Good to know. I have some brand new ones sitting around that I could swap in. Thanks On Dec 11, 2014 8:39 AM, "Mike Hammett via Af" <af@afmug.com> wrote:
> You have old, old units. The new ones do 2024 or better. Still Rocket Ms. > They changed that 2 - 3 years ago. > > > > ----- > Mike Hammett > Intelligent Computing Solutions > http://www.ics-il.com > > ------------------------------ > *From: *"Kade Sullivan via Af" <af@afmug.com> > *To: *af@afmug.com > *Sent: *Thursday, December 11, 2014 8:35:41 AM > *Subject: *Re: [AFMUG] EoIP over fiber - high latency? > > So looks like this may be a reason not to use UBNT stuff for our backup > links. Looks like the highest I can set the MTU is 1515 on a couple units > and 1524 on another. Neither capable of 1528 or more. > > I'll have to find some brand new hardware and see if it can go higher. > > How big of a performance hit are we talking here? Potentially requiring > double the pps to move the same amount of large packets? I could that > potentially being a pretty big problem. > > > On Wed, Dec 10, 2014 at 4:49 PM, Shayne Lebrun via Af <af@afmug.com> > wrote: > >> To my understanding, it works like this: >> >> >> >> Say you take an IP packet coming into ether1, and it’s full MTU; 1500 >> bytes. >> >> >> >> Now, you want to bridge ether1 to an EoIP tunnel. EoI is GRE, and >> there’s a 28 byte overhead for the GRE encapsulation. Now you have a 1528 >> byte packet. >> >> >> >> Unless every device between that router and the EoIP endpoint has layer2 >> MTUs of at least 1528 bytes, you’re going to transmit two packets to move >> that one original packet. One packet will have something like 1472 bytes >> of the original packet, plus GRE overhead for 1500, and one will have the >> remaining 28 bytes of the original packet, plus 28 GRE overhead, so, >> something like 56 bytes. >> >> >> >> This introduces the obvious slowdowns, as well as not so obvious ones, >> like maybe you have a device in the middle that’s not so good at PPS. Or >> that queues up small packets into one big air frame, and therefore you’re >> waiting for reassembly on the far end. >> >> >> >> Now, if you’re going from a 1500 byte LAN across a 9000 byte fiber >> connection, you’ll not notice this. If you’re going to a satellite office >> behind DSL with PPPoE, or a cable modem, or whatever, you’re going to >> notice. >> >> >> >> >> >> *From:* Af [mailto:af-boun...@afmug.com] *On Behalf Of *Kade Sullivan >> via Af >> *Sent:* Wednesday, December 10, 2014 5:17 PM >> *To:* af@afmug.com >> *Subject:* Re: [AFMUG] EoIP over fiber - high latency? >> >> >> >> Could you elaborate on this? We have a couple EOIP links across "other" >> networks and have never adjusted the MTU anywhere. I just pulled up the >> EOIP interfaces on each router and they are all set for 1500. Should we be >> increasing this number as a best practice when building EOIP Tunnels? >> >> >> >> On Wed, Dec 10, 2014 at 3:52 PM, Shayne Lebrun via Af <af@afmug.com> >> wrote: >> >> Bear in mind that unless you’ve increased your MTU from end to end, or >> dropped the MTU on your two devices that the EoIP are bridging, you’re >> going to get packet fragmentation. >> >> >> >> Otherwise, what RouterOS version? >> >> >> >> *From:* Af [mailto:af-boun...@afmug.com] *On Behalf Of *Erich Kaiser via >> Af >> *Sent:* Wednesday, December 10, 2014 4:25 PM >> *To:* af@afmug.com >> *Subject:* [AFMUG] EoIP over fiber - high latency? >> >> >> >> So I have an EoIP tunnel setup over two fiber connections for a customer, >> I am seeing high latency over the tunnel any idea? MTU Issue? Using >> RB1100AHx2 on both ends. >> >> >> > > >
