I am curious how you came to the conclusion that ICMP has anything to do with 
it?  We also do ICMP probes to UEs once a minute and we do not (seemingly) have 
this problem.  We will also occasionally run continuous pings to UEs that we 
are actively troubleshooting, and the more technically-savvy customers of 
course will also run their own ping tests.  If ICMP were causing devastation of 
this magnitude then I think we would be seeing it...

We do not use SNMP to UEs, only TR-069.  So maybe SNMP is somehow the key.  
Though I can't imagine why SNMP would be any different than other UDP traffic 
(e.g., VoIP, DNS, etc.).

Have you tried shutting off only the SNMP polling for a few cycles, but leaving 
the ICMP polling tests running?

-- Nathan

From: [email protected] [mailto:[email protected]] On Behalf Of 
Jeremy Austin
Sent: Monday, April 03, 2017 7:43 AM
To: [email protected]
Subject: Re: [Telrad] Uplink throughput again


On Thu, Feb 16, 2017 at 12:24 AM, Nathan Anderson 
<[email protected]<mailto:[email protected]>> wrote:
Jeremy mentioned his periodic traffic dips to me recently off-list.

I have found the source of the dips, if not the final cause. For a refresher, 
this was a 10-second long forced dip in traffic — nearly to nil on the uplink, 
and 75-80% reduction on downlink. Easily replicable by pushing iperf and 
watching interfaces as well as per-second results.

ICMP+SNMP polling once per minute was causing them. My guess is still that it's 
EPC related, as ICMP and SNMP results do get returned — eventually. The poller 
puts out a peak of 600 pps for the whole network, so the actual load for the 
LTE bearer ought to be much less than that.

Anyone else doing ICMP or SNMP polling, perhaps at longer intervals? (My guess 
is that we wouldn't have caught this at 5 minute intervals or longer.)

--
Jeremy Austin

(907) 895-2311 office
(907) 803-5422 cell
[email protected]<mailto:[email protected]>

Heritage NetWorks
Whitestone Power & Communications
Vertical Broadband, LLC

Schedule a meeting: http://doodle.com/jermudgeon
_______________________________________________
Telrad mailing list
[email protected]
http://lists.wispa.org/mailman/listinfo/telrad

Reply via email to