Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

What network bandwidth should I expect on GCE?

I want to host a webradio station, using an icecast2 server, on a google compute engine instance in the europe-west1-a zone. This service is mainly bandwidth limited so I've searched the documentation to found an estimate of how much bandwidth is available (from my server to the internet) and I couldn't found it.

I've run som preliminary tests, by simulating listeners with curl processes on a single remote server with 100 Mbps of bandwith. My tests started failing at 500 listeners. The stream was encoded at 64kbps so the server was outputing ~32Mbps. From the way the test is run, I can't conclude the bandwidth is the limiting factor but still I'd like to know how much bandwidth I should expect to estimate the number of concurrent listeners I can have.

like image 797
Yann Biancheri Avatar asked Aug 27 '13 07:08

Yann Biancheri


People also ask

What is GCE network?

Google Compute Engine (GCE) is an infrastructure as a service (IaaS) offering that allows clients to run workloads on Google's physical hardware. Google Compute Engine provides a scalable number of virtual machines (VMs) to serve as large compute clusters for that purpose.

What is network bandwidth in cloud computing?

Bandwidth specifically refers to the capacity at which a network can transmit data. For example, if the bandwidth of a network is 40 Mbps, it implies that the network cannot transmit data faster than 40 Mbps in any given case.

How do I increase my bandwidth on Google cloud?

By selecting larger N2, N2D, or C2 machine types you can reach up to 100 Gbps of TIER 1 bandwidth throughput.

What is the network bandwidth?

Network bandwidth is a measurement indicating the maximum capacity of a wired or wireless communications link to transmit data over a network connection in a given amount of time. Typically, bandwidth is represented in the number of bits, kilobits, megabits or gigabits that can be transmitted in 1 second.


2 Answers

100 Megabit/second is expected. Your test might not be accurate if you are testing from single host, how do you know that the bottleneck is not the receiving end.

Here is my test 1 vCPU, 3.8 GB memory GCE instance:

  • install nginx
  • download debian-update-7.1.0-amd64-CD-1.iso to the instance www root

use two remote servers in Asia as clients parallel with 5 threads each using axel http://axel.alioth.debian.org/

  • host 1: Downloaded 639.1 megabytes in 1:31 seconds. (7132.22 KB/s)
  • host 2: Downloaded 639.1 megabytes in 1:41 seconds. (6452.50 KB/s)

total: 13584.72 KB/s = 108.67 Mbits/s

like image 148
Feczo Avatar answered Nov 02 '22 22:11

Feczo


Test result from my 'g1-small' instance:

CPU model :  Intel(R) Xeon(R) CPU E5-2689 0 @ 2.60GHz
Number of cores : 1
CPU frequency :  2599.998 MHz
Total amount of ram : 1699 MB
Total amount of swap : 0 MB
I/O speed :  55.1 MB/s

Download speed from CacheFly, CDN: 16.5MB/s
Download speed from Linode, Tokyo, JP: 10.1MB/s
Download speed from Linode, London, UK: 14.7MB/s
Download speed from Linode, Newark, NJ: 27.5MB/s
Download speed from Linode, Atlanta, GA: 17.8MB/s
Download speed from Linode, Dallas, TX: 7.57MB/s
Download speed from Linode, Fremont, CA: 22.2MB/s
Download speed from SoftLayer, Dallas, TX: 52.6MB/s
Download speed from SoftLayer, Seattle, WA: 28.0MB/s
Download speed from SoftLayer, Washington, DC: 35.6MB/s
Download speed from SoftLayer, Houston, TX: 47.3MB/s
Download speed from SoftLayer, San Jose, CA: 26.6MB/s
Download speed from SoftLayer, Singapore, Singapore: 8.09MB/s
Download speed from SoftLayer, Amsterdam, NL: 14.9MB/s
like image 25
billzhong Avatar answered Nov 03 '22 00:11

billzhong