Da hast du natürlich vollkommen Recht, das würde den Traffic beeinflussen.
Aber der Server hängt ganz alleine an einer 1 GBit/s Leitung.
Und im Monitoring sehe ich die Auslastung des Servers am Ethernet-Port.
Zeitgleich funktioniert der Download aber an einem TELEKOM DEUTSCHLAND LAN IP FIBER 1.000 "korrekt" 91,7 MB/s (733,6 MBit/s), da limitiert dann aber die CPU und der WGET-Befehl zeigt 99% CPU-Last an (ist ein Quad 1.2 GHz MIPSEL 64-Bit Router auf dem ich den Download durchführe). Der Upload wird von einer "großen" Maschine durchgeführt, welche bei 6,8 GBit/s seine Probleme findet bei MTU 1500b.
Fakt ist, dass die KD mehrere Routen zum Ziel (mich als Kunde) hat. Keine Ahnung wie die Netz-Topologie gebaut ist, aber sicherlich gehen keine 100G von der Mitte durch das ganze Mesh bis zum CMTS. Ich gehe davon aus, dass n * 10G oder n * 40G verwendet wurde, und als aggregation-Protokoll LACP / LAG mit Layer2+Layer3+Layer4 Hash verwendet wird. Das würde zumindest erklären, warum bei mehrfacher Wiederholung der Tests die Geschwindigkeit so immens schwankt.
Hier Test von Telekom Business Fiber 1000
Code: Alles auswählen
root@ubnt:/home/rack# curl -o /dev/null --interface 192.168.178.2 http://download.rsm-connect.net/1gb.iso
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
67 1108M 67 745M 0 0 88.6M 0 0:00:12 0:00:08 0:00:04 91.3M
Kabel Deutschland Business 500
Code: Alles auswählen
C:\Users\Michael Rack>wget -O NUL --report-speed=bits http://download.rsm-connect.net/1gb.iso
--2019-11-14 07:49:10-- http://download.rsm-connect.net/1gb.iso
Resolving download.rsm-connect.net (download.rsm-connect.net)... 185.58.31.10
Connecting to download.rsm-connect.net (download.rsm-connect.net)|185.58.31.10|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 1162083328 (1,1G) [application/octet-stream]
Saving to: 'NUL'
NUL 41%[=====================================> ] 454,90M 437Mb/s eta 12s
^C
C:\Users\Michael Rack>wget -O NUL --report-speed=bits http://download.rsm-connect.net/1gb.iso
--2019-11-14 07:50:06-- http://download.rsm-connect.net/1gb.iso
Resolving download.rsm-connect.net (download.rsm-connect.net)... 185.58.31.10
Connecting to download.rsm-connect.net (download.rsm-connect.net)|185.58.31.10|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 1162083328 (1,1G) [application/octet-stream]
Saving to: 'NUL'
NUL 0%[ ] 6,63M 27,6Mb/s
^C
C:\Users\Michael Rack>wget -O NUL --report-speed=bits http://download.rsm-connect.net/1gb.iso
--2019-11-14 07:50:08-- http://download.rsm-connect.net/1gb.iso
Resolving download.rsm-connect.net (download.rsm-connect.net)... 185.58.31.10
Connecting to download.rsm-connect.net (download.rsm-connect.net)|185.58.31.10|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 1162083328 (1,1G) [application/octet-stream]
Saving to: 'NUL'
NUL 100%[=====================================================================>] 1,08G 301Mb/s in 35s
2019-11-14 07:50:43 (266 Mb/s) - 'NUL' saved [1162083328/1162083328]
Man sieht, dass die Geschwindigkeit einmal bei 437 MBit/s war.
Es ist also möglich.... Ich gehe einfach von einem überlasteten Segment aus.
Und den LAG / LACP sieht man auch gut... Manche Streams sind langsamer, bis 20 MBit/s runter.