有線網絡中的丟包是非常低的。我之前在ns3上仿真,網絡中的數據流採用擁塞控制機制,可以使得網絡很穩定。根據Sprint.net中統計的網絡性能參數[6],網絡中的丟包率低的令人讚歎!
無線網絡中的丟包情況,是什麼樣的,我也不知道,大概也是非常低的。
當前網絡中tcp數據流,採用的擁塞控制是cubic,把丟包當做擁塞信號。隨機丟包對數據流的吞吐量有負面影響。根據我的仿真,當有0.03的丟包率,TCP的速率就很低了。
根據[1]的測試,感覺丟包率在其測試的無線網絡中很嚇人!
We run VoIP traffic between our server and a laptop via wifi hotspot or USB modem and 3G internet connection. Results are different for the 4 operators:
Tele2 3G during 45 hours (wifi, 1 G.729 channel): uptime = 99.89%, worst MOS = 2.5, max packet loss = 6.94%
MTS 3G during 19 hours (wifi, 1 G.729 channel): uptime = 100%, worst MOS = 3.45, max packet loss = 16.34%
MTS 3G during 96 hours (USB, 1 G.729 channel): uptime = 99.94%, worst MOS = 2.41, max packet loss = 6.06%
MTS 3G during 96 hours (USB, 3 G.729 channel): uptime = 99.87%, worst MOS = 1, max packet loss = 27.94%
Megafon 3G during 24 hours (wifi, 1 G.729 channel): uptime = 100%, worst MOS = 3.89, max packet loss = 0.47%
Yota 4G during 1 hour (wifi, 1 G.729 channel): uptime = 0%, worst MOS = 1, max packet loss = 100%
同時[2]也聲明:it it is not unusual for wireless networks to experience high packet loss rates
來自[3]的數據:
Country | Network Latency | Packet loss |
---|---|---|
UK WiFi | 10 ms | 0.2 % |
UK Mobile (4G) | 25 ms | 0.2 % |
US WiFi | 25 ms | 0.2 % |
USMobile (4G) | 50 ms | 0.5 % |
Germany WiFi | 25 ms | 0.2 % |
GermanyMobile (4G) | 50 ms | 0.2 % |
Japan WiFi | 10 ms | 0.1 % |
Japan Mobile (4G) | 50 ms | 0.2 % |
[3]博文的作者還構造了一個網絡模擬器[4]。
[5]設計了一個工具,能夠探測出,路徑上的瓶頸鍊路到底是在wireless段還是在wlan段。
文中聲明:
the cellular last-hop link is the bottleneck in 68.9% and 25.7% of the total bottleneck cases for 3G and LTE clients, respectively, suggesting that there is a significant fraction of cases where the poor performance experienced by the user is due to the WAN and could potentially be routed around
[1] http://startrinity.com/VoIP/NetworkTester/IpNetworksPerformanceResearch.aspx
[2] The Effect of Packet Loss on Redundancy Elimination in Cellular Wireless Networks
[3] Real-world latency and packet loss
[4] network-emulator https://gitlab.com/codavel/network-emulator/tree/master
[5] QProbe: Locating the Bottleneck in Cellular Communication
[6] SprintLink