mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   chalsall (https://www.mersenneforum.org/forumdisplay.php?f=169)
-   -   A request for assistance on a very interesting problem... (https://www.mersenneforum.org/showthread.php?t=26556)

chalsall 2021-03-02 22:43

A request for assistance on a very interesting problem...
 
So... I find myself in a very interesting situation...

All my instrumentation tells me everything should be working perfectly. And it does when testing.

But it doesn't /seem/ to /always/ when at "scale". Doing a particular thing.

Unfortunately, this scaling can't be simulated. There have to be "Humans in the loop".

Specifically, 20 or so Humans all holding Android phones, all sharing the same WiFi network, all at the same time. All trying to do the same thing at once...

I was wondering if there might be anyone out there who might have a class of undergraduates who might be willing to spend 30 minutes of their day (to be scheduled; perhaps after a lecture) to help debug a very interesting problem.

I can't go into a whole lot of details. Please PM me if you (or someone you know) might be able to assist and I can share more.

Namaste.

Uncwilly 2021-03-02 22:53

Can't you buy that kind of service from a Chinese "click" farm? I have seen pictures of a single individual with racks of phones and tablets and rooms of people doing the same thing on PCs.

chalsall 2021-03-02 23:31

[QUOTE=Uncwilly;572884]Can't you buy that kind of service from a Chinese "click" farm? I have seen pictures of a single individual with racks of phones and tablets and rooms of people doing the same thing on PCs.[/QUOTE]

Hadn't thought of that, but possibly...

Not exactly the kind of supporting evidence that I'd like to reference, though... "Yeah... I went on the Dark web' and found 'a guy'. He told me these were the results... Only 0.0001 bitcoins!" :wink:

jwaltos 2021-03-03 05:16

I obtained a diploma from here (quite) a while back and found the school to be responsive.
Here's a link: [url]https://www.conestogac.on.ca/engineering-and-technology/contact[/url] (Rafik Guindi).
As the birthplace of the Blackberry and present tech. hub I'm certain there's some assistance available locally which could be beneficial to both of you.

As you may have experienced in the past, flushing out one bug sometimes allows you to discover a nest of the damn things..good luck.

Nick 2021-03-03 09:10

Scaling is precisely what POTS tech did well and what digital tech does badly.

Dr Sardonicus 2021-03-03 12:49

[QUOTE=chalsall;572882]So... I find myself in a very interesting situation...

All my instrumentation tells me everything should be working perfectly. And it does when testing.[/QUOTE]Apparently there is something relevant your instrumentation is not measuring. The question then is, what?

Apart from that, why is the situation "interesting?"

M344587487 2021-03-03 13:20

All I know is that (at least some) WAPs have concurrent device limitations regardless of bandwidth that may be related to your issue. I know a guy who (god knows why) likes off-the-shelf IoT devices for home automation, he had to upgrade to a beefy consumer solution when hitting concurrency issues at ~20 devices and is now looking at upgrading to something more robust as he's hitting concurrency issues again at ~40 devices.



So maybe add another WAP to spread the users across? I don't know much about networking other than it's a spider's nest that should be avoided or set on fire whenever possible.

kriesel 2021-03-03 14:35

It's not clear which problem you're facing, server hosted application scaling, or wireless access scaling. The following assumes the latter.
How many limits are there? At least:
[LIST=1][*] Access point chipset constraints, typically in the hundreds[*]Limits in practice are typically much lower [url]https://community.cisco.com/t5/wireless-mobility-documents/how-many-wireless-clients-an-access-point-can-support/ta-p/3108926[/url][*]Presence of consumer grade firmware, perhaps imposing a much lower device count limit[*] Bandwidth consumption per device, depends on the app & usage patterns[*] Adjacent-channel crowding by someone else's WAP next door (or they're channel hopping, sometimes landing on the same channel) reducing usable bandwidth[*] Other activity/bandwidth consumption on the same access point/router/uplink, especially bursty high peak use[*] EMI from other devices lowering S/N so lowering usable data rate (break room microwave, unshielded open frame multigpu mining rig, nearby analog radio or tv transmitters including amateurs, someone else's garage shop arc welder or a failing fluorescent ballast etc. I once had a computer monitor whose CRT high voltage supply leakage became visible tracking across its insulator surface.)[*] Distance & antenna gain considerations, attenuation by walls and other material perhaps including the humans holding the phones, again lowering signal strength and so lowering achievable data rate[*]Interference between multiple close-spaced phones[*]Multipath interference[*] ISP-imposed limits?[*]unaccounted for IOT devices inflating the device count (what do the AP/router MAC address tables say?)[*]Open access or compromise and someone's [URL="https://en.wikipedia.org/wiki/Wardriving"]wardriving[/URL] and [URL="https://en.wikipedia.org/wiki/Piggybacking_(Internet_access)"]piggybacking[/URL][*]Occasional router-enforced client or upstream-enforced router DHCP expiration and renewal can be disruptive[*]Marginal NIC in the wired section causing packet storms or table overflow, or a marginal piece of wireless gear[*]Timeout issues[*]If n wireless range extenders or bridges are being used in series, divide nominal bandwidth by ~2[SUP]n[/SUP][*] What else?[/LIST]Or is the scenario 20 students in a classroom all hitting the same cell tower upon command by the instructor?

Due to your occupation, I expect you're quite familiar with all that. Sometimes making the whole list can help. The vagueness of the initial post makes an on target response a challenge. I'm guessing there is contractual or law enforcement basis for the lack of information, perhaps as benign as a nondisclosure agreement.

Batalov 2021-03-03 18:20

[QUOTE=chalsall;572882]So... I was wondering if there might be anyone out there who might have a class of undergraduates who might be willing to spend 30 minutes of their day (to be scheduled; perhaps after a lecture) to help debug a very interesting problem.
[/QUOTE]
[url]https://www.fiverr.com/[/url] !!

chalsall 2021-03-03 22:58

[QUOTE=kriesel;572918]It's not clear which problem you're facing. Due to your occupation, I expect you're quite familiar with all that. Sometimes making the whole list can help.[/QUOTE]

Indeed. Thanks. To all of your (and others') above.

This has helped me model this in my head.

NDA's suck.

jwaltos 2021-03-04 04:01

NDA's can also save your butt.
I was just leafing through the latest 2600 and noticed on p.16 "Technical Details of Layer 1" that this kind of material may help you think outside the tesseract and provide you with an assist..if you're thinking about topology.


All times are UTC. The time now is 03:59.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.