Is 0ms of ping possible?

Forums › New Zealand Broadband › Is 0ms of ping possible?

Paulthagerous

698 posts

Ultimate Geek

Trusted

#58017 4-Mar-2010 14:52

Did a speedtest and, aside from speeds being less than 1/20th what they usually are, noticed ping was 0ms:

Now I know this is not the best measure and is quite possibly higher but: is 0ms technically possible under ideal conditions, or can it never be 0? Can it even be 0 if we allow for rounding to the nearest ms?

Detruire

1611 posts

Uber Geek

#304441 4-Mar-2010 15:45

0ms would most likely indicate a latency lower than 1ms that has been rounded down.
That would be possible.

An actual latency of 0ms would indicate something akin to teleportation.

rm *

GoodSync

GoodSync. Easily back up and sync your files with GoodSync. Simple and secure file backup and synchronisation software will ensure that your files are never lost (affiliate link).

oldmaknz

536 posts

Ultimate Geek

#304444 4-Mar-2010 15:47

I doubt it's exactly 0 (null), it may be <=0.49ms though. Uni's tend to have epic internet (and that Speedtest server must be extremely close, if not local).

michaeln

235 posts

Master Geek

Trusted

#304448 4-Mar-2010 15:56

No, it's not strinctly possible, but rounded to the nearest ms it's very possible.

Speed of light 299,792,458 m/s
Speed in copper 75% approx
Speed in copper 224,844,344 m/s
Distance (1-way) 10 km
Distance (1-way) 10,000 m
Travel time 0.00 s
Travel time 0.04 ms

Packet size 64 B
Packet size 512 b
Line rate 10 Mbps
Line rate 10000000 bps
Serialisation delay 0.00 s
Serialisation delay 0.05 ms
1-way 0.10 ms
RTT 0.19 ms

I.e., for up to 10km and on a 10Mbps Ethernet, you'd expect a minimum of 0.19ms round trip for 64byte packets. Actual latency could be higher, for various reasons, but given that processing speed on modern routers is microseconds per packet, you are not going to get above the .5ms needed to round to 1ms.

Note that ADSL imposes several ms of latency due to forward error correction, and cable somewhat less, so you should never see zero ms on current consumer broadband. You on the other hand appear to be fibre connected.

vinnieg

2260 posts

Uber Geek

Trusted

#304459 4-Mar-2010 16:13

Paulthagerous: Did a speedtest and, aside from speeds being less than 1/20th what they usually are, noticed ping was 0ms:

Now I know this is not the best measure and is quite possibly higher but: is 0ms technically possible under ideal conditions, or can it never be 0? Can it even be 0 if we allow for rounding to the nearest ms?

Nah not possible at the Uni here,

while we do have a 1gb link between campuses and to citylink
The new proxy that was implemented is caching it I think, as well as sharing the load, because when students come back, it gets owned!!!, using the mac Manual staff proxy I get(I'm using a PC):

I have moved across the ditch. Now residing in Melbourne as a VOIP/Video Technical Trainer/Engineer.

Paulthagerous

698 posts

Ultimate Geek

Trusted

#304469 4-Mar-2010 16:42

michaeln: No, it's not strinctly possible, but rounded to the nearest ms it's very possible.

I.e., for up to 10km and on a 10Mbps Ethernet, you'd expect a minimum of 0.19ms round trip for 64byte packets. Actual latency could be higher, for various reasons, but given that processing speed on modern routers is microseconds per packet, you are not going to get above the .5ms needed to round to 1ms.

Note that ADSL imposes several ms of latency due to forward error correction, and cable somewhat less, so you should never see zero ms on current consumer broadband. You on the other hand appear to be fibre connected.

I assume that of the 1Gb, I have more than enough at any time. It is my understanding it is connected directly to Citylink with Fibre all the way, and well less than 10km away. Question now is, in real life with real life interferences etc is it likely to be under the 0.5ms threshold?

vinnieg:
Nah not possible at the Uni here,

while we do have a 1gb link between campuses and to citylink
The new proxy that was implemented is caching it I think, as well as sharing the load, because when students come back, it gets owned!!!, using the mac Manual staff proxy I get(I'm using a PC):

My only query would be this: if the speedtest was cached on a uni server, should my test not have recorded a blazing speed as opposed to the slowest I have ever seen it by at least 10 time? Or am I not getting the point?

Detruire

1611 posts

Uber Geek

#304472 4-Mar-2010 16:47

It's possible that the cache's connection/s to the network is/are being saturated.

rm *

Paulthagerous

698 posts

Ultimate Geek

Trusted

#304473 4-Mar-2010 16:51

Will try tomorrow morning at 8 am while all the other students are still asleep lol. Hopefully can replicate ping with blazing speed

vinnieg

2260 posts

Uber Geek

Trusted

#304475 4-Mar-2010 16:57

I think it's something to do with speedtest being a really unreliable measure on a proxy server connection. Not exactly sure what the proxy is doing, but some things are going really weird. Think the speedtest pings have been up and down all day. I think a better test to try is probably the Slingshot NZ online speed test.. although i don't think it'll show the ping, since I think it's blocked on the proxy.

I think it still must be the cache causing hte false 0ms ping test...because everytime I turn it to manual it's got a ping of about 6+ or so, 4-10 is about right for the distance we are from Telstra. Not sure why speedtest.net does it, but running a tracert to Telstraclear brings back a ping of about 8ms for the first hop which sounds about right

I have moved across the ditch. Now residing in Melbourne as a VOIP/Video Technical Trainer/Engineer.

rvangelder

352 posts

Ultimate Geek

#304476 4-Mar-2010 16:59

maknz: I doubt it's exactly 0 (null), it may be <=0.49ms though.

a real geek might have said <0.5ms, in case their ping was 0.491ms

Screeb

688 posts

Ultimate Geek

#304482 4-Mar-2010 17:13

is 0ms technically possible under ideal conditions, or can it never be 0? Can it even be 0 if we allow for rounding to the nearest ms?

Just to spell it out, latency is the time taken for an internet signal to get from one computer to another. This is ALWAYS greater than 0 - after all, no information can move faster than the speed of light. 0ms is either rounding (very nearby server (unlikely) or cache) or a bug.

oldmaknz

536 posts

Ultimate Geek

#304487 4-Mar-2010 17:21

rvangelder:
maknz: I doubt it's exactly 0 (null), it may be <=0.49ms though.

a real geek might have said <0.5ms, in case their ping was 0.491ms

Could've, should've, would've.

Paulthagerous

698 posts

Ultimate Geek

Trusted

#304490 4-Mar-2010 17:28

Screeb:
is 0ms technically possible under ideal conditions, or can it never be 0? Can it even be 0 if we allow for rounding to the nearest ms?

Just to spell it out, latency is the time taken for an internet signal to get from one computer to another. This is ALWAYS greater than 0 - after all, no information can move faster than the speed of light. 0ms is either rounding (very nearby server (unlikely) or cache) or a bug.

Figured as much but wanted to check :)

wjw

162 posts

Master Geek

#304799 5-Mar-2010 15:12

It's got to be a proxy issue.

From a linux box on the above network I get:

PING alias.snap.net.nz (202.37.101.17) 56(84) bytes of data.
64 bytes from alias.snap.net.nz (202.37.101.17): icmp_seq=1 ttl=62 time=0.405 ms
64 bytes from alias.snap.net.nz (202.37.101.17): icmp_seq=2 ttl=62 time=0.351 ms
64 bytes from alias.snap.net.nz (202.37.101.17): icmp_seq=3 ttl=62 time=0.323 ms
64 bytes from alias.snap.net.nz (202.37.101.17): icmp_seq=4 ttl=62 time=0.360 ms
64 bytes from alias.snap.net.nz (202.37.101.17): icmp_seq=5 ttl=62 time=0.445 ms

Windows ping just shows <1ms

vinnieg

2260 posts

Uber Geek

Trusted

#304811 5-Mar-2010 15:23

Yeah it is, we started implementing Bluecoat early in Jan(Dec for the trials), seems to be working ok...

I have moved across the ditch. Now residing in Melbourne as a VOIP/Video Technical Trainer/Engineer.

Detruire

1611 posts

Uber Geek

#304814 5-Mar-2010 15:30

vinnieg: Yeah it is, we started implementing Bluecoat early in Jan(Dec for the trials), seems to be working ok...

Just don't let the license expire.

rm *