Jump to content
IGNORED

Managed switch...best settings for optimal performance?


Recommended Posts

I recently acquired a Cisco SG300-10 managed switch for my audio system. I use to for the 2 optical ports to connect my server pc running HQ Player and the NAA pc. While I got it working last night I am frankly overwhelmed with all of the features (L2 vs L3, VLAN, Mulitcast, QOS, Jumbo Frames, etc). In my case since I only use this switch for my audio devices, all I should care about is speed and minimal latency? I was hoping that some networking guru could provide some advice on which features I should enable and what settings to use for absolute best performance in my audio only scenario.

 

Thank you.

12TB NAS >> i7-6700 Server/Control PC >> i3-5015u NAA >> Singxer SU-1 DDC (modded) >> Holo Spring L3 DAC >> Accustic Arts Power 1 int amp >> Sonus Faber Guaneri Evolution speakers + REL T/5i sub (x2)

 

Other components:

UpTone Audio LPS1.2/IsoRegen, Fiber Switch and FMC, Windows Server 2016 OS, Audiophile Optimizer 3.0, Fidelizer Pro 6, HQ Player, Roonserver, PS Audio P3 AC regenerator, HDPlex 400W ATX & 200W Linear PSU, Light Harmonic Lightspeed Split USB cable, Synergistic Research Tungsten AC power cords, Tara Labs The One speaker cables, Tara Labs The Two Extended with HFX Station IC, Oyaide R1 outlets, Stillpoints Ultra Mini footers, Hi-Fi Tuning fuses, Vicoustic/RealTraps/GIK room treatments

Link to comment
Share on other sites

L2 vs L3, VLAN, Mulitcast, QOS, Jumbo Frames, etc

 

These are all advanced network features that you're never likely to use in a single home install. If you had a lot of people using the network you might create a distinct VLAN dedicated for audio, or you could use QoS to limit the rest of the house from gobbling too much bandwidth with Netflix or something. Putting a lot of people on the network seems the opposite of your goals though.

 

It's possible to drop the CPU usage on your hardware by turning on Jumbo Frames, which used to be necessary to get full performance from 1000Mbps links. But everything has to support that and be configured for it, which is a pain even when it's needed. There's very little reason to consider it at all nowadays, and no reason I can think of for it in an audio environment.

 

The latency of this grade of network switch is so low you might as well be worried about sunspots as it impacting your sound quality. The one thing that can ripple out to audible blips is automatic rate negotiation. Your network devices and the switch are going to talk to determine the highest speed they can both handle. Rarely, that negotiating can happen more than it's supposed to, and the switch will bounce between two speeds--"flapping" is the usual word used.

 

I wouldn't worry about that either, not unless you notice the link is dropping regularly or there's a lot of messages about rate changes. If you really want to be paranoid or you run into a problem, what you do is fix the rate of the equipment so it will only run at one speed. Instead of letting it pick 100Mbps or 1000Mbps, you only let it run at the speed you know works. This also used to be a regular thing to tune, but you'd have to be pretty unlucky to see flapping pop up at home here in 2016.

Link to comment
Share on other sites

Hi @gregs1104 - I agree with you 100%.

 

In my experience, just having a managed switch is all one will need in such a system as the op. Managed switches frequently have greater bandwidth or are less over subscribed etc... I have a Netgear GS108 that I can stick in my network to slow everything down. Most people think Gigabit is Gigabit, but that just not the case.

Founder of Audiophile Style

Announcing The Audiophile Style Podcast

Link to comment
Share on other sites

Hi @gregs1104 - I agree with you 100%.

 

In my experience, just having a managed switch is all one will need in such a system as the op. Managed switches frequently have greater bandwidth or are less over subscribed etc... I have a Netgear GS108 that I can stick in my network to slow everything down. Most people think Gigabit is Gigabit, but that just not the case.

 

I've just bought an unmanaged switch to replace a managed switch on the basis that I couldn't tell if I was messing things up by inadvertent settings in the managed switch. So I'm sorry but I don't think it really helps people like me or the OP to say all you need is a managed switch without saying what you should do with the settings. It's far from clear how to avoid having any of its 'unnecessary' features on unhelpful or conflicting settings. Needless to say the manuals are written for the people who need them least.

Link to comment
Share on other sites

Agree with this in probably 99% of cases.

 

Only thing I can think of in a busy home... If your NAS has two Ethernet ports you can configure 'port trunking' (e.g. LACP) to aggregate the speed of the two connections (ports).

 

Even so, this usually only benefits very heavy load with multiple concurrent devices using the NAS.

 

LACP has to be configured on both the switch and the NAS to work properly - I have done this with mine, can't tell if there is any gain because I can't load it with enough data heavy connections.

 

Just my 2p.

 

These are all advanced network features that you're never likely to use in a single home install.

Source:

*Aurender N100 (no internal disk : LAN optically isolated via FMC with *LPS) > DIY 5cm USB link (5v rail removed / ground lift switch - split for *LPS) > Intona Industrial (injected *LPS / internally shielded with copper tape) > DIY 5cm USB link (5v rail removed / ground lift switch) > W4S Recovery (*LPS) > DIY 2cm USB adaptor (5v rail removed / ground lift switch) > *Auralic VEGA (EXACT : balanced)

 

Control:

*Jeff Rowland CAPRI S2 (balanced)

 

Playback:

2 x Revel B15a subs (balanced) > ATC SCM 50 ASL (balanced - 80Hz HPF from subs)

 

Misc:

*Via Power Inspired AG1500 AC Regenerator

LPS: 3 x Swagman Lab Audiophile Signature Edition (W4S, Intona & FMC)

Storage: QNAP TS-253Pro 2x 3Tb, 8Gb RAM

Cables: DIY heavy gauge solid silver (balanced)

Mains: dedicated distribution board with 5 x 2 socket ring mains, all mains cables: Mark Grant Black Series DSP 2.5 Dual Screen

Link to comment
Share on other sites

Hi @gregs1104 - I agree with you 100%.

 

In my experience, just having a managed switch is all one will need in such a system as the op. Managed switches frequently have greater bandwidth or are less over subscribed etc... I have a Netgear GS108 that I can stick in my network to slow everything down. Most people think Gigabit is Gigabit, but that just not the case.

Hi Chris,

could you elaborate your findings on the Netgear GS108, as that product exists in different flavours as unmanaged and managed switches? I have one GS108T v2 (managed) waiting to be inserted in the network. After your comment I am wondering whether it is a good idea or not.

Tom

Link to comment
Share on other sites

I've just bought an unmanaged switch to replace a managed switch on the basis that I couldn't tell if I was messing things up by inadvertent settings in the managed switch. So I'm sorry but I don't think it really helps people like me or the OP to say all you need is a managed switch without saying what you should do with the settings. It's far from clear how to avoid having any of its 'unnecessary' features on unhelpful or conflicting settings. Needless to say the manuals are written for the people who need them least.

 

Hi @craighartley - Maybe I could have been clearer. What I mean is that getting a managed switch and leaving it in the default configuration is all you need to do. I know this sounds strange but I have experience with this. Managed switches are often built better than unmanaged.

Founder of Audiophile Style

Announcing The Audiophile Style Podcast

Link to comment
Share on other sites

Hi Chris,

could you elaborate your findings on the Netgear GS108, as that product exists in different flavours as unmanaged and managed switches? I have one GS108T v2 (managed) waiting to be inserted in the network. After your comment I am wondering whether it is a good idea or not.

Tom

Hi Tom - I have an unmanaged GS108 and can cause my network performance to go down when it's inserted into my network. For example, some audio components actually have a hard time accepting all the data that can be sent to them on a very fast network. If the network is very good you can get audio dropouts because the end point is dropping packets. Kind of like catching water from a fire hose with a shot glass. If I place the GS108 in the network I can solve the issue without changing anything else. It slows traffic enough to make everything work. This is actually a help not a hinderance in this specific case, but in a perfect world I'd go for the best performing network.

Founder of Audiophile Style

Announcing The Audiophile Style Podcast

Link to comment
Share on other sites

I tend to use HP Procurve switches, they never seem to miss a beat (at home and at work)

 

in a perfect world I'd go for the best performing network.

Source:

*Aurender N100 (no internal disk : LAN optically isolated via FMC with *LPS) > DIY 5cm USB link (5v rail removed / ground lift switch - split for *LPS) > Intona Industrial (injected *LPS / internally shielded with copper tape) > DIY 5cm USB link (5v rail removed / ground lift switch) > W4S Recovery (*LPS) > DIY 2cm USB adaptor (5v rail removed / ground lift switch) > *Auralic VEGA (EXACT : balanced)

 

Control:

*Jeff Rowland CAPRI S2 (balanced)

 

Playback:

2 x Revel B15a subs (balanced) > ATC SCM 50 ASL (balanced - 80Hz HPF from subs)

 

Misc:

*Via Power Inspired AG1500 AC Regenerator

LPS: 3 x Swagman Lab Audiophile Signature Edition (W4S, Intona & FMC)

Storage: QNAP TS-253Pro 2x 3Tb, 8Gb RAM

Cables: DIY heavy gauge solid silver (balanced)

Mains: dedicated distribution board with 5 x 2 socket ring mains, all mains cables: Mark Grant Black Series DSP 2.5 Dual Screen

Link to comment
Share on other sites

Hi Tom - I have an unmanaged GS108 and can cause my network performance to go down when it's inserted into my network. For example, some audio components actually have a hard time accepting all the data that can be sent to them on a very fast network. If the network is very good you can get audio dropouts because the end point is dropping packets. Kind of like catching water from a fire hose with a shot glass. If I place the GS108 in the network I can solve the issue without changing anything else. It slows traffic enough to make everything work. This is actually a help not a hinderance in this specific case, but in a perfect world I'd go for the best performing network.

Thank you, Chris, for that information/clarification.

Tom

Link to comment
Share on other sites

I found this presentation by Audinate that explains really well some of the advanced features my managed switch has. I started messing around with QoS yesterday a bit to ensure priority is given to certain data streams. I am not sure if it makes a difference but its good piece of mind for me to know I am optimizing as much of my audio chain as I can.

 

https://www.audinate.com/sites/default/files/PDF/advanced-dante-networking-avnw-2015-audinate.pdf

12TB NAS >> i7-6700 Server/Control PC >> i3-5015u NAA >> Singxer SU-1 DDC (modded) >> Holo Spring L3 DAC >> Accustic Arts Power 1 int amp >> Sonus Faber Guaneri Evolution speakers + REL T/5i sub (x2)

 

Other components:

UpTone Audio LPS1.2/IsoRegen, Fiber Switch and FMC, Windows Server 2016 OS, Audiophile Optimizer 3.0, Fidelizer Pro 6, HQ Player, Roonserver, PS Audio P3 AC regenerator, HDPlex 400W ATX & 200W Linear PSU, Light Harmonic Lightspeed Split USB cable, Synergistic Research Tungsten AC power cords, Tara Labs The One speaker cables, Tara Labs The Two Extended with HFX Station IC, Oyaide R1 outlets, Stillpoints Ultra Mini footers, Hi-Fi Tuning fuses, Vicoustic/RealTraps/GIK room treatments

Link to comment
Share on other sites

Agree with this in probably 99% of cases.

 

Only thing I can think of in a busy home... If your NAS has two Ethernet ports you can configure 'port trunking' (e.g. LACP) to aggregate the speed of the two connections (ports).

 

Even so, this usually only benefits very heavy load with multiple concurrent devices using the NAS.

 

LACP has to be configured on both the switch and the NAS to work properly - I have done this with mine, can't tell if there is any gain because I can't load it with enough data heavy connections.

 

Just my 2p.

 

You are thinking about LAG(link aggregate group) not Port Trunking. Port Trunking is dedicated links from switch to switch.

 

LACP allows for LAG's to be dynamically built. Also be careful. LAG's generally only guarantee redundancy and not increased bandwidth.

 

Ports on a Switch have three configuration modes: Access, Dynamic, Trunk.

Link to comment
Share on other sites

I found this presentation by Audinate that explains really well some of the advanced features my managed switch has. I started messing around with QoS yesterday a bit to ensure priority is given to certain data streams. I am not sure if it makes a difference but its good piece of mind for me to know I am optimizing as much of my audio chain as I can.

 

https://www.audinate.com/sites/default/files/PDF/advanced-dante-networking-avnw-2015-audinate.pdf

 

Don't use QOS when Vlan's are readily at your disposal and the switch support intervlan routing.

Link to comment
Share on other sites

Do: Use Vlan with InverVlan routing. Don't use QOS since you won't need it.

Do: Use jumbo frames if your endpoints support it

 

Using Vlan's has the benefit of giving you another, smaller, collision domain. It also sets up a smaller broadcast domain. QOS does neither of those for you.

Link to comment
Share on other sites

Do: Use Vlan with InverVlan routing. Don't use QOS since you won't need it.

Do: Use jumbo frames if your endpoints support it

 

Using Vlan's has the benefit of giving you another, smaller, collision domain. It also sets up a smaller broadcast domain. QOS does neither of those for you.

 

Thank you for the reply. Perhaps a dumb question but what on the end point needs to support Jumbo Frames? Is it the OS or the application receiving the data? Is there anything I need to do at the end points if its supported? My end points are two different Server R2 machines, one running HQ Player and the other NAA.

12TB NAS >> i7-6700 Server/Control PC >> i3-5015u NAA >> Singxer SU-1 DDC (modded) >> Holo Spring L3 DAC >> Accustic Arts Power 1 int amp >> Sonus Faber Guaneri Evolution speakers + REL T/5i sub (x2)

 

Other components:

UpTone Audio LPS1.2/IsoRegen, Fiber Switch and FMC, Windows Server 2016 OS, Audiophile Optimizer 3.0, Fidelizer Pro 6, HQ Player, Roonserver, PS Audio P3 AC regenerator, HDPlex 400W ATX & 200W Linear PSU, Light Harmonic Lightspeed Split USB cable, Synergistic Research Tungsten AC power cords, Tara Labs The One speaker cables, Tara Labs The Two Extended with HFX Station IC, Oyaide R1 outlets, Stillpoints Ultra Mini footers, Hi-Fi Tuning fuses, Vicoustic/RealTraps/GIK room treatments

Link to comment
Share on other sites

Thank you for the reply. Perhaps a dumb question but what on the end point needs to support Jumbo Frames? Is it the OS or the application receiving the data? Is there anything I need to do at the end points if its supported? My end points are two different Server R2 machines, one running HQ Player and the other NAA.

 

The manufacturer of the NIC is what determines all the various features they support. This will be available from the properties of the adapter.

 

Are you running GUI's? PowerShell? or a machine with RSAT to manage them?

 

The application has zero idea what is going on since things are packetized in the Session layer.

Link to comment
Share on other sites

Well yes if you are referring to what Cisco call it, they call trunking 'switch to switch', but HP and QNAP call LACP (802.3ad) 'port trunking'.

 

https://www.qnap.com/en/tutorial/con_show.php?op=showone&cid=216

 

To be fair it's probably not worth doing for most home cases. I did it because I'm familiar with setting up managed networks and thought why not.

 

At work I can see both LANs being load balanced when two or more PCs access the NAS.

 

 

You are thinking about LAG(link aggregate group) not Port Trunking. Port Trunking is dedicated links from switch to switch.

 

LACP allows for LAG's to be dynamically built. Also be careful. LAG's generally only guarantee redundancy and not increased bandwidth.

 

Ports on a Switch have three configuration modes: Access, Dynamic, Trunk.

Source:

*Aurender N100 (no internal disk : LAN optically isolated via FMC with *LPS) > DIY 5cm USB link (5v rail removed / ground lift switch - split for *LPS) > Intona Industrial (injected *LPS / internally shielded with copper tape) > DIY 5cm USB link (5v rail removed / ground lift switch) > W4S Recovery (*LPS) > DIY 2cm USB adaptor (5v rail removed / ground lift switch) > *Auralic VEGA (EXACT : balanced)

 

Control:

*Jeff Rowland CAPRI S2 (balanced)

 

Playback:

2 x Revel B15a subs (balanced) > ATC SCM 50 ASL (balanced - 80Hz HPF from subs)

 

Misc:

*Via Power Inspired AG1500 AC Regenerator

LPS: 3 x Swagman Lab Audiophile Signature Edition (W4S, Intona & FMC)

Storage: QNAP TS-253Pro 2x 3Tb, 8Gb RAM

Cables: DIY heavy gauge solid silver (balanced)

Mains: dedicated distribution board with 5 x 2 socket ring mains, all mains cables: Mark Grant Black Series DSP 2.5 Dual Screen

Link to comment
Share on other sites

Well yes if you are referring to what Cisco call it, they call trunking 'switch to switch', but HP and QNAP call LACP (802.3ad) 'port trunking'.

 

https://www.qnap.com/en/tutorial/con_show.php?op=showone&cid=216

 

To be fair it's probably not worth doing for most home cases. I did it because I'm familiar with setting up managed networks and thought why not.

 

At work I can see both LANs being load balanced when two or more PCs access the NAS.

 

I tend to stick with Cisco's nomenclature since I started using them in 99. The reasoning is that if you have a port in Trunk mode, which is a real configuration option on switches, you'll have issues connecting a workstation to that port.

 

If an end user here gets into the command line or another UI and actually configures the port as a trunk port because they are led to believe 'trunk' = 'LAG/LACP' they are going to run into issues.

Link to comment
Share on other sites

The manufacturer of the NIC is what determines all the various features they support. This will be available from the properties of the adapter.

 

Are you running GUI's? PowerShell? or a machine with RSAT to manage them?

 

The application has zero idea what is going on since things are packetized in the Session layer.

 

Thank you for helping a networking noob. I read a bit more into this and as you said it looks like its as easy as editing the properties of the NIC. On both of my Server machines I am using powershell though I can switch back to GUI for ease of editing the adapter settings. I will give this a try and see what happens.

12TB NAS >> i7-6700 Server/Control PC >> i3-5015u NAA >> Singxer SU-1 DDC (modded) >> Holo Spring L3 DAC >> Accustic Arts Power 1 int amp >> Sonus Faber Guaneri Evolution speakers + REL T/5i sub (x2)

 

Other components:

UpTone Audio LPS1.2/IsoRegen, Fiber Switch and FMC, Windows Server 2016 OS, Audiophile Optimizer 3.0, Fidelizer Pro 6, HQ Player, Roonserver, PS Audio P3 AC regenerator, HDPlex 400W ATX & 200W Linear PSU, Light Harmonic Lightspeed Split USB cable, Synergistic Research Tungsten AC power cords, Tara Labs The One speaker cables, Tara Labs The Two Extended with HFX Station IC, Oyaide R1 outlets, Stillpoints Ultra Mini footers, Hi-Fi Tuning fuses, Vicoustic/RealTraps/GIK room treatments

Link to comment
Share on other sites

Yep, there's quite a bit of confusion over this on the web. I don't why there can't be a standard terminology, it would save everyone a lot of time and frustration.

 

 

 

I tend to stick with Cisco's nomenclature since I started using them in 99. The reasoning is that if you have a port in Trunk mode, which is a real configuration option on switches, you'll have issues connecting a workstation to that port.

 

If an end user here gets into the command line or another UI and actually configures the port as a trunk port because they are led to believe 'trunk' = 'LAG/LACP' they are going to run into issues.

Source:

*Aurender N100 (no internal disk : LAN optically isolated via FMC with *LPS) > DIY 5cm USB link (5v rail removed / ground lift switch - split for *LPS) > Intona Industrial (injected *LPS / internally shielded with copper tape) > DIY 5cm USB link (5v rail removed / ground lift switch) > W4S Recovery (*LPS) > DIY 2cm USB adaptor (5v rail removed / ground lift switch) > *Auralic VEGA (EXACT : balanced)

 

Control:

*Jeff Rowland CAPRI S2 (balanced)

 

Playback:

2 x Revel B15a subs (balanced) > ATC SCM 50 ASL (balanced - 80Hz HPF from subs)

 

Misc:

*Via Power Inspired AG1500 AC Regenerator

LPS: 3 x Swagman Lab Audiophile Signature Edition (W4S, Intona & FMC)

Storage: QNAP TS-253Pro 2x 3Tb, 8Gb RAM

Cables: DIY heavy gauge solid silver (balanced)

Mains: dedicated distribution board with 5 x 2 socket ring mains, all mains cables: Mark Grant Black Series DSP 2.5 Dual Screen

Link to comment
Share on other sites

Another question...should I disable flow control? My network is all gigabit.

12TB NAS >> i7-6700 Server/Control PC >> i3-5015u NAA >> Singxer SU-1 DDC (modded) >> Holo Spring L3 DAC >> Accustic Arts Power 1 int amp >> Sonus Faber Guaneri Evolution speakers + REL T/5i sub (x2)

 

Other components:

UpTone Audio LPS1.2/IsoRegen, Fiber Switch and FMC, Windows Server 2016 OS, Audiophile Optimizer 3.0, Fidelizer Pro 6, HQ Player, Roonserver, PS Audio P3 AC regenerator, HDPlex 400W ATX & 200W Linear PSU, Light Harmonic Lightspeed Split USB cable, Synergistic Research Tungsten AC power cords, Tara Labs The One speaker cables, Tara Labs The Two Extended with HFX Station IC, Oyaide R1 outlets, Stillpoints Ultra Mini footers, Hi-Fi Tuning fuses, Vicoustic/RealTraps/GIK room treatments

Link to comment
Share on other sites

could you elaborate your findings on the Netgear GS108, as that product exists in different flavours as unmanaged and managed switches? I have one GS108T v2 (managed) waiting to be inserted in the network.

 

I ended up buying 3 GS108T switches around 2011 or 2012 because they were one of the cheapest management switches around, and I was already a long time user of their unmanaged switches. They have been very hit or miss for me. Never had any issues on the performance, they've always been fast and easy to manage when they're working.

 

First: GS108T V1H1, Software 3.0.4.7. Works perfectly.

 

Second: GS108T V1H1, Software 3.0.4.10. Each time there's a power spike in the house, this switch has to be rebooted. The switch itself is on a UPS, but not every single device is. It loses its mind from that pretty routine situation. And, before someone asks, it did not go away downgrading back to 3.0.4.7.

 

Third: GS108T V2H1: This one gave me so many weird problem just during installation and testing that it lives in the spare parts bin instead of the live network. That was many years ago so the later firmware may have improved things. I suspect though that I just got a lemon because the V2 hardware was brand new.

 

My take: if your switch works for you, try some testing that involves nasty power cycling of the unit or devices attached to it. Flipping a breaker is how I stress test; I'm pretty hardcore about this. If your switch works through the testing, I wouldn't stress about it further. Every manufacturer has bad models and you may not see one the way I did.

Link to comment
Share on other sites

LAG's generally only guarantee redundancy and not increased bandwidth.

 

And going through the LAG processing makes latency less predictable. Does it slow down delivering to two ports instead of one? If both are active, what if one of them has an error but not both? I would not want to touch that feature unless I have a lot of time on my hands to experiment with it.

Link to comment
Share on other sites

And going through the LAG processing makes latency less predictable. Does it slow down delivering to two ports instead of one? If both are active, what if one of them has an error but not both? I would not want to touch that feature unless I have a lot of time on my hands to experiment with it.

 

Error would be framing and the network stack should take care of it be requesting a frame resent. LACP is the dynamic, on the fly configuration of LAG. You can omit LACP and statically configure the LAG.

 

Or you can roll with Microsoft SMB 3.0 and just let it team adapters and use bog standard $20 switches. They've moved all this functionality into their networking stack.

 

Plus not only is their solution link redundant it's bandwidth aggregation. I'm able to max out the SSD write performance over the network with my laptop and a SandDisk SSD.

 

But none of this is needed for audio. Just Vlan it, and Jumbo frame and turn on as many CPU off load features the NIC supports.

 

My 2.0 sits on it's own wireless network into VLAN 2 on my Dell PowerConnect switch.

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
 Share



×
×
  • Create New...