Installed and - OH MY GOD THE DUST - sorry. The power jack on the front of the switch is very unusual but it has its advantages - the cables all come in from the front anyway and this allows the switch to be pushed right up to the back of something without crushing the thin power cable like on my EdgeRouter Lite beside it.
From left to right, 1 GbE uplink to my core gigabit switch, my workstation, and to my 2.5 GbE OpenMediaVault server, currently shut off but with its NIC at 10 Mbps standby for WOL.
Send a magic packet to the server, which always works well, waking it up. Amber light remains for a while, then goes out, then comes back green, 2.5 GbE.
Damn I need to dust back there. Anyway.
Workstation connection:
Server connection:
Performance testing. Note I have two connections to my server - the 2.5 GbE connection through this switch for file sharing and a gigabit connection for management.
Ping testing over the switch:
Code:
$ ping 192.168.1.15 -c 10
PING 192.168.1.15 (192.168.1.15) 56(84) bytes of data.
64 bytes from 192.168.1.15: icmp_seq=1 ttl=64 time=0.293 ms
64 bytes from 192.168.1.15: icmp_seq=2 ttl=64 time=0.192 ms
64 bytes from 192.168.1.15: icmp_seq=3 ttl=64 time=0.226 ms
64 bytes from 192.168.1.15: icmp_seq=4 ttl=64 time=0.241 ms
64 bytes from 192.168.1.15: icmp_seq=5 ttl=64 time=0.194 ms
64 bytes from 192.168.1.15: icmp_seq=6 ttl=64 time=0.224 ms
64 bytes from 192.168.1.15: icmp_seq=7 ttl=64 time=0.212 ms
64 bytes from 192.168.1.15: icmp_seq=8 ttl=64 time=0.227 ms
64 bytes from 192.168.1.15: icmp_seq=9 ttl=64 time=0.216 ms
64 bytes from 192.168.1.15: icmp_seq=10 ttl=64 time=0.216 ms
--- 192.168.1.15 ping statistics ---
10 packets transmitted, 10 received, 0% packet loss, time 9222ms
rtt min/avg/max/mdev = 0.192/0.224/0.293/0.026 ms
Versus testing over the gigabit switch:
Code:
$ ping 192.168.1.53 -c 10
PING 192.168.1.53 (192.168.1.53) 56(84) bytes of data.
64 bytes from 192.168.1.53: icmp_seq=1 ttl=64 time=0.213 ms
64 bytes from 192.168.1.53: icmp_seq=2 ttl=64 time=0.217 ms
64 bytes from 192.168.1.53: icmp_seq=3 ttl=64 time=0.224 ms
64 bytes from 192.168.1.53: icmp_seq=4 ttl=64 time=0.221 ms
64 bytes from 192.168.1.53: icmp_seq=5 ttl=64 time=0.224 ms
64 bytes from 192.168.1.53: icmp_seq=6 ttl=64 time=0.217 ms
64 bytes from 192.168.1.53: icmp_seq=7 ttl=64 time=0.194 ms
64 bytes from 192.168.1.53: icmp_seq=8 ttl=64 time=0.235 ms
64 bytes from 192.168.1.53: icmp_seq=9 ttl=64 time=0.226 ms
64 bytes from 192.168.1.53: icmp_seq=10 ttl=64 time=0.219 ms
--- 192.168.1.53 ping statistics ---
10 packets transmitted, 10 received, 0% packet loss, time 9213ms
rtt min/avg/max/mdev = 0.194/0.219/0.235/0.010 ms
Nothing to note here. It's not impeding anything and behaves the same as my gigabit switch. So similar that it may be a clue - more on this later.
iperf throughput tests:
Code:
$ iperf -c 192.168.1.15
------------------------------------------------------------
Client connecting to 192.168.1.15, TCP port 5001
TCP window size: 935 KByte (default)
------------------------------------------------------------
[ 3] local 192.168.1.77 port 35518 connected with 192.168.1.15 port 5001
[ ID] Interval Transfer Bandwidth
[ 3] 0.0-10.0 sec 2.74 GBytes 2.36 Gbits/sec
Versus iperf tests over my gigabit switch:
Code:
$ iperf -c 192.168.1.53
------------------------------------------------------------
Client connecting to 192.168.1.53, TCP port 5001
TCP window size: 944 KByte (default)
------------------------------------------------------------
[ 3] local 192.168.1.77 port 48332 connected with 192.168.1.53 port 5001
[ ID] Interval Transfer Bandwidth
[ 3] 0.0-10.0 sec 2.74 GBytes 2.36 Gbits/sec
Oops, that's weird, that went over the 2.5 GbE connection even though that IP is assigned to the gigabit port. More on this later, but at least I confirmed the switch is operating at wire speed. 2.36 Gbits/sec agrees with a direct connection I tried before I had this switch.
And finally, file sharing tests. I'll be using NFS because it's the fastest for my use (I run Linux and tested this - NFS tested the fastest. It's more efficient and multi-threaded). First clearing caches then "pushing" the file from the workstation to the server, twice because I screwed up the name:
Code:
# echo 3 >/proc/sys/vm/drop_caches
# pv Windows\ 10\ Educational.iso.xz > /home/Shadowfax/Backup
3.71GiB 0:00:13 [ 288MiB/s] [================================>] 100%
# pv Windows\ 10\ Educational.iso.xz > /home/Shadowfax/test
3.71GiB 0:00:13 [ 271MiB/s] [================================>] 100%
Even faster than my previous tests, 288MiB/s = 2.42Gb/s = 302MB/s. Agrees with iperf plus a little bonus.
Now a "pull" from the server to the workstation.
Code:
# pv /home/Shadowfax/test > test
3.71GiB 0:00:34 [ 110MiB/s] [================================>] 100%
# echo 3 >/proc/sys/vm/drop_caches
# pv /home/Shadowfax/test > test
3.71GiB 0:00:34 [ 111MiB/s] [================================>] 100%
'ere now, wot's dis? 111MiB/s = 0.931 Gb/s = 116MB/s.
Looks like that one went over gigabit despite me specifically mounting the IP address of the 2.5 GbE interface of the server. That would explain the weirdness earlier. Some sort of routing issue or something. Easily solvable, just disconnect the gigabit connection to the server.
Code:
# echo 3 >/proc/sys/vm/drop_caches
# pv /home/Shadowfax/test > test
3.71GiB 0:00:13 [ 277MiB/s] [================================>] 100%
Proper!
I should mention throughout these tests the switch didn't get hot. It's just barely above ambient directly above the heatsink. An IR thermometer shows 31.9°C maximum on the top of the switch, in a 23.4°C room. So the heatsink is more than adequate, this switch does not need fan cooling. Its very open venting and heatsink should be more than enough even with all ports active.
So that's it. This switch does exactly what it says on the box, no more, no less. It seems well built and well cooled and works perfectly both at gigabit and at 2.5 GbE for a price per 2.5 GbE port no other switch can match.
This is not a "forever" switch - much like I outgrew 4 10/100 ports off my router to 4 gigabit ports off my router to an 8-port gigabit switch, and much like I outgrew that to an 8-port managed gigabit switch, and much like I outgrew that to a 16-port managed gigabit switch, it would be better to have an 8- or 16-port managed 2.5 GbE switch. But for now it works perfectly and allows room for expansion - I can add a connection to my FreeNAS server when I can get a 2.5 GbE (or low-priced 10 GbE) card with auto-negotiation. I can also add a WiFi 6 / AX access point with a 2.5 GbE port, though I'd have to use a PoE injector.
It was worth the wait and the price I paid for it.