我可以在相同Cloud Service中的兩臺Standard_D2_V2 West Europe Classic(ASM)VM之間獲得亞毫秒延遲,並將雲服務分配給一個相關組。如何在同一(區域)VNET上的兩個Azure VM(ARM)之間獲得亞毫秒的延遲?
我還可以在Azure VM Scale Set中的VM之間獲得亞毫秒的延遲。
但我似乎無法獲得兩個Azure的ARM虛擬機之間的子毫秒的延遲(共享相同的(區域)VNET)。有沒有一種方法可以在ARM虛擬機之間可靠地獲得亞毫秒級的延遲,而無需像使用Azure classic/ASM中的親和性組一樣部署爲Scale Set?
下面是我對各種情況的測量。我已經多次部署了各種方案,以盡我所能排除幸運與不幸的部署。
兩個ASM(經典)虛擬機中分配給親和基團相同的雲服務(通過VNET未連接!!)
PS C:\Users\simon> .\psping -l 1k -n 10000 -h 13.81.205.180:6
PsPing v2.10 - PsPing - ping, latency, bandwidth measurement utility
Copyright (C) 2012-2016 Mark Russinovich
Sysinternals - www.sysinternals.com
TCP latency test connecting to 13.81.205.180:60123: Connected
10005 iterations (warmup 5) sending 1024 bytes TCP latency test: 100%
TCP roundtrip latency statistics (post warmup):
Sent = 10000, Size = 1024, Total Bytes: 10240000,
Minimum = 0.11ms, Maxiumum = 6.23ms, Average = 0.45ms
Latency Count
0.11 6042
0.43 2721
0.75 846
天青ARM VM量表組VM
PS C:\Users\simon> .\psping -l 1k -n 20000 -h 10.0.0.7:6
TCP roundtrip latency statistics (post warmup):
Sent = 20000, Size = 1024, Total Bytes: 20480000,
Minimum = 0.18ms, Maxiumum = 9.62ms, Average = 0.48ms
Latency Count
0.18 19733
0.68 178
天青ARM VM天青ARM VM上相同(區域)的vnet
PS C:\Users\simon> .\psping -l 1k -n 20000 -h 10.0.0.7:6
TCP roundtrip latency statistics (post warmup):
Sent = 20000, Size = 1024, Total Bytes: 20480000,
Minimum = 1.59ms, Maxiumum = 10.60ms, Average = 1.81ms
Latency Count
1.59 19401
2.07 379
2.54 98
你有沒有考慮網絡RDMA密集型vms? https://docs.microsoft.com/en-us/azure/virtual-machines/virtual-machines-windows-a8-a9-a10-a11-specs#access-to-the-rdma-network,另外,什麼是錯的與規模集。 – 4c74356b41