Funnily enough when we ping a machine of his, we seem to be jumping up and down in the ping rates. At times as low as 4ms and as high as 55ms over a direct frame.. What could it be? RAM? On Mon, 1 Apr 2002, Craig Spiers wrote:
If he pings www.net4u.co.nz and gets no packet loss, his circuit is fine..
If he pings www.winz.co.nz and gets packet losss the problem is between you and winz.co.nz
Quoting Sahil Gupta - Net4U Limited
: As below with "###"
On Sun, 31 Mar 2002, Malcolm Lockyer wrote:
Let me just jump in here, since I was on site..
On Sun, 31 Mar 2002, Craig Whitmore wrote:
A few things which may help (not in any order)
1) Look at Stats of Frame-Relay Connection (Any Errors??) (Geting Telecom
to
do BIR TEST)
Don't know - forgot the router's ip address, so I can't telnet in and check - sahil will though. (just do a show frame-relay pvc right?)
2) Are they getting what they are paying for? Are they getting CIR before before any errors occur? Remember the 384K is not promised, only 64K is promised, any extra bandwidth depends on ALOT of things.
Don't know - originally it was being traffic shapped at 384K - which I thought could give the same behavior (ie - the cisco over subscribes the bandwidth). After removing the traffic shaping - the pings were a little lower.
### I think that the guy needs to up his PVC, I feel he is pushing through more than the CIR of the circuit and the circuit seems to be falling close to the CIR most of the times.
3) Define "Packet loss"? slow downloads?
Well - we've only heard from the asian guy (no offence, but kinda a
little
bad at english) that they get packet losses up to 45%, and when they get packet loss the whole internet is "slow", and people get the "cannot find server errors".
### He gets packet losses the moment they begin to download, he unfortunately pings "www.winz.co.nz" and "www.net4u.co.nz". While he pings www.net4u.co.nz he doesn't get any packet loss even while downloading, only thing when he jumps out of our network and onto the WWW he begins facing packet loss i.e. while pinging www.winz.co.nz
4) Check Settings on Routers on either end (Are both ends different
hardware
(Any compatibility issues between hardware))?
Nope - there is a cisco 2501 and a 2503.
5) Check Telecom has Circuit defined right.. Dunno - we should ask though.
###I think Telecom have set it right. Only thing using MRTG I have seen them pull upto 440k. Which is why I decided to traffic-shape their link as it was affecting other users.
6) Are routers running at high CPU? For example 2500's can't handle
much
bandwidth thru them without going Kaput and slowing things down
Umm.. both the routers were running at about 45ish% cpu usage. However, there was very little ram on the 2503 (4mb) - and it is handling about a meg of shaped traffic now - could that be it?
###Today, I went and swapped the 2 DRAM Chips around (BTW, Malcolm that bent case isn't shutting.. you may need to come around) it seems better, but not the BEST. I have also ordered in some 16MB Parity Kingston RAM, I'll give that a go.
7) Check External Devices (ethernet convertor on 2500/Serial Cables/Ehternet)
The AUI trancivers seem fine.
8) On Frame Relay connections, you have not "over subscribed" the Access
(ie
trying to total of 2M CIR's over a 1M Circuit)?
Well - its possible with the traffic shaping rules we were using (ie - shaped to the PIR - when it is running a 25% CIR)
###Nope, definately NOT. Our end has 3 PVC's off our 1Meg Access and they sum up to 1024k
9) MTU set right on both ends?
Dunno - Sahil - check
### They both say "0 01"
Thats just some info about what happenes - any more ideas?
Thanks
Thanks Craig Whitmore Orcon Internet http://www.orcon.net.nz
--snip--
Best regards,
Malcolm Lockyer
-- owner/founder/managing director
-= Nuron Net =- You're on the net!
- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
- To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
On Mon, Apr 01, 2002 at 07:52:19PM +1200, Sahil Gupta - Net4U Limited wrote:
Funnily enough when we ping a machine of his, we seem to be jumping up and down in the ping rates. At times as low as 4ms and as high as 55ms over a direct frame..
What could it be? RAM?
You know, it could just be straight congestion. Did you thin kof that? When you congest a link too much, you're going to get packet loss. A 55ms ping probably just means that the link was being used when the ping was sent. If there are lots of clients on a slow link, then you should *expect* a lot of variation in latency across the link, maybe 200 ms on a link as slow as that, maybe even more; things such as tcp window sizes and speed of sending host etc all matter there. I'd make sure that no hosts are using large tcp window sizes, Linux is especially bad at defaulting to large window sizes that will overly congest a slow link with a single tcp session. 64kbytes in the air, when you can only send a maximum of 48kbytes (- overheads) a second, could straight away mean you're over a second in latency from a fast/close link. Ben. - To unsubscribe from nznog, send email to majordomo(a)list.waikato.ac.nz where the body of your message reads: unsubscribe nznog
participants (2)
-
Ben Aitchison
-
Sahil Gupta - Net4U Limited