SP 2016 - Only TTFB slow

We have a pre-production SP 2016 instance (single server) that we are commissioning but for some reason the landing page TTFB is slow. The rest of the resources and page load in reasonable time but the initial TTFB is usually between 5-10 seconds.

We have a DEV server configured the same (with about a quarter of the resources assigned) and the page loads fine.

Here are some Chrome ‘Network’ stats:

And here is the tree view to show compared to the other resources:

I am at the end of my rope trying ot figure this out. CPU, memory and disk are basically zero on both the single app server and the SQL instance, there is basically no content in the site at all yet and there are only two admins even accessing it.

I’ve checked every log I can find and I can’t seem to find anything that would cause this single issue. The strangest part is that every couple of minutes it will go back down to a reasonable amount (less than 500ms) to load the entire page.

To get some more metrics I set a PRTG sensor to scrape the page every 30 seconds and the load time fluctuates wildly, generally 5-10 seconds but dipping all the way down also.

Any advice on where to look?

Things I’ve looked at:

• Wait times in SQL (basically 0ms)
• Bandwidth is essentially 0
• CPU, RAM, DISK I/O is basically 0

EDIT : There appears to actually be several 401 errors with different resources (mostly images) which I’m having our SP admin look into as well. Some of them seem to have several 401 errors, then a 200 success which seems to indicate to me that it is trying different authentication methods before finally succeeding.

Looking at both the working DEV box and the slow production box they appear to have the same permissions configure though so we will have to see.

EDIT2 : Authentication issues do not appear to be related. I’ve run some fiddler tests on the DEV and PRD and it shows cleary there is some sort of delay on the app side.

This is the response from the DEV box (working correctly):

Request Count:   1
Bytes Sent:      1,359      (headers:1,359; body:0)
Bytes Received:  175,636        (headers:659; body:174,977)

ACTUAL PERFORMANCE
--------------
ClientConnected:    10:36:09.958
ClientBeginRequest: 10:36:24.414
GotRequestHeaders:  10:36:24.414
ClientDoneRequest:  10:36:24.414
Determine Gateway:  0ms
DNS Lookup:         0ms
TCP/IP Connect: 0ms
HTTPS Handshake:    0ms
ServerConnected:    10:36:24.377
FiddlerBeginRequest:    10:36:24.415
ServerGotRequest:   10:36:24.415
ServerBeginResponse:    10:36:24.578
GotResponseHeaders: 10:36:24.578
ServerDoneResponse: 10:36:24.590
ClientBeginResponse:    10:36:24.590
ClientDoneResponse: 10:36:24.591

    Overall Elapsed:    0:00:00.177

and this is from the PRD box (slow response):

Request Count:   1
Bytes Sent:      1,321      (headers:1,321; body:0)
Bytes Received:  159,171        (headers:661; body:158,510)

ACTUAL PERFORMANCE
--------------
ClientConnected:    10:36:55.024
ClientBeginRequest: 10:36:55.391
GotRequestHeaders:  10:36:55.391
ClientDoneRequest:  10:36:55.391
Determine Gateway:  0ms
DNS Lookup:         0ms
TCP/IP Connect: 0ms
HTTPS Handshake:    0ms
ServerConnected:    10:36:55.373
FiddlerBeginRequest:    10:36:55.391
ServerGotRequest:   10:36:55.391
ServerBeginResponse:    10:37:08.374
GotResponseHeaders: 10:37:08.374
ServerDoneResponse: 10:37:08.391
ClientBeginResponse:    10:37:08.391
ClientDoneResponse: 10:37:08.392

    Overall Elapsed:    0:00:13.001

This is actually an extreme example of 13 seconds, almost consistently it is 4.9 seconds from ‘ServerGotRequest’ to ‘ServerBeginResponse’

EDIT 3 : The SPRequestDuration header returns : 13375.