Web Hosting Talk : Web Hosting Main Forums : Colocation and Data Centers : Will Pinging reveal the speed of the DataCenter

[Eddy2010]

Hi

will pinging reveal the speed of the datacenter network.

If i ping from webcoast, east coast and in some location in Asia may be india.

Just calculating the 3 averages and can i conclude the speed. I know it depends on network, hops and distance. Try to understand how far this test will be helpful.

So just finding the speed, I can decide which datacenter to choose to buy the dedicated server.

Thanks

[luki]

Ping will give you an estimate of the latency between the end points (i.e. path length), and of the packet loss (there shouldn't be any). I say estimate because often ICMP Pings have lower priority than other traffic. It will not give you any indication of the throughput or bandwidth capacity (i.e. thickness of the pipe).

For TCP, the latency does affect the maximum achievable throughput. The perceived and theoretical speed of a single TCP connection will be better with lower latency. But we're talking about differences on the scale of 20 ms vs. 100 ms, not 20 ms vs. 25 ms.

See also: http://en.wikipedia.org/wiki/Network..._and_protocols

[Mike Johnson]

lol luki, that's the best signature ever.

[LiquidWebBenny]

I like MTR, too. It combines traceroute with pings, and gives you all kinds of data.

http://www.bitwizard.nl/mtr/

[devonblzx]

Ping times don't represent network speed in anyway. Also, don't judge packet loss by pings. Most routers and such treat ICMP packets with the lowest priority therefore packet loss through pings does not represent packet loss through TCP/UDP.

[iTom]

Speed can mean a few things:

Latency, this is what ping measures, how long it takes for a packet to go from one place to another.

Bandwidth, this is the amount of data that you can move in a time, such as 1GB in a minute or a second. The two examples of bandwidth are extremely different, however can maintain the same ping/latency..

You need to work out what your application is effected by, and then make tests to match.

For a test of bandwidth I would suggest asking the data centre for a test file which you can download.

[BradQ]

Originally Posted by devonblzx

Ping times don't represent network speed in anyway. Also, don't judge packet loss by pings. Most routers and such treat ICMP packets with the lowest priority therefore packet loss through pings does not represent packet loss through TCP/UDP.

That's really not true. Most routers will deprioritize ICMP directed *at* them, not passing through. The former isn't routing and thereby not handled by ASICs...and core routers don't have a surplus of CPU to deal with your need to ping them.

There's really no reason to expect ICMP to the destination to be any slower than any other form of traffic. Layer 4 filtering would be pretty weird to see at the core level. The effect you're talking about is relevant for intermediary hops visible with a traceroute, not the final one.

But luki is right. Rely on test file downloads from each datacenter's physical sites, not your ping times if you truly mean "speed" as throughput.

[Eddy2010]

Yes that is the best bet, downloading test files. Did make sense on how your web pages will load.

thanks

[acezonehosting]

A lot of hosts will have a Speedtest.net app or file download to let you see what kind of speed to expect from their service. a ping is a good start, but not really a good tell. If you are running *nix you can use the My Trace Route "mtr" command and it will tell you live ping times and packet loss to any given ip. The command is

mtr IPADDRESS

Great stuff.