PING replies - how many dropped packets isn't good?

[gerardsweeney]

Hi, folks.

 

Just looking for a sort of straw poll here.

 

Our High schools are connected to the Data Centre via 200Mb links.

Our High schools have (on average) 600 wired PCs and a few hundred laptops running on 8021x.

 

From our High Schools - if we do a looped PING of (say) the server which hosts our home folders (which lives in the DC), we're seeing

 

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=3ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=3ms TTL=122

Reply from 10.11.12.13: bytes=32 time=3ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Request timed out.

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=3ms TTL=122

Request timed out.

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=3ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=3ms TTL=122

Reply from 10.11.12.13: bytes=32 time=2ms TTL=122

Reply from 10.11.12.13: bytes=32 time=3ms TTL=122

 

And so forth.

Packets: Sent = 215, Received = 208, Lost = 7 (3% loss),

Approximate round trip times in milli-seconds:

Minimum = 2ms, Maximum = 4ms, Average = 2ms

 

 

My question is - is that an acceptable packet loss?

I know in an ideal world, we shouldn't drop any - but what's the Real World realistic answer?

 

Thanks,

Gerard

[TomHD]

I'm happy with anything below 5%

 

Just had a look at traffic and I have 1 switch dropping 16/10204 packets, the rest are dropping none

Edited June 6, 2019 by TomHD

[HPlum78]

That's an interesting question, 5% is a lot! And what is "acceptable" depends on what the traffic is. 5% while streaming audio/ video would be i suppose acceptable but not the best experience, where <1% is what you would be looking for.

 

Could do with narrowing down where the packets are being dropped, so is it at the switch/ ap or at the server so on and....

[TomHD]

I should have been clearer I guess sustained loss at 5% is not good, but I won't really investigate anything below 5% unless it is sustained and affecting quality of service

[HPlum78]

@TomHD I was not having a dig at you there (you had a smilely face) just used that figure as you already had.

[TomHD]

I didn't feel any malice in your response <3 but after a quick Google its clear my 5% is a tad high and most are saying 0-1% is OK

 

Looking I am dropping <20 packets per second across the entire network so well within acceptable numbers... sorry, not meaning to rub it in @OP

Edited June 6, 2019 by TomHD

[Oaktech]

I'm dropping 0% on internal connections. I'm dropping approx 0.5-1% to a mapped drive in a data centre. 5% would be the top end of what I'd accept before I got the link supplier to look into that for me.

[pete]

Are you consistently experiencing 3% loss (say across weeks/months) or is this a new thing?

[RobD]

If you see packet loss that does not get recovered in later hops, it would suggest a potential problem. Is the data centre local or somewhere else? The problem with ping is FW's often dont prioritise it so can and will drop ping if needed i.e. hight bandwidth and QoS kicks in.

 

This doesnt really help fix the issue but if you want to use a different tool to measure, then WinMTR is really nice for bing:

https://github.com/White-Tiger/WinMTR/releases

 

Also consider using iperf to test tcp and bandwidth, might get more reliable results.

 

https://bohemiangrove.co.uk/iperf/

Edited June 6, 2019 by RobD

[gerardsweeney]

Are you consistently experiencing 3% loss (say across weeks/months) or is this a new thing?

 

Truthfully - I don't know.

We've never really had much call to test the connectivity back to the DC because things have (generally) been OK.

It was just that one of my colleagues had noticed things were slow when accessing EXEs in their home folder while in their school and noticed the dropped packets.

I did a looped PING in my school, and the results were the same - so figured I should ask here before I go making myself unpopular by suggesting that things are not well with the WAN

[PotNoodleTech]

You shouldn't be getting any packet loss at all. You should always be getting 0% loss on internal connections and I wouldn't be expecting loss on external connections either.

 

Heres an internal network ping:

Reply from 10.[iNTERNAL]: bytes=32 time<1ms TTL=254

Reply from 10.[iNTERNAL]: bytes=32 time<1ms TTL=254

 

Ping statistics for 10.[iNTERNAL]:

Packets: Sent = 50, Received = 50, Lost = 0 (0% loss),

Approximate round trip times in milli-seconds:

Minimum = 0ms, Maximum = 1ms, Average = 0ms

 

Here's a ping to datacentre:

Reply from 10.[DATACCENTRE]: bytes=32 time=4ms TTL=250

Reply from 10.[DATACCENTRE]: bytes=32 time=14ms TTL=250

 

Ping statistics for 10.[DATACCENTRE]:

Packets: Sent = 50, Received = 50, Lost = 0 (0% loss),

Approximate round trip times in milli-seconds:

Minimum = 4ms, Maximum = 48ms, Average = 11ms

 

Obs ping is worse going up to the DC via leased line but zero loss and that's how it should be!

 

If all your schools are getting the same loss then the problem must be in your datacentre or your datacentres leased line.

[localzuk]

Yeah, seeing packet loss is indicative of something being wrong.