It wasn't any one specific port or ports. It was certain ports in combination with certain other ports. More specifically, if traffic was being hairpinned through the router from a Compact to the EPC and then out through the router to the internet (PDN VLAN), and the ports that the Compact, EPC, and internet connection were plugged into were all physically wired to the same switch chip, and if the traffic was being either routed or bridged by the CPU instead of switched by the switch chip, then we had major problems. But if we made sure to only use one port per switch chip for any function, everything was fine. So, although there are some exceptions, to be on the safe side, use any port ether1-ether5 (once one port in that group is used, other ports are unusable), any port ether6-ether10 (same conditions), and then ether11, ether12, and/or ether13 (which are all discrete).
I will dig up the full write-up I composed about the issue, and post it here, in case anyone is interested in reading more details. -- Nathan From: [email protected] [mailto:[email protected]] On Behalf Of Adam Moffett Sent: Saturday, November 19, 2016 9:46 AM To: [email protected] Subject: Re: [Telrad] Issues Nathan, We happen to have a few 1100AHx2. Which ports did you have to avoid to resolve the performance issue? The last two? Thanks, Adam ------ Original Message ------ From: "Nathan Anderson" <[email protected]<mailto:[email protected]>> To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Sent: 11/18/2016 2:00:18 PM Subject: Re: [Telrad] Issues When we had problems with S1 trunking, the primary symptom was as it sounds like you are describing: TCP performance to LTE UEs got worse the more hops out on the network you got. Of course, overall total capacity to a particular UE could be achieved by opening a bajillion parallel TCP sockets, thus demonstrating at a radio link level that things were okay, but that didn't mean that there wasn't a problem. If, as a test, we threw an EPC out at a site where we were seeing the TCP performance issues, those issues vanished completely. In our case, our problem turned out to be blamable on a particular architectural quirk of certain MikroTik router models that happen to incorporate switch chips (our PDN gateway is an 1100AHx2): by being strategic about what specific ethernet ports we used, the problem could be avoided. We never even really managed to get to the point where we were able to completely understand what was happening at a lower level to cause the performance degradation. If we ran unidirectional UDP tests, there was no measureable difference in packet loss before we worked around the issue vs. after. But TCP performance difference was night and day. In one-off cases where we have had a relatively small amount of measureable packet loss on a backhaul that services a Compact, we have seen similar TCP performance issues. The path from the UE all the way to your internet gateway needs to be pristine. -- Nathan From: [email protected]<mailto:[email protected]> [mailto:[email protected]<mailto:[email protected]>] On Behalf Of Justin Davis Sent: Wednesday, November 16, 2016 8:35 PM To: [email protected]<mailto:[email protected]> Subject: Re: [Telrad] Issues Correct. Iperf tests show good results and so does dslreports speedtest with a bunch of streams open it yields great results. Like I said it does not cause actual performance issues just lack of performance on sites like speedtest.net<http://speedtest.net>. Trying to explain that to customers who don't understand and do not have service issues just that their speedtest results are not showing what they are paying for. Even though they can stream 3 movies at one time. We have close to 600 UE's online now and very rarely get calls from our LTE customers. I was just curious if anybody else had experienced or had a solution for the difference in site vs remote eNB's. On Wed, Nov 16, 2016 at 9:25 PM, Matthew Carpenter <[email protected]<mailto:[email protected]>> wrote: Are two EPC are located at the Datacenter. None of our eNB are located in the same location as the Datacenter. I mostly use iperf from the UE back to the Datacenter for a good speedtest. I know customers use speedtest.net<http://speedtest.net> (we have a speedtest.net<http://speedtest.net> service in our Datacenter also). Results between iperf and speedtest.net<http://speedtest.net> are very close to the same. I have neve taken a EPC out to the tower where a eNB is located. Packages we sell on LTE are 5,10,15,20. Most are on the 10 or 15 Mbit plan. Upload is 1 or 2 mbit and most are on the 1 Mbit. So you are saying that the S1 tunnel between the EPC and eNB could be causing an issue? I have not really seen any speed problems until we got to around 25-30UE’s on a eNB, at that point I have been unable to get the upload speeds. You would think on a 80x10 plan (for us only to test with), and the eNB only moving small amount of data I should be able to 60x7, but I only see around 14Mbit x 600K. Matt Carpenter On Nov 16, 2016, at 10:13 PM, Skywerx Support <[email protected]<mailto:[email protected]>> wrote: What about on say for example speedtest.net<http://speedtest.net/> with a UE connected to and eNB where the EPC is located vs a UE connected to an eNB that is not. We are putting 80 UE's on our eNB's with no problem. Performance is always there but we do see a difference with web based speedtests when the eNB's are not at the EPC site. -- Justin Davis COO SkyWerx Industries, LLC On Nov 16, 2016, at 9:06 PM, Matthew Carpenter <[email protected]<mailto:[email protected]>> wrote: We have a new eNB with only 3 UE’s on it, speed tests (iperf) this evening are 63x7.5 4x4TM4 Mode and 2/0 frame config. If I run a iperf on a UE connected to a eNB with 40+ UE’s then I find the download is in the 14-18Mbit range, but the upload is 300-600K. Checking the eNB in Breezeview it shows very little usage at the time I tested. I need to check all eNB if see if they exhibit this same upload speed issues. From little rumblings in the Service Dept I think I will find they all do this. Nick is working on this and I am sure he will come up with a solution soon. Matt Carpenter On Nov 16, 2016, at 8:58 PM, Skywerx Support <[email protected]<mailto:[email protected]>> wrote: We have two 2020 centralized also with 8 eNB talking to one and 7 eNB to the other. Are you seeing similar throughout on all eNB's that are not located where the EPC's are? -- Justin Davis COO SkyWerx Industries, LLC On Nov 16, 2016, at 7:28 PM, Matthew Carpenter <[email protected]<mailto:[email protected]>> wrote: Two Centralized EPCs. Right now they are 1010RPL (2020L), but we are software upgrading them to the 2020CS this week. Matt Carpenter On Wed, Nov 16, 2016 at 8:07 PM, Skywerx Support <[email protected]<mailto:[email protected]>> wrote: Matt, Are you using site based or centralized EPC? Thanks -- Justin Davis COO SkyWerx Industries, LLC On Nov 16, 2016, at 4:43 PM, Matthew Carpenter <[email protected]<mailto:[email protected]>> wrote: From what I have seen running 4573: 1. 4x4TM4 is working good. I have 2 eNB’s using it with good speeds through trees. In one location I saw 63Mbit x 7.3Mbit and a CINR of 27. UBNT 5ghz at this same location was around 79db and we would not install. 2. Better performance. Jury is still out on this one. Like a few of you have stated the Upload speeds are not where they should be 2/0 config and only 1mbit of upload and I only see 700-800K on iperf test. 3. There are some timer changes that extend the amount of time a action happens. Stable, with 7000 and 8000’s both working good, except for the speed issues that I stated above. Telrad support is working on them now. Matt Carpenter Amarillo Wireless On Nov 16, 2016, at 4:37 PM, Ian Fraser <[email protected]<mailto:[email protected]>> wrote: No release notes that I find (?) So what's the benefit ? -- Ian Fraser goZoom.ca<http://gozoom.ca/> Inc. 195 Libby's Rd. McNab-Braeside K7S0E1 877(613) 622 0093 ext 21<tel:%28613%29%20622%200093%20ext%2021> On 16/11/2016 4:39 PM, Jeremy Austin wrote: On Wed, Nov 16, 2016 at 12:37 PM, Ian Fraser <[email protected]<mailto:[email protected]>> wrote: I'm still on 4013. Has 4573 fixed anything for anyone ? Like the high upload utilization ? I'm using sf2/ssf0. Changing subframe is not an option for me - I need to co-exist with a another provider who will never change. 4573 seems fairly stable for us with 7ks running .105 and 8ks. Didn't fix the upload KPI or (as far as we can tell) performance. -- Jeremy Austin (907) 895-2311 (907) 803-5422<tel:%28907%29%20803-5422> [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]<mailto:[email protected]> _______________________________________________ Telrad mailing list [email protected]<mailto:[email protected]> http://lists.wispa.org/mailman/listinfo/telrad _______________________________________________ Telrad mailing list [email protected]<mailto:[email protected]> http://lists.wispa.org/mailman/listinfo/telrad _______________________________________________ Telrad mailing list [email protected]<mailto:[email protected]> http://lists.wispa.org/mailman/listinfo/telrad -- Matthew Carpenter 806-316-5071<tel:806-316-5071> office 806-236-9558<tel:806-236-9558> cell [https://docs.google.com/uc?export=download&id=0BxDRq5UV7HPOaEM4LXVaVnk5cWM&revid=0BxDRq5UV7HPOTDdiVjM0TXRIc3ZzMXVUVDdDVjBiaFU0bHJNPQ] _______________________________________________ Telrad mailing list [email protected]<mailto:[email protected]> http://lists.wispa.org/mailman/listinfo/telrad _______________________________________________ Telrad mailing list [email protected]<mailto:[email protected]> http://lists.wispa.org/mailman/listinfo/telrad _______________________________________________ Telrad mailing list [email protected]<mailto:[email protected]> http://lists.wispa.org/mailman/listinfo/telrad _______________________________________________ Telrad mailing list [email protected]<mailto:[email protected]> http://lists.wispa.org/mailman/listinfo/telrad _______________________________________________ Telrad mailing list [email protected]<mailto:[email protected]> http://lists.wispa.org/mailman/listinfo/telrad -- Sincerely, Justin Davis SkyWerx COO 970-731-9790
_______________________________________________ Telrad mailing list [email protected] http://lists.wispa.org/mailman/listinfo/telrad
