Images Load and TTFB

#1
Hello, first of all I want to say I am very happy to be testing the trial of LSws Enterprise 2 CPU and will buy 1-CPU license once the testing period is over.

My client has magento virtual store, last month I configured an instance in AWS with Nginx, everything looked pretty good, except that TTFB remained high, however the site had a very fast load time.

My client wanted to reduce costs and went to Soyoustart, although now the server was on another continent (USA - Canada) and we here in Brazil, I was extremely surprised with load, the same load as before, when we were in DC with AWS in our country.

Now I TTFB variation in the times is F but when I do the second test, get A, whereas this test is performed in WebPagetest site and reference the location of São Paulo, the server is in the U.S. and am testing to Sao Paulo, actually a result of standing to applaud.

I'll start adopting litespeed as an alternative to Nginx for my clients, I have been working with nginx to a year, but now it seems I'll have to think of an alternative for those who do not mind paying a bit more to have a performance even higher.

I need the help of friends, I have two questions:

1 - Is there any setting to reduce TTFB time low, to reduce the response time?

2 - My only dissatisfaction at the moment are the pictures, see the link below the image, everything loads very fast, but the images (3 banners 150kb each) are actually causing major slowdowns, one of these images comes charging in almost 5 sec if it could be solved or improved I would be very grateful.

Image Link http://screencast.com/t/3geTQudSTQi
 

NiteWave

Administrator
#2
can you run the test a few more times, to confirm if the 3 images is slowest all the time ?

both litespeed and nginx is event-driven, static files (including images) performance have been optimized to max(as we did benchmarks before, the bottle neck is network card, both can saturate 1G / 10G NIC easily when serve static files). I think TTFB mainly depending on the network condition between client and server. the only related setting in lsws may be
lsws web admin->Server->Tuning->Smart Keep-Alive:No
you can on/off it to see if any difference.
 
#3
Yes, qhen i am optimizing a server or site i run a test many times

This test was performed in the OVH dedicated server with link 250-300 mbits.

How can I identify if the bottleneck is the network card?

Current tests are now like the image below, I used tinypng to improve compression of images

http://screencast.com/t/EyAoLE8y
 

NiteWave

Administrator
#4
our benchmark tests tell us network card is the bottle neck.
in real world, network card itself is hard to be saturated and not the bottle neck at all. rather, "network connection" between client and server may be the main delay factor. in contrast, you can run same tests in LAN, the network delay should be reduced to minimum.

have you tried
lsws web admin->Server->Tuning->Smart Keep-Alive:No/Yes
any difference?
 
Top