Speedtest Results Questions

[pcklutz]

Paying ISP $62/mo for “up to 400/20 Mbps” internet.

Using testmy.net, recorded 5 minute samples for 16 hours (0000 to 1600 CST) on three days; Tuesday before Thanksgving, Thanksgiving Day, and Saturday after Thanksgiving. 

Test Setup:

Modem: ARRIS DG2470A (ISP provided)

Test PC: Dell Inspiron (64 bit Win 8.1, i5-3230M CPU, 8 GB RAM, 1 TB HD)

Cable: New 1 foot, CAT6 cable between PC and modem; nothing else connected to modem ports, wireless not used

Plotted TMN results in Excel.  Results were similar each day (see graphs)

Observations:

Download speed was very erratic

Peak samples all three days were ~244 Mbps (3 – 6 AM)

Speed typically <100 Mbps before 3 AM and during day

Upload speed not as erratic; either consistently good or consistently poor.

Upload speed dropped so low at ~1230 Thanksgiving Day that test stopped; manually restarted ~1345 when I discovered test had stopped.

Comment:

At 1030 AM on 27 Nov, ISP tech on scene to check out service.  OOKLA spot test said 418 Mbps; my PC spot test said 423 Mbps.  No surprise there - GIGO.

Questions:

Is the short-term wide variation in DL speed typical?

Is the longer-term wide variation in DL speed between middle-of-night and rest-of-time typical?  (I’m not surprised that there is a difference between busy and not-so-busy time speeds, just surprised that variation is as great as what TMN recorded.)

What can cause the UL speed to be consistently good and consistently poor (see 21 & 23 Nov graphs)?

With normal setups in the real (consumer) world, would OOKLA speeds ever be realized?  If not, it seems to me that the advertised/billed speeds would be fraudulent were it not for the weasel words “up to”

Would any professional network engineer use OOKLA or accept its results?

 

 

 

 

 

21 Nov speedtest.pdf

23 Nov speedtest.pdf

25 Nov speedtest.pdf