Gigabit speed and gear thread

Forums › Spark New Zealand (including Skinny and BigPipe) › Gigabit speed and gear thread

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | ... | 11

370 posts

Ultimate Geek

#1647634 8-Oct-2016 10:21

dt:

Orcon, Auckland, North Shore, Albany.

Running a pfSense router made up of an i3, 8GB memory and a quad port Intel Pro/1000 PT server NIC.

CAT6 from the ONT to router, CAT6 from the router to my PC running a PCI-E Intel 1340-t2 server NIC.

I was getting between 200-500 up and down on the Orcon supplied fritzbox..

Have you thought about creating an 2Gbps link via NIC bonding between your pfsense box and your PC?

Just wondered if those speeds would increase a little.

Would be interested to hear if anyone is running gigabit fibre through a stock FreeBSD or OpenBSD router.

lagbort

263 posts

Ultimate Geek

ID Verified

#1647638 8-Oct-2016 10:27

michaelmurfy:

@lagbort before you buy a new router could you please try the following:

1) Download Advanced Tomato and flash it on your router.
2) Set it up as normal then go into Advanced -> Miscellaneous to enable Hardware NAT (can't remember what it is actually called), reboot.
3) Test again over Ethernet - see if you see any improvement.

This router has the capability to do Gigabit with Hardware NAT enabled so will be interested to see if you see any changes.

After installing Advanced Tomato

Reading around it appears the standard firmware cant do CTF over PPPOE.

All is well now, thanks for the help!

Intravix

110 posts

Master Geek

#1647716 8-Oct-2016 16:16

dt:

Orcon, Auckland, North Shore, Albany.

Running a pfSense router made up of an i3, 8GB memory and a quad port Intel Pro/1000 PT server NIC.

CAT6 from the ONT to router, CAT6 from the router to my PC running a PCI-E Intel 1340-t2 server NIC.

I was getting between 200-500 up and down on the Orcon supplied fritzbox..

Awesome, is pfsense virtualized? I'm running it (on 100/20Mbps Bigpipe) in ESXi with 1 core and 512MB, I think it runs around 30% memory utilization and similar CPU but sometimes goes up to 100% (with no noticeable changes in performance). Will look at increasing the resources and possibly replacing/upgrading the host soon especially when I go to gigabit.

Xeon

302 posts

Ultimate Geek

#1647728 8-Oct-2016 16:56

I feel like I'm doing something wrong.

BigPipe/Dunedin.

CAT6 via Asus RT-AC87U w/ Merlin (hardware acceleration CTF only):

Direct to ONT:

cynnicallemon

370 posts

Ultimate Geek

#1647735 8-Oct-2016 17:06

Intravix:

Awesome, is pfsense virtualized? I'm running it (on 100/20Mbps Bigpipe) in ESXi with 1 core and 512MB, I think it runs around 30% memory utilization and similar CPU but sometimes goes up to 100% (with no noticeable changes in performance). Will look at increasing the resources and possibly replacing/upgrading the host soon especially when I go to gigabit.

I don't think dt was running that virtualized and the golden rule is that you really shouldn't run anything like pfSense as a VM. If your version of pfSense is 2.2 (or above) then you should be running with dual (or more cores) as pf itself is now SMP compatible. On a 100Mbps connection it should really be dual core and say 2GB RAM. For 1Gbps then ideally dual/quad core and 2-4GB RAM.

But your virtualized pfSense VM needs a beef up.

Intravix

110 posts

Master Geek

#1647742 8-Oct-2016 17:14

cynnicallemon:

Intravix:

Awesome, is pfsense virtualized? I'm running it (on 100/20Mbps Bigpipe) in ESXi with 1 core and 512MB, I think it runs around 30% memory utilization and similar CPU but sometimes goes up to 100% (with no noticeable changes in performance). Will look at increasing the resources and possibly replacing/upgrading the host soon especially when I go to gigabit.

I don't think dt was running that virtualized and the golden rule is that you really shouldn't run anything like pfSense as a VM. If your version of pfSense is 2.2 (or above) then you should be running with dual (or more cores) as pf itself is now SMP compatible. On a 100Mbps connection it should really be dual core and say 2GB RAM. For 1Gbps then ideally dual/quad core and 2-4GB RAM.

But your virtualized pfSense VM needs a beef up.

I've seen a lot of recommendations along those lines, and some the other way too stating VM isolation is good enough. I followed a guide which suggested specs, but I read that the new version is written for multiple cores, I'll probably migrate it soon to the host it's on and use another PC as a web/other server.

zyo

513 posts

Ultimate Geek

#1647846 9-Oct-2016 06:51

michaelmurfy:
@lagbort before you buy a new router could you please try the following:

1) Download Advanced Tomato and flash it on your router.
2) Set it up as normal then go into Advanced -> Miscellaneous to enable Hardware NAT (can't remember what it is actually called), reboot.
3) Test again over Ethernet - see if you see any improvement.

This router has the capability to do Gigabit with Hardware NAT enabled so will be interested to see if you see any changes.

Speaking of advanced tomato on r7000
Do you know if it supports 11ac ? In the wireless mode drop down for the 5ghz band there is only a and n.

There is an option for 80mhz channel width.

Sharesies

Trade NZ and US shares and funds with Sharesies (affiliate link).

michaelmurfy

meow
13242 posts

Uber Geek

Moderator

ID Verified

Trusted

Lifetime subscriber

#1647855 9-Oct-2016 08:01

zyo:
michaelmurfy:
@lagbort before you buy a new router could you please try the following:

1) Download Advanced Tomato and flash it on your router.
2) Set it up as normal then go into Advanced -> Miscellaneous to enable Hardware NAT (can't remember what it is actually called), reboot.
3) Test again over Ethernet - see if you see any improvement.

This router has the capability to do Gigabit with Hardware NAT enabled so will be interested to see if you see any changes.

Speaking of advanced tomato on r7000
Do you know if it supports 11ac ? In the wireless mode drop down for the 5ghz band there is only a and n.

There is an option for 80mhz channel width.

In my experience it does (@mdf can comment further as he bought my router).

I have not used one since last year. Had the Edgerouter since.

Michael Murphy | https://murfy.nz
Referral Links: Quic Broadband (use R122101E7CV7Q for free setup)

Are you happy with what you get from Geekzone? Please consider supporting us by subscribing.
Opinions are my own and not the views of my employer.

dan

1134 posts

Uber Geek

Lifetime subscriber

#1647861 9-Oct-2016 08:38

Xeon:

I feel like I'm doing something wrong.

BigPipe/Dunedin.

CAT6 via Asus RT-AC87U w/ Merlin (hardware acceleration CTF only):

Direct to ONT:

you may be limited by the capabilities of the PC/laptop/whatever actually doing the speed test?

cynnicallemon

370 posts

Ultimate Geek

#1647863 9-Oct-2016 08:48

Xeon:

I feel like I'm doing something wrong.

BigPipe/Dunedin.

CAT6 via Asus RT-AC87U w/ Merlin (hardware acceleration CTF only):

If you're in Dunedin then it would be better to select a server in Dunedin to do your speedtest from as it may yield better results but those test to Auckland should be way better.

Talkiet

4792 posts

Uber Geek

Trusted

#1647868 9-Oct-2016 08:49

cynnicallemon:

Xeon:

I feel like I'm doing something wrong.

BigPipe/Dunedin.

CAT6 via Asus RT-AC87U w/ Merlin (hardware acceleration CTF only):

If you're in Dunedin then it would be better to select a server in Dunedin to do your speedtest from as it may yield better results but those test to Auckland should be way better.

This shouldn't make a significant difference with speedtest.net servers. I have personally reviewed the reporting from Spark's 3 speedtest.net servers and I can see 900+Mbps results from DUnedin against our Christchurch, Wellington and Auckland servers.

Cheers- N

Please note all comments are from my own brain and don't necessarily represent the position or opinions of my employer, previous employers, colleagues, friends or pets.

cynnicallemon

370 posts

Ultimate Geek

#1647904 9-Oct-2016 10:25

Talkiet:

cynnicallemon:

Xeon:

I feel like I'm doing something wrong.

BigPipe/Dunedin.

CAT6 via Asus RT-AC87U w/ Merlin (hardware acceleration CTF only):

If you're in Dunedin then it would be better to select a server in Dunedin to do your speedtest from as it may yield better results but those test to Auckland should be way better.

This shouldn't make a significant difference with speedtest.net servers. I have personally reviewed the reporting from Spark's 3 speedtest.net servers and I can see 900+Mbps results from DUnedin against our Christchurch, Wellington and Auckland servers.

Cheers- N

Agreed, it shouldn't make a great difference but if you're troubleshooting speed issues I would think it's best to pick the Dunedin servers in this case as you rule out potential networking problems over that 1400km distance that might reflect a non optimal test.

sbiddle

30853 posts

Uber Geek

Retired Mod

Trusted

Biddle Corp

Lifetime subscriber

#1647908 9-Oct-2016 10:58

cynnicallemon:

Talkiet:

cynnicallemon:

Xeon:

I feel like I'm doing something wrong.

BigPipe/Dunedin.

CAT6 via Asus RT-AC87U w/ Merlin (hardware acceleration CTF only):

If you're in Dunedin then it would be better to select a server in Dunedin to do your speedtest from as it may yield better results but those test to Auckland should be way better.

This shouldn't make a significant difference with speedtest.net servers. I have personally reviewed the reporting from Spark's 3 speedtest.net servers and I can see 900+Mbps results from DUnedin against our Christchurch, Wellington and Auckland servers.

Cheers- N

Agreed, it shouldn't make a great difference but if you're troubleshooting speed issues I would think it's best to pick the Dunedin servers in this case as you rule out potential networking problems over that 1400km distance that might reflect a non optimal test.

But that depends if the provider is peering in Dunedin. We know that Big Pipe's only BNG is Auckland so if you're testing to a Dunedin server on a Big Pipe connection the traffic will go to Auckland and then back.

It's also very safe to assume there are a number of NZ speedtest servers only running on 1GB connections.

dt

1152 posts

Uber Geek
Inactive user

#1648792 10-Oct-2016 22:30

Intravix:

Awesome, is pfsense virtualized? I'm running it (on 100/20Mbps Bigpipe) in ESXi with 1 core and 512MB, I think it runs around 30% memory utilization and similar CPU but sometimes goes up to 100% (with no noticeable changes in performance). Will look at increasing the resources and possibly replacing/upgrading the host soon especially when I go to gigabit.

I'm running a physical box. This setup appears to be way overkill however. CPU very rarely goes up to 15% and memory sits around 5-10%. Pretty much running a full UTM suite as well.

It was a re purposed HTPC I had laying around. currently looking to build something less beefy with lower power requirements.

I did see a blog where someone put in the effort running multiple tests in virtual/physcial configurations and there was next to no difference in the results

cynnicallemon:

Have you thought about creating an 2Gbps link via NIC bonding between your pfsense box and your PC?

Just wondered if those speeds would increase a little.

Would be interested to hear if anyone is running gigabit fibre through a stock FreeBSD or OpenBSD router.

Will give it a crack tomorrow evening and post up the results

My connection seems to be getting faster by the day at the moment

mdf

3513 posts

Uber Geek

Trusted

#1648794 10-Oct-2016 22:39

michaelmurfy:
zyo:
michaelmurfy:

@lagbort before you buy a new router could you please try the following:

1) Download Advanced Tomato and flash it on your router.
2) Set it up as normal then go into Advanced -> Miscellaneous to enable Hardware NAT (can't remember what it is actually called), reboot.
3) Test again over Ethernet - see if you see any improvement.

This router has the capability to do Gigabit with Hardware NAT enabled so will be interested to see if you see any changes.

Speaking of advanced tomato on r7000
Do you know if it supports 11ac ? In the wireless mode drop down for the 5ghz band there is only a and n.

There is an option for 80mhz channel width.

In my experience it does (@mdf can comment further as he bought my router).

I have not used one since last year. Had the Edgerouter since.

You know, that is an excellent question. Is there a way to determine conclusively whether you're connecting to a 5GHz AC or N network? Other than the speed? I'm being somewhat limited by the AC devices I have available.

From the looks of things, there isn't an option to force AC only.

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | ... | 11