Skip to Content.
Sympa Menu

perfsonar-user - Re: [perfsonar-user] Low Cost Hardware-Intel Nuc

Subject: perfSONAR User Q&A and Other Discussion

List archive

Re: [perfsonar-user] Low Cost Hardware-Intel Nuc


Chronological Thread 
  • From: Hervey Allen <>
  • To:
  • Subject: Re: [perfsonar-user] Low Cost Hardware-Intel Nuc
  • Date: Tue, 9 Feb 2016 22:55:27 -0800
  • Organization: Network Startup Resource Center

On 2/9/16 8:06 PM, John W. O'Brien wrote:
> On 2/9/16 5:59 PM, Manglos, Andrew P (173E) wrote:
>> I was wondering if anyone has used an Intel Nuc for iperf testing? I
>> noticed that it is mentioned on the low cost hardware page but did not
>> see any reference of testing on these devices.
>
> I have not, but I am the proud owner of multiple NUCs, and would be glad
> to give it a whirl. Are there test that are of specific interest to you?
>

We have used both Intel NUC5i5RYH

http://www.intel.com/content/www/us/en/nuc/nuc-kit-nuc5i5ryh.html

and the Intel NUC5i7RYH

http://www.intel.com/content/www/us/en/nuc/nuc-kit-nuc5i7ryh.html

We use Samsung 850 Pro SSDs and 16GB of DDR3L 1600MHz PC3 12800 RAM.

With the 5th gen, i5-5250U processor we did local iPerf tests with the
on-board NIC and against another box that had a 10G NIC.

owping statistics between the two boxes looked like this:

--- owping statistics from [ps-test-02.nsrc.org]:8763 to
[perfsonar.nsrc.org]:9800 ---
SID: 80df9d2cd99dd0e22e633e7ecaa109b3
first: 2015-09-11T22:14:59.294
last: 2015-09-11T22:15:09.761
100 sent, 0 lost (0.000%), 0 duplicates
one-way delay min/median/max = -20.3/-20/-19.4 ms, (err=16.8 ms)
one-way jitter = 0.5 ms (P95-P50)
Hops = 1 (consistently)
no reordering

and

--- owping statistics from [perfsonar.nsrc.org]:9839 to
[ps-test-02.nsrc.org]:8796 ---
SID: 80df5583d99dd0e23390f966b900cc27
first: 2015-09-11T22:14:59.293
last: 2015-09-11T22:15:10.884
100 sent, 0 lost (0.000%), 0 duplicates
one-way delay min/median/max = 19.8/20.3/20.7 ms, (err=16.8 ms)
one-way jitter = 0.4 ms (P95-P50)
Hops = 1 (consistently)
no reordering

These were pretty typical test results fir bwctl (using iperf):

bwctl -c perfsonar.nsrc.org -t 30 -f m -i 1
bwctl: Using tool: iperf
bwctl: 36 seconds until test results available

RECEIVER START
------------------------------------------------------------
Server listening on TCP port 5061
Binding to local address 128.223.157.44
TCP window size: 0.08 MByte (default)
------------------------------------------------------------
[ 15] local 128.223.157.44 port 5061 connected with 128.223.85.131 port
40890
[ ID] Interval Transfer Bandwidth
[ 15] 0.0- 1.0 sec 109 MBytes 915 Mbits/sec
[ 15] 1.0- 2.0 sec 109 MBytes 916 Mbits/sec
[ 15] 2.0- 3.0 sec 109 MBytes 915 Mbits/sec
[ 15] 3.0- 4.0 sec 109 MBytes 916 Mbits/sec
[ 15] 4.0- 5.0 sec 109 MBytes 913 Mbits/sec
[ 15] 5.0- 6.0 sec 109 MBytes 915 Mbits/sec
[ 15] 6.0- 7.0 sec 109 MBytes 914 Mbits/sec
[ 15] 7.0- 8.0 sec 109 MBytes 916 Mbits/sec
[ 15] 8.0- 9.0 sec 109 MBytes 913 Mbits/sec
[ 15] 9.0-10.0 sec 109 MBytes 915 Mbits/sec
[ 15] 10.0-11.0 sec 109 MBytes 915 Mbits/sec
[ 15] 11.0-12.0 sec 109 MBytes 913 Mbits/sec
[ 15] 12.0-13.0 sec 109 MBytes 914 Mbits/sec
[ 15] 13.0-14.0 sec 109 MBytes 914 Mbits/sec
[ 15] 14.0-15.0 sec 109 MBytes 916 Mbits/sec
[ 15] 15.0-16.0 sec 109 MBytes 916 Mbits/sec
[ 15] 16.0-17.0 sec 109 MBytes 913 Mbits/sec
[ 15] 17.0-18.0 sec 109 MBytes 916 Mbits/sec
[ 15] 18.0-19.0 sec 106 MBytes 893 Mbits/sec
[ 15] 19.0-20.0 sec 109 MBytes 915 Mbits/sec
[ 15] 20.0-21.0 sec 109 MBytes 915 Mbits/sec
[ 15] 21.0-22.0 sec 109 MBytes 917 Mbits/sec
[ 15] 22.0-23.0 sec 109 MBytes 916 Mbits/sec
[ 15] 23.0-24.0 sec 109 MBytes 917 Mbits/sec
[ 15] 24.0-25.0 sec 109 MBytes 917 Mbits/sec
[ 15] 25.0-26.0 sec 109 MBytes 918 Mbits/sec
[ 15] 26.0-27.0 sec 109 MBytes 916 Mbits/sec
[ 15] 27.0-28.0 sec 109 MBytes 917 Mbits/sec
[ 15] 28.0-29.0 sec 109 MBytes 916 Mbits/sec
[ 15] 29.0-30.0 sec 109 MBytes 917 Mbits/sec
[ 15] 0.0-30.0 sec 3272 MBytes 915 Mbits/sec
[ 15] MSS size 1448 bytes (MTU 1500 bytes, ethernet)


We saw slightly better results when we connected 2 x NUCs back-to-back,
but I don't have those tests.

Bottom line - the 5th gen, i5, dual core model appears to sustain
1Gbits/sec throughput levels using iperf.

I have not tested the less expensive 5th gen, i3 models. We have not
tested the i7 NUC we have yet. I'm assuming it would be as good, or
better, but that just an assumption.

Cheers,
- Hervey Allen





--
Hervey Allen Assistant Director, Network Startup Resource Center

http://nsrc.org/ : http://facebook.com/nsrc.org
GPG Fingerprint: AC08 31CB E453 6C65 2AB3 4EDB CEEB 5A74 C6E5 624F



Archive powered by MHonArc 2.6.16.

Top of Page