NS5 and PS5 how many PPS ?

[thomaxxx]

Hello! Do you know me say how many PPS (packages per second)
the PS5 and NS5 support?

because the high traffic is directly linked to the PPS.


Thank you.


Thomas.
Brasil - Actual Network Telecom.

:?:

[UBNT-Mike.Ford]

Hello Thomaxxx,

We do not measure our units in terms of PPS. The number of PPS can vary greatly depending on the type of packets being sent and the size of the packets. We measure our devices in sheer total throughput (TCP/IP) as most type of packets are sent using this protocol.

Thank you,

Mike

[Frotihngdog.ca]

Knowing the full throughput of a unit is nice but in my experience the PPS a unit can process is the greatest limiting factor of a wireless radio.

I usually run into PPS issues before I run into throughput issues.

Would the engineers beable to determine the PPS limit?

[physical]

A PPS benchmark isn't meant to be a realistic benchmark. PPS only really means anything if you use same sized packets in your tests across different hardware. Its a synthetic benchmark meant to test a theoretical maximum number of packets a network device can handle.

I think you should do some PPS benchmarks.

Here's my theory on how it would work...
1) Pick an arbitrary packet size (Minimum packet size would give an interesting number)
2) Config source as a computer with a dedicated radio card.
3) Config destination as PS or NS radio bridged to a computer.
4) Source machine sends packets to Destination machine (as much or more than the wireless link can transfer).
5) Destination machine counts the number of packets it recieves.
6) Run for 24 hours.
7) Avg results.

Any concerns or objections to this methodology to test a theoretical maximum PPS for the LS/PS/NS radio?

[UBNT-Mike.Ford]

Hey Guys,

Only problem is I can skew the results by utilizing different packet sizes. This is the reason I do not trust PPS readings. Also I do not know of any reliable software (that is truly accurate) to do PPS testing. If you guys can find some accurate software/hardware that can do PPS testing I will look into it.

Thanks,

Mike

[physical]

Thats why my suggestion was that we use the minimum packet size.
Send packets with zero length data segments across the interface and count those.

What we're trying to test here is the raw capability of the board to take a packet and hand it off, not the amount of data that the board can transfer.

The way I understand it, you'd only be able to skew your numbers lower by increasing the size of packets used with this approach.
It would be interesting to see the numbers with other packet sizes as well.
Might make an interesting graph to see packet size vs PPS capability.

Community Question:
Is this the number you guys are looking for? Or, do you have some other specifications for the test?

[UBNT-Mike.Ford]

Hey Physical,

Sending a packet with zero length data segments seems like a bogus test. That number would have no bearing on real world conditions. Never in a real world would you be sending packets with this. So the PPS number would be very misleading as it would be something never achieved in a real world environment. I dunno, I just always look at things in real world numbers and not something to fluff our marketing brochure.

But like I said if you can find a program or hardware ill test it.

Thanks,

Mike

[UBNT-Mike.Ford]

Hey Guys,

Ive been doing some research to put your minds at ease. I have found a program that should help measure PPS. I will post my results for NS/PS/LS once i figure the program out ;)

Thanks,

Mike

[UBNT-Mike.Ford]

Hey Guys,

The one thing that is different in my testing, is ALL of MikroTik's testing is done on the Ethernet side of things. They are testing Ethernet port to Ethernet port and not over the wireless. Wireless has a theoretical maximum that is far below the specs for a wired network like MikroTik tests with. Can someone point me to a palce that has hard facts about MikroTik's WIRELESS PPS.

Thanks,

Mike

[physical]

wait... when did we start talking about mikrotik stuff?

[UBNT-Mike.Ford]

Hey,

Well so far that the only wireless company I know of that deals in PPS. I went through Tranzeo, MikroTik, Deliberant, ect. and the only one I saw mentioning PPS on a device that can do wireless was MTIK. So i just assumed :)

Thanks,

Mike

[physical]

This is a test of the processor, not the radio...

Now that you mentioned it, feel free to do an ethernet -> ethernet PPS test as well. It'd be interesting to see compared to the wireless one. I'm curious to see how much more the kernel juggles around a wireless packet over an ethernet one.

While you have a test rig up and boards lying around... you might as well do some tests on other boards as well.
I'm not interested in numbers comparing your value segment integrated solution to other companies value segment integrated solution.
I'm interested in seeing where these units fit in the big picture. Where they are appropriate to deploy, and when a stronger unit is required... Considering even our bigger boxes have your cards in em, you're still the winner in the end.

Your choice of brands to compare your board to are pretty poor... None of them provide source code...

[UBNT-Mike.Ford]

Hey Physical,

Only problem with testing ethernet to ethernet is the fact that PowerStation 5, 9, NS2, NS5, PicoStation, LiteStation 5, LSX ect only have a single ethernet interface.

Also, since most of these are single radio boards, and not routers, the processor is not the limiting factor in PPS. Its the radio.

Thanks,

Mike

[physical]

I guess thats why I use the LS2 and PS2
or are my second ethernet ports just for show?

Technically speaking, the radio is part of the processor. And, I always thought that this breed of device as some sort of wireless router. I mean... most home routers are WiSoC style. Was prolly silly to think of these things as anything more than a home router in an industrial enclosure.

The PPS test is to determine the routing capability of the board, and to help identify where they are insufficient for the needs of the network.

I only ever played with the 2.4 boards, so I actually didn't know that the 5g counterparts only had a single ethernet. Just out of curiosity... why only one on the LS5 and PS5? I can understand on the NS and pico.

So, just for fun, add on to the list of requested PPS tests a home linksys wireless AP

[sldnkarm]

Mike,
You could try Vivinet Assessor.
http://www.netiq.com/products/va/default.asp

You can also use iperf to create streams of UDP data at varying packet sizes to get very close to what VA can tell you, minus all the VOIP statistics.

I use iperf for total throughput and PPS measurements on all the equipment we deploy.

sldnkarm

[Frotihngdog.ca]

Mike did you ever get an answer for this post?

[UBNT-Mike.Ford]

Hey Guys,

I still have not found a reliable way to test PPS.

Thanks,

Mike

[sldnkarm]

Mike, did you give iperf a try? I can pass along my testing process and some sample data if you would like. I will try and do so asap.

We test both maximum throughput at both a small packet size (64byte) and large packet size (1470). These two tests will show you max throughput and max pps rates. Which are both things you need to understand when planning a wireless point to point or multi-point system. I have not found a great deal of this data from many manufacturers. Unless your name is Alvarion and you can wirelesly transfering 50,000 pps...

[UBNT-Mike.Ford]

Hello,

Please do. Send the info and instructions to Mike@ubnt.com and ill do some control tests on it.

Thanks,

Mike