[FOLIO-1118] folio-demo.gbv.de takes 50 seconds to load the first page Created: 13/Mar/18  Updated: 12/Nov/18  Resolved: 19/Mar/18

Status: Closed
Project: FOLIO
Components: None
Affects versions: None
Fix versions: None

Type: Bug Priority: P3
Reporter: Julian Ladisch Assignee: Julian Ladisch
Resolution: Won't Do Votes: 0
Labels: sprint34
Remaining Estimate: Not Specified
Time Spent: 2 hours
Original estimate: Not Specified

Attachments: PNG File 2018-03-13-folio-demo.gbv.de-startup-time.png     PNG File 2018-03-14-page-load-speed.png     PNG File screenshot_52.png     PNG File screenshot_54.png     PNG File screenshot_55.png     PNG File screenshot_56.png     PNG File screenshot_57.png    
Issue links:
Relates
relates to STCOR-26 Async module loading Closed
Sprint:

 Comments   
Comment by Julian Ladisch [ 13/Mar/18 ]

I cannot reproduce this report. For me it takes 5 seconds until the login screen is ready so that I can enter the credentials. Jason Skomorowski also reports that is is way faster than 50 seconds.

Please invoke Strg+Shift+I and select the "Network" tab. Then enter https://folio-demo.gbv.de in the URL bar and attach a screenshot of the resulting network time chart. I attach my network time chart.

Comment by Peter Murray [ 13/Mar/18 ]

I cannot reproduce the problem this morning.

I did look at the network inspector last night, but I didn't make a screenshot of it. The file that was slowing things down was the `bundle.....js` file, although I don't recall what part of the transfer process was taking 50 seconds. I'll follow-up with the members of the Accessibility SIG to see if the issue has disappeared for them as well.

Comment by Julian Ladisch [ 13/Mar/18 ]

Peter, thanks for the screenshot_52.png. It shows that downloading 3 MB takes 8 seconds. The server can send the 3 MB in less than 1 second. This looks as if the internet connectivity is slow.

Comment by Julian Ladisch [ 13/Mar/18 ]

I've added $request_time to folio-demo.gbv.de's nginx log to see how long the request-response-times are.

Comment by Peter Murray [ 13/Mar/18 ]

I heard back from Gene Hayworth at University of Colorado and he was seeing the issue. I tried again, and it does seem to be some sort of slow connectivity. This request was made at about 15.00 GMT from IP address 76.219.105.194.

Comment by Julian Ladisch [ 14/Mar/18 ]

Thanks for the feedback, I see 127 seconds in the log at 16:01:46 +0100 = 15:01:46 GMT.

0.0.0.0 - - [13/Mar/2018:14:27:43 +0100] "GET /bundle.a09a40df3c860c0b90b4.js HTTP/1.1" 200 3072027 "-" "-" 1.016
0.0.0.0 - - [13/Mar/2018:15:32:48 +0100] "GET /bundle.a09a40df3c860c0b90b4.js HTTP/1.1" 200 3072027 "-" "-" 90.029
0.0.0.0 - - [13/Mar/2018:15:51:00 +0100] "GET /bundle.a09a40df3c860c0b90b4.js HTTP/1.1" 200 3072027 "-" "-" 0.116
0.0.0.0 - - [13/Mar/2018:15:59:37 +0100] "GET /bundle.a09a40df3c860c0b90b4.js HTTP/1.1" 200 3072027 "-" "-" 34.154
0.0.0.0 - - [13/Mar/2018:16:01:46 +0100] "GET /bundle.a09a40df3c860c0b90b4.js.map HTTP/1.1" 200 10638312 "-" "-" 127.019
0.0.0.0 - - [13/Mar/2018:17:10:11 +0100] "GET /bundle.a09a40df3c860c0b90b4.js HTTP/1.1" 200 3072027 "-" "-" 1.124
0.0.0.0 - - [13/Mar/2018:17:15:26 +0100] "GET /bundle.a09a40df3c860c0b90b4.js HTTP/1.1" 200 3072027 "-" "-" 1.242
0.0.0.0 - - [13/Mar/2018:17:15:50 +0100] "GET /bundle.a09a40df3c860c0b90b4.js HTTP/1.1" 200 3072027 "-" "-" 7.780
0.0.0.0 - - [13/Mar/2018:19:15:52 +0100] "GET /bundle.a09a40df3c860c0b90b4.js HTTP/1.1" 200 3072027 "-" "-" 4.749
0.0.0.0 - - [13/Mar/2018:19:15:57 +0100] "GET /bundle.a09a40df3c860c0b90b4.js HTTP/1.1" 200 3072027 "-" "-" 2.986
0.0.0.0 - - [13/Mar/2018:19:16:02 +0100] "GET /bundle.a09a40df3c860c0b90b4.js HTTP/1.1" 200 3072027 "-" "-" 12.111
0.0.0.0 - - [13/Mar/2018:19:18:20 +0100] "GET /bundle.a09a40df3c860c0b90b4.js HTTP/1.1" 200 3072027 "-" "-" 3.283

Other requests are much faster, see the last number showing the request-response-times.
Now I've enabled ip and browser name (http_user_agent) logging to see whether this only affects a few people (connectivity issue) or all.

Comment by Peter Murray [ 14/Mar/18 ]

Thanks for adding the extra debugging. I tried this morning in a new incognito window in Chrome (1323 UTC):

Safari on MacOS is showing the same network capacity issue (1330 UTC):

Oddly, a traceroute with a 1024 packet size doesn't show significant delays:

$ traceroute -I folio-demo.gbv.de 1024
traceroute to esx-181.gbv.de (193.174.240.181), 64 hops max, 1024 byte packets
 1  192.168.10.1 (192.168.10.1)  1.545 ms  1.909 ms  1.169 ms
 2  * * *
 3  76-219-104-1.lightspeed.clmboh.sbcglobal.net (76.219.104.1)  28.721 ms  38.074 ms  24.098 ms
 4  71.151.141.21 (71.151.141.21)  26.625 ms  31.154 ms  24.513 ms
 5  * * *
 6  12.83.38.109 (12.83.38.109)  25.215 ms  25.651 ms  26.808 ms
 7  cgcil402igs.ip.att.net (12.122.133.161)  34.160 ms  35.171 ms  35.603 ms
 8  * 192.205.37.150 (192.205.37.150)  32.977 ms  32.908 ms
 9  ae-1-5.bar1.hamburg1.level3.net (4.69.142.209)  128.868 ms  128.508 ms  128.881 ms
10  195.122.181.62 (195.122.181.62)  130.723 ms  130.276 ms  130.717 ms
11  cr-han2-hundredgige0-0-0-5.x-win.dfn.de (188.1.144.38)  143.572 ms  143.279 ms  145.818 ms
12  kr-goe12-1.x-win.dfn.de (188.1.231.126)  146.781 ms  145.250 ms  144.792 ms
13  bm-cluster1-stargate-ext.net.gwdg.de (134.76.250.251)  145.840 ms *  145.241 ms
14  xr-fmz1-stargate.net.gwdg.de (134.76.250.4)  144.912 ms  144.295 ms  145.209 ms
15  xr-theo1-attractor.net.gwdg.de (134.76.147.19)  138.872 ms  139.240 ms  138.730 ms
16  gr-vbz-theo.gwdg.de (134.76.249.213)  138.738 ms  139.698 ms  138.455 ms
17  * * *
18  * *^C
Comment by Julian Ladisch [ 14/Mar/18 ]

I've enabled gzip compression. This should reduce the download time of /bundle.a09a40df3c860c0b90b4.js to 25%.

The speed test at https://www.dotcom-tools.com/website-speed-test.aspx shows load times between 0.6 and 8 seconds.

Comment by Julian Ladisch [ 19/Mar/18 ]

It looks as if this is caused by the throughput speed of the internet connection.
Enabling gzip compression has reduced the download time to 25%.
Speed tests from different locations result in page load times of 8 seconds or less for the first visit, following visits are much faster because of browser cache.
I close this issue because gzip compression reduced the priority and there isn't any other measure with resonable effort.

Comment by Peter Murray [ 19/Mar/18 ]

I agree, and I'll relay this to the accessibility SIG member.

For what its worth, the test I ran this morning in an incognito window has the JavaScript bundle downloaded in just under four seconds.

Generated at Thu Feb 08 23:10:58 UTC 2024 using Jira 1001.0.0-SNAPSHOT#100246-sha1:7a5c50119eb0633d306e14180817ddef5e80c75d.