OpenVPN Slow - local network test

spyder0552

@spyder0552
Back on this again now. Anyone have suggestions?
This is all on the same network using Cat6 cables. Just testing from a laptop to the wall to a device on the other side.
No VPN - Speed is fantastic.
OpenVPN tunnel - speed is pathetic.

Any help would be appreciated before I have to dump this and look to another solution. I just can't believe that OpenVPN can't handle at least 400-600 Mbs.

Thanks all.

Rico

Did you try with Encryption Algorithm set to None for testing?
If you really aim for highspeed VPN traffic I think you need to try with IPsec.

-Rico

johnpoz

I'm about ready to head out - but I I want to try and see what my 4860 can do via this local sort of testing.. If find some time will do a local only test..

spyder0552

@johnpoz
Hi all.
So, still having the same issue. I have now setup a home lab and get the exact same results.
Test Lab:
Intel i5 with 32Gb Ram.
Using Windows HyperV as the host.
Pfsense and Windows10 as guest machines.

Test 1 = laptop -> Firewall -> Windows 10
Get ~900 Mb/s

Test 2 = Laptop with OpenVP -> Firewall -> Windows 10
Get ~ 170 Mb/s

This is the same results I am seeing at my office using Dell R610 with 24 CPU Xenon's.

Things I have tried:

Enable/disable the network options Hardware TCP Seg and Large Receive Offload (check box set to disable them).
Enable fast I/O (with it turned off, I go down to ~155-160)...so currently enabled.
Snd/rcv buffers. Tried default and 2Mb. No significant change.
Encryption at AES-128-CBC and GCM. No significant change.

So for anyone that wants to setup my test enviroment:

Just setup a HyperV box.
Add 2 virtual switches (one external and one internal).
Create a VM for Pfsense and give it both virtual switches (WAN = External, LAN = Internal switch).
Create a VM for a test box
Put Iperf on the test box
Create your PFsense with default OpenVPN (using TLS). Just run through the wizard.

I am sure if someone did this they would get to the same spot I am where the speed is just no where at what it should be capable of.
You can rule the hardware out to some extent as testing without the VPN shows near full speed (meaning the nics are talking to each other properly).

Anyone willing to try my test?

spyder0552

@Rico
Hey, just replied below. Yes, I did try that on Friday and it did not make a difference.

Is anyone out there trying this with Windows 10 clients? Perhaps this is a windows issue?

Pippin

@spyder0552 said in OpenVPN Slow - local network test:

[2.4.5-RELEASE][admin@pfSense.seradex.local]/root: openvpn --genkey --secret /tm p/secret
[2.4.5-RELEASE][admin@pfSense.seradex.local]/root: time openvpn --test-crypto -- secret /tmp/secret --verb 0 --tun-mtu 20000 --cipher aes-128-gcm
Fri Jun 19 14:47:16 2020 disabling NCP mode (--ncp-disable) because not in P2MP client or server mode
10.007u 0.000s 0:10.00 100.0% 843+177k 0+0io 0pf+0w
[2.4.5-RELEASE][admin@pfSense.seradex.local]/root:

By the looks of that ^^^ it theoretically could do ~320Mbps

How about testing both ways? (client>server - server>client)
Which I5 is that?

Your top -aSH screenshot is timed wrong, ..... I guess.

spyder0552

@Pippin
Home test lab is i5-4690K (yeah..little dated...best I can get atm).

I can't really get it to test from the inside out. The wall seems to be blocking it out. I guess the IP I would connect to would be the end point of the tunel...but not working it seems.

Pippin

You can reverse test mode by using iperf3 -R
"-R, Reverse test mode – Server sends, client receives"

spyder0552

@Pippin
WOW...actually slower in reverse. Averaged only 100 Mb/s

But if I drop the VPN and just go through the firewall, it goes up to 900

spyder0552

@spyder0552
Just to show the config on the client/server side
Client:
dev tun
persist-tun
persist-key
cipher AES-128-GCM
ncp-ciphers AES-128-GCM
auth SHA256
tls-client
client
resolv-retry infinite
remote 192.168.2.98 1194 udp4
verify-x509-name "OpenVPN-Server-Cert" name
auth-user-pass
pkcs12 pfSense-UDP4-1194-vpn.p12
tls-crypt pfSense-UDP4-1194-vpn-tls.key
remote-cert-tls server

Server:
dev ovpns1
verb 1
dev-type tun
dev-node /dev/tun1
writepid /var/run/openvpn_server1.pid
#user nobody
#group nobody
script-security 3
daemon
keepalive 10 60
ping-timer-rem
persist-tun
persist-key
proto udp4
cipher AES-128-GCM
auth SHA256
up /usr/local/sbin/ovpn-linkup
down /usr/local/sbin/ovpn-linkdown
client-connect /usr/local/sbin/openvpn.attributes.sh
client-disconnect /usr/local/sbin/openvpn.attributes.sh
local 192.168.2.98
tls-server
server 192.168.99.0 255.255.255.0
client-config-dir /var/etc/openvpn-csc/server1
username-as-common-name
plugin /usr/local/lib/openvpn/plugins/openvpn-plugin-auth-script.so /usr/local/sbin/ovpn_auth_verify_async user TG9jYWwgRGF0YWJhc2U= false server1 1194
tls-verify "/usr/local/sbin/ovpn_auth_verify tls 'OpenVPN-Server-Cert' 1"
lport 1194
management /var/etc/openvpn/server1.sock unix
push "route 192.168.1.0 255.255.255.0"
push "dhcp-option DOMAIN test.local"
push "dhcp-option DNS 8.8.8.8"
client-to-client
ca /var/etc/openvpn/server1.ca
cert /var/etc/openvpn/server1.cert
key /var/etc/openvpn/server1.key
dh /etc/dh-parameters.2048
tls-crypt /var/etc/openvpn/server1.tls-crypt
ncp-ciphers AES-128-GCM
persist-remote-ip
float
topology subnet
fast-io

johnpoz

@spyder0552 said in OpenVPN Slow - local network test:

Just setup a HyperV box.

That doesn't seem like a legit test.. My test will be this

windows 10 PC --- sg4860--- NAS

Will set this up later.. But you don't show any of the mtu settings or txqueuelen..

johnpoz

Ok I just set this up...

With default everything on open seeing..

$ iperf3 -c 192.168.9.10
warning: Ignoring nonsense TCP MSS 334848
Connecting to host 192.168.9.10, port 5201
[  5] local 10.0.100.2 port 52251 connected to 192.168.9.10 port 5201
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-1.00   sec  25.5 MBytes   214 Mbits/sec
[  5]   1.00-2.00   sec  29.5 MBytes   248 Mbits/sec
[  5]   2.00-3.00   sec  32.4 MBytes   272 Mbits/sec
[  5]   3.00-4.00   sec  31.1 MBytes   261 Mbits/sec
[  5]   4.00-5.00   sec  29.5 MBytes   247 Mbits/sec
[  5]   5.00-6.00   sec  30.8 MBytes   258 Mbits/sec
[  5]   6.00-7.00   sec  30.6 MBytes   257 Mbits/sec
[  5]   7.00-8.00   sec  31.6 MBytes   265 Mbits/sec
[  5]   8.00-9.00   sec  31.0 MBytes   260 Mbits/sec
[  5]   9.00-10.00  sec  32.0 MBytes   268 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-10.00  sec   304 MBytes   255 Mbits/sec                  sender
[  5]   0.00-10.01  sec   304 MBytes   255 Mbits/sec                  receiver

Let me see if can tweak that a bit..

win10 pc (192.168.200.10) --- switch - 192.168.200.1 (test igb4) pfsense (sg4860) ( igb0 lan) 192.168.9.253 -- switch -- 192.168.9.10 (NAS)

Tweaking didn't make much difference to be honest, but using aes-128-gcm vs cbc was huge difference.. When changed to cbc vs gcm when to 130 vs mid 200's

spyder0552

@johnpoz
Thanks for doing the testing.
I am surprised to see that the best we can get on a full 1Gb/s link is ~250 Mb/s
So I guess my ~150Mb/s on the first crack is not horrible?

I know there is overhead to deal with...just surprised it is so much. This is looking like a 70% loss of speed using OpenVpn.
Is this what others are seeing as well?

johnpoz

openvpn is single threaded.. its easy to use - its never been "speedy" ;)

Keep in mind my sg4860, not a rocketship vpn endpoint concentrater either..

It has enough umph to get the job done with lower power requirements.. But prob not what I would use for my vpn endpoint if what I wanted to as much throughput as possible.. Nor would openvpn be my first choice in that area - ipsec is better geared for throughput..

Openvpn advantage is ease of use, and deployment, etc.

But overall your sort of test with everything on the same vm host is not really a valid sort of testing.. It works for poc, etc. But its not going to be a good indicator of what sort of bandwidth you could expect when using in the real world..

Pippin

But overall your sort of test with everything on the same vm host is not really a valid sort of testing

Yes I think so too.
Looking at the hardware it should be capable of more.
.

Nor would openvpn be my first choice in that area

Wait a bit, who knows ;)

johnpoz

@Pippin said in OpenVPN Slow - local network test:

Wait a bit, who knows ;)

If the goal was pure throughput, openvpn would not be on the top of the list of choices.. It has many other attributes it shines at.. But if what I am looking for is closest to line speed using least amount of horsepower.. Then no its not on the top of really any list ;)

spyder0552

@johnpoz
That is what I am concluding.
OpenVPN = easy to deploy and flexible....just not that speedy.

I am in the process of setting up another test to use IPSEC and compare. I will post my results for those that are interested.

For this initial issue, I am calling it closed as it seems we have beaten the heck out of it. Thank you all for the fantastic support on this. So very much appreciated.

Pippin

If the goal was pure throughput, openvpn would not be on the top of the list of choices

Sure, but maybe it will get to that top some time in the future.

johnpoz

^ we can hope yeah ;)

Pippin

@johnpoz said in OpenVPN Slow - local network test:

^ we can hope yeah ;)

Although for Linux but here it is:
https://github.com/OpenVPN/ovpn-dco