Hi Trampas, thanks for sharing.
It was not clear, are you using some custom loading of resources (eg script and 
css files) ? Or do you mean just data (such as JSON)?
Could you share details how that is performed?
I was looking for some way that I can manage the loading manually and request 
each resource one by one instead of browser doing it..
but I couldn't find a way.

Unfortunately from what I search, there is no way to configure the browsers how 
should they behave on loading the multiple resources files.
I tried some tags for preload etc and none made any effect (at leas in firefox, 
where I was testing..)

Mario

________________________________________
From: lwip-users <lwip-users-bounces+mrluzeiro=ua...@nongnu.org> on behalf of 
Trampas Stern <tram...@gmail.com>
Sent: 10 May 2020 13:21
To: Mailing list for lwIP users
Subject: Re: [lwip-users] httpd, tls, multiple get requests, delays

I have ran into similar issues.  I ended up changing my HTML source files.

To understand let me start with the background.  I have a web page where ever 
second I request data from the embedded server and display it.  So in the 
Javascript for the web page I had a 1000ms timeout that requested new 
data(file).  This worked great on the bench.  However the customer was testing 
it and complaining about it being slow.  As it turns out the issue was the 
customer had a slow laptop and when they started a zoom meeting the network 
traffic on the laptop went up and my device got really really slow.
What I found was happening is I would make a request for data, it would time 
out inside chrome which would retry the request.  That is as the requests timed 
out in the browser it would retry, this meant as time moved forward I would 
start with browser trying to request one copy of data, then if that timed out 
it would request it again while my timer in javascript also requested it again, 
now I had two connections for the same file on the embedded device. Then three, 
then four...

The solution was to change the javascript such that it made request and if I 
did not get a response in 500ms it would time out the request, before I made a 
second request. This way at any time the browser would only allow one request 
per file. I also could increase the timeout between requests to match network 
speed.   I found this also needed to be done with CSS files.  That is sometimes 
I found that I would get random failure requesting files.  For example at one 
time my Phy chip was getting too hot and causing random network failures that 
cause havoc on my system.  I found that this would cause random failures, and 
retries.  By having the javascript control the time out on the file it helped.  
That is with the random failures and the time out for the http connection in 
lwip I would again get where I had two open connections for the same 
resource/file.  Hence timing out on javascript would send the close to the 
server and thus kill one connection before starting a new one.

This also leads to the other issue, when you use LWIP you have think about how 
the system handles multiple requests for the same resource.  That is if you are 
sending out a file can lwip have two open connections to the file?  For example 
two computers accessing your server at the same time.  This is one reason I 
implemented authentication system, as it restricts the number of connections, 
by restricting the number of users.

Trampas



On Sat, May 9, 2020 at 5:03 PM Mário Luzeiro 
<mrluze...@ua.pt<mailto:mrluze...@ua.pt>> wrote:
Hello all,

I'm looking for some ideas how can I improve an issue I have.
I have a httpd over TLS. As you may know, browsers fire multiple requests at 
same time to get the resources.

I had a lot of issues related with TLS / LWIP connections to handle this 
multiple connections but the major are now working ok...
but as you can see in the screenshot, despite most of the requests are handled 
very quickly but for some reason.. some will take too long to handle.
There is some pattern on this, sometimes a bit random on the time it takes..
The first time the browser connects to the server it takes a lot to initialize 
the TLS, but its ok because the next ones are very fast.. except that random 
get requests

Any idea how can it be improved ?

I already support 20 TCP connections ( MEMP_NUM_TCP_PCB is 20 * 2 ) and played 
with lots of parameters that I tuned for the best results except this last one..

Regards,
Mario Luzeiro
_______________________________________________
lwip-users mailing list
lwip-users@nongnu.org<mailto:lwip-users@nongnu.org>
https://lists.nongnu.org/mailman/listinfo/lwip-users

_______________________________________________
lwip-users mailing list
lwip-users@nongnu.org
https://lists.nongnu.org/mailman/listinfo/lwip-users

Reply via email to