Hello,
I'm interested, if you have met with the following problem:
when using NV2 and mixed type of ethernet (100mbps and Gig) the thruput is hardly decreased.
In case of 'clean 100mbps' or 'clean gigE' network, the thruput is slightly better.
This NOT happens in case of nstreme.
What i'm talking about is:
1)
Btest source (PC) - (gigE) - Switch -(100M)- RB433AH -(wireless,NV2 2x2 40mhz)- RB433AH - Btest end (RB493AH as switch and multiple RB433AHs)
~40mbps(tcp)
2)
Btest source (RB433AH) - (100M) - Switch -(100M)- RB433AH -(wireless,nsteme turbo, single polarity 40mhz)- RB433AH - Btest end (RB493AH as switch and multiple RB433AHs)
~65mbps (tcp - or little bit more, if i run btest on multiple RBs on each side)
3)
Btest source (PC) - (gigE) - Switch -(gigE)- RB493AH -(wireless 2x2 40mhz)- RB433AH - Btest end (RB493AH as switch and multiple RB433AHs)
up to 100mbps TCP (433AH cannot give more, since it's 100mbps only)
What is strange: I cannot reproduce it on desk, but happens on many working links.
ps: replacing the first hop results much better thruput on second and third hop in multihop environment.
Re: Nstreme vs NV2 effective thruput with mixed type of ethe
[img]
http://forum.mikrotik.com/download/file ... w&id=12468
[/img]
As seen in the structure above, the measurements happens with btest to/from point "A" on top left. It's connected to a full GigE switch,
what's connected to a 24x100 and 2xGE switch with GE uplink. Measurement done with TCP.
the 433AHs connected to FE port, while 493G connects to a GigE port (ok, thru a RB2011's GigE port - but everything is full gig in this line).
What is found is:
- From "A" to SiteB router (what's an atom PC) can happily do HDX 150mbps (90+60mbps since the remote radio is in a 433AH as visible)
before the 493G, the local radio was in 433AH as well. Speed was ~40M down and 80M up (but with bad direction)
- From "A" to SiteA AP1 and AP2 i can measure ~ HDX 80mbps (split by the two receiver, since ~60mbps eat's up a 433AH's CPU)
as i know, ~ 80mbps is the limit of single polarity nstreme link
- From "A" to SiteC AP1 and AP2 i can push ~60mbps down and ~80-90mbps up (also in bad direction)
(hopefully today the 433AH feeding SiteC will be replaced to RB600 with GE uplink - fingers crossed)
I also can measure higher speed from a 433AH connected to a switch than from "A" PC.
My conclusion is:
using NV2, if the traffic is originated from GigE port and destinated to FastE port, speed is much less than expected.
If the datapath is sent from FastE to FastE - it's seems OK.
If the wireless protocol is nstreme (instead of nv2) speed it much better (bottleneck is wireless at 80mbps).
That makes me think of NV2 issue with TCP ACK or ethernet backpressure. Maybe issue within bridging.
(setup is not yet tested with routing since it's a production system)
http://forum.mikrotik.com/download/file ... w&id=12468
[/img]
As seen in the structure above, the measurements happens with btest to/from point "A" on top left. It's connected to a full GigE switch,
what's connected to a 24x100 and 2xGE switch with GE uplink. Measurement done with TCP.
the 433AHs connected to FE port, while 493G connects to a GigE port (ok, thru a RB2011's GigE port - but everything is full gig in this line).
What is found is:
- From "A" to SiteB router (what's an atom PC) can happily do HDX 150mbps (90+60mbps since the remote radio is in a 433AH as visible)
before the 493G, the local radio was in 433AH as well. Speed was ~40M down and 80M up (but with bad direction)
- From "A" to SiteA AP1 and AP2 i can measure ~ HDX 80mbps (split by the two receiver, since ~60mbps eat's up a 433AH's CPU)
as i know, ~ 80mbps is the limit of single polarity nstreme link
- From "A" to SiteC AP1 and AP2 i can push ~60mbps down and ~80-90mbps up (also in bad direction)
(hopefully today the 433AH feeding SiteC will be replaced to RB600 with GE uplink - fingers crossed)
I also can measure higher speed from a 433AH connected to a switch than from "A" PC.
My conclusion is:
using NV2, if the traffic is originated from GigE port and destinated to FastE port, speed is much less than expected.
If the datapath is sent from FastE to FastE - it's seems OK.
If the wireless protocol is nstreme (instead of nv2) speed it much better (bottleneck is wireless at 80mbps).
That makes me think of NV2 issue with TCP ACK or ethernet backpressure. Maybe issue within bridging.
(setup is not yet tested with routing since it's a production system)
You do not have the required permissions to view the files attached to this post.
Re: Nstreme vs NV2 effective thruput with mixed type of ethe
Did you found solution of this problem, I have exactly same problem on RB450G<->METAL<->Groove and RB450G is wired by 1gb port. Version 6.7. Is there hack to avoid this?
My conclusion is:
using NV2, if the traffic is originated from GigE port and destinated to FastE port, speed is much less than expected.
If the datapath is sent from FastE to FastE - it's seems OK.
If the wireless protocol is nstreme (instead of nv2) speed it much better (bottleneck is wireless at 80mbps).
That makes me think of NV2 issue with TCP ACK or ethernet backpressure. Maybe issue within bridging.
(setup is not yet tested with routing since it's a production system)
Re: Nstreme vs NV2 effective thruput with mixed type of ethe
well, you can
1) replace 450G with 2011 (or 450) and use only 100mbps port
2) use nstream instead of nv2
3) replace groove with 952 (or alike with gigE port)
1) replace 450G with 2011 (or 450) and use only 100mbps port
2) use nstream instead of nv2
3) replace groove with 952 (or alike with gigE port)
Who is online
Users browsing this forum: No registered users and 55 guests