You will have the same limitation using LACP.
> On Jan 12, 2018, at 5:00 PM, Steve Jones <thatoneguyst...@gmail.com> wrote: > > that will limit single stream to single port speed, will it not? So I would > end up saturating one link while not using the other if a single stream were > to get heavy? > >> On Fri, Jan 12, 2018 at 3:50 PM, Eric Kuhnke <eric.kuh...@gmail.com> wrote: >> Don't try to do it at L2. Build it as router-to-router OSPF+BGP adjacency >> across the two separate Integra links. >> >> Build it as two OSPF /30 links and use OSPF cost to adjust traffic >> accordingly. >> >> >> >>> On Fri, Jan 12, 2018 at 1:48 PM, Steve Jones <thatoneguyst...@gmail.com> >>> wrote: >>> So we will be doing this integra 2+0 link. We got dinged by sprint though >>> on the PCN. so we have to drop one sides power on one channel since this >>> path has no other channels. This drops that one chain to 256qam (for >>> reliability) from 1024 so 643-514mbps. This model 2+0 the radios dont >>> communicate so its really just 2 separate links handled externally >>> >>> so I go from (643+643) / (643+643) to (514+643) / (643+643) >>> >>> Is there any way with LACP to account for this single path that will be >>> lower than the other two? >>> >>> There is nothing that fully ties me to LACP. I have the option of HP >>> procurve switches or Mikrotik CCR routers to handle the aggregation. >>> >>> As best I can tell LACP doesnt have any granular throughput definition, >>> just splits traffic across all interfaces (last i read, routeros and the hp >>> OS both allow full aggregate speed instead of single streams being limited >>> to individual port throughput) >>> >>> In my case with 1.2gbps i still have an 800mbpsish overflow issue. so If >>> there is an aggregation thats semi dynamic and granular to actual link >>> capacity, that would tickle my goat >>> >>> any advice from the sages? >>> >>> Id like to keep my switch/routers solution to under 1k per side, much less >>> if possible. I already have HP 1810g-24 that i believe will handle this, so >>> theyre effectively free >> >
