmic...@plan.ceibal.edu.uy wrote: Can someone explain to me why these registers have been set to such values? Is this a mistake or it is something done for some reason in particular? Hi Franco, What values are you expecting, and why? cheers, m -- martin.langh...@gmail.com mar...@laptop.org -- School Server Architect - ask interesting questions - don't get distracted with shiny stuff - working code first - http://wiki.laptop.org/go/User:Martinlanghoff ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: XO 1.5 - CONTENTION WINDOW
On Wed, 2009-12-02 at 09:33 -0200, Franco Miceli wrote: What I was expecting was to have CWmax closer to 1024. What I think happens when you set CWmax to 31 is that -as you have more and more machines trying to compete for the medium- you could be having more colisions -since they would wait less time slots to tx and collisions would be more likely to happen. When you set CWmax to a higher value you prevent this situation. Of course this is for a higher number of machines using the medium. Could you directly reference the code that you are referring to, and the register documentation that you are looking at? Thanks, Daniel ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: XO 1.5 - CONTENTION WINDOW
Sure, I'm looking directly at the registers in the XO 1.5 B2 laptop. The process I follow is the following one: # mount -t debugfs null /mnt # cd /mnt/lbs_wireless/eth0/ To read register 0x # echo 0x rdmac ; cat rdmac MAC[0x0] = 0x0033fa05 I've read the following registers: 0xA0A0 - CWMin queue 0 0xA0A4 - CWMax queue 0 0xA0A8 - CWMin queue 1 0xA0AC - CWMax queue 1 0xA0B0 - CWMin queue 2 0xA0B4 - CWMax queue 2 0xA0B8 - CWMin queue 3 0xA0BC - CWMax queue 3 Now, this are the four queues I've read, and they have the values: CWmax = 31 CWmin = 7 The eight registers that follow this ones have the exact same pattern (7 31 7 31). That's why I mentioned eight queues (but I might be wrong and those other queues could be backups of the first ones). Thanks, Franco 2009/12/2 Daniel Drake d...@laptop.org On Wed, 2009-12-02 at 09:33 -0200, Franco Miceli wrote: What I was expecting was to have CWmax closer to 1024. What I think happens when you set CWmax to 31 is that -as you have more and more machines trying to compete for the medium- you could be having more colisions -since they would wait less time slots to tx and collisions would be more likely to happen. When you set CWmax to a higher value you prevent this situation. Of course this is for a higher number of machines using the medium. Could you directly reference the code that you are referring to, and the register documentation that you are looking at? Thanks, Daniel ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: XO 1.5 - CONTENTION WINDOW
On Wed, 2009-12-02 at 10:06 -0200, Franco Miceli wrote: Now, this are the four queues I've read, and they have the values: CWmax = 31 CWmin = 7 The eight registers that follow this ones have the exact same pattern (7 31 7 31). That's why I mentioned eight queues (but I might be wrong and those other queues could be backups of the first ones). Ok, and what about the documentation? Are you sure that a value of 31 actually means 31 slots? On other wireless hardware I have worked with, this was not the case. Daniel ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: XO 1.5 - CONTENTION WINDOW
On Wed, Dec 2, 2009 at 1:32 PM, Daniel Drake d...@laptop.org wrote: Ok, and what about the documentation? Also - what does the firmware on the card do with that? Could it be overwritten by a DCW algorythm? m -- martin.langh...@gmail.com mar...@laptop.org -- School Server Architect - ask interesting questions - don't get distracted with shiny stuff - working code first - http://wiki.laptop.org/go/User:Martinlanghoff ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: XO 1.5 - CONTENTION WINDOW
Daniel, I've not got such documentation, could you tell me where can I get it? Also - what does the firmware on the card do with that? Could it be overwritten by a DCW algorythm? -Martin Martin, No idea if that happens, I've only read such parameters since they were used also by the XO1.0 and assumed that they remained there for the XO 1.5. If I can get a copy of the documentation I may be able to understand a little bit more about this. Thanks, Franco 2009/12/2 Daniel Drake d...@laptop.org On Wed, 2009-12-02 at 10:06 -0200, Franco Miceli wrote: Now, this are the four queues I've read, and they have the values: CWmax = 31 CWmin = 7 The eight registers that follow this ones have the exact same pattern (7 31 7 31). That's why I mentioned eight queues (but I might be wrong and those other queues could be backups of the first ones). Ok, and what about the documentation? Are you sure that a value of 31 actually means 31 slots? On other wireless hardware I have worked with, this was not the case. Daniel ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
Re: XO 1.5 - CONTENTION WINDOW
On Wednesday 02 December 2009 10:49:38 am Martin Langhoff wrote: On Wed, Dec 2, 2009 at 1:32 PM, Daniel Drake d...@laptop.org wrote: Ok, and what about the documentation? Also - what does the firmware on the card do with that? Could it be overwritten by a DCW algorythm? Are you sure that you're using a PHY mode with fixed bounds for the CW? The default values you suggest (31-1023) are for DCF. I think those are set by the AP in HCF. You could check if those values change in different environments (with an AP that properly supports those extensions) to test that. Here's what 802.11-2007 says about the Hybrid Coordination Function: The contention window limits aCWmin and aCWmax, from which the random backoff is computed, are not fixed per PHY, as with DCF, but are variable (contained in the MIB attribute tables dot11QAPEDCACWmin and dot11QAPEDCACWmax for an AP and in the MIB attribute tables dot11EDCATableCWmin and dot11EDCATableCWmax for a non-AP STA) and assigned by a management entity or by an AP. m -- -Andrés ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel
XO 1.5 - CONTENTION WINDOW
Hi, Looking for the reasons for low throughput on the XO 1.5 I've come across with the contention window's registers. I've found that there are like eight queues and that for all of them the CWmax and min are set as follows: CWmin = 7 slots CWmax = 31 slots As I see it this would make for a decrease in performance in situations with a high number of customers connected to the same AP. Can someone explain to me why these registers have been set to such values? Is this a mistake or it is something done for some reason in particular? Thanks. Franco ___ Devel mailing list Devel@lists.laptop.org http://lists.laptop.org/listinfo/devel