Network Isolation

[taipan254]

2 hours ago, jabbr said:

Interestingly, I haven't used it but Solid-Run has a 10G SFP(+) single board:

https://shop.solid-run.com/product/SRS91-EVKCB-R01/

 

Cool board - looks a lot cheaper than an opticalrendu! Also lots of flexibility on the power supply. 

 

I assume you flash an OS onto the the SDCard and then tunnel in to configure given there's no HDMI out? 

 

What is everyone's favorite way to get NAA up and running on this kind of set up? 

[jabbr]

1 hour ago, taipan254 said:

 

Cool board - looks a lot cheaper than an opticalrendu! Also lots of flexibility on the power supply. 

 

I assume you flash an OS onto the the SDCard and then tunnel in to configure given there's no HDMI out? 

 

What is everyone's favorite way to get NAA up and running on this kind of set up? 

 

Yes! Load the OS onto SD card and then install onto eMMC. I'm pretty sure microUSB is console

[wklie]

9 hours ago, taipan254 said:

Can you run your 10G Fiber directly into your Lumin and just run it at 1G speed?

 

I think there's a high probability that this won't work.

[SQFIRST]

23 minutes ago, wklie said:

10 hours ago, taipan254 said:

Can you run your 10G Fiber directly into your Lumin and just run it at 1G speed?

 

I think there's a high probability that this won't work.

Correct. It wasn't working that way. The switch would not connect 10g to 1g. I could connect by using 1g spfs on both the lumin and the 10g-switch, but that does not provide the 10g isolation by itself. The only time I saw the isolation benefit was when both switches were connected and operating at 10g speed. Just need that one connection for the isolation to be in place. 

 

As simple as this solution sounds, it is not intuitive or at all easy to solve because most people are only dealing with 1g speed rated devices. It will be great few years from now when other possible enhancements mitigate this type of issue, or when 1g is no longer the norm.

[jabbr]

9 hours ago, wklie said:

 

I think there's a high probability that this won't work.

There are 10/1g hybrid SPF(+) that plug into the switch and communicate with 1g SFP modules on the endpoint. The protocol runs at 1000base-X.

 

e.g. https://ii-vi.com/product/10g-1g-dual-rate-10gbase-sr-and-1000base-sx-400m-multimode-datacom-sfp-optical-transceiver/

 

If you want to send me one of your DACs I promise to get it working else I'll send it back 😉

[jabbr]

8 hours ago, SQFIRST said:

The only time I saw the isolation benefit was when both switches were connected and operating at 10g speed. Just need that one connection for the isolation to be in place. 

If you want to use one of the baby Mikrotik switches with 2 SFP(+) ports as a cost effective FMC $120 and you want guaranteed 10g link, then use one to feed into the Mikrotik you already have.

[SQFIRST]

51 minutes ago, jabbr said:

There are 10/1g hybrid SPF(+) that plug into the switch and communicate with 1g SFP modules on the endpoint. The protocol runs at 1000base-X.

Would the 10g related isolation still apply when connecting down at 1000base-x using just one 10g switch?

[jabbr]

9 minutes ago, SQFIRST said:

Would the 10g related isolation still apply when connecting down at 1000base-x using just one 10g switch?

When the switch receives the packets over the 10Gbase-X protocol then it is required to reduce jitter and noise. Typically when the switch receives packets using the 1000base-X protocol, it will go through the same circuitry as 10G but that's not guaranteed. If I, for example connect my Mellanox switch using a 25G SFP28 port to a 1g device then isolation occurs. 

 

So ... if your data comes into the 10g switch a 1g rate you might assume that the 10g isolation occurs but if you want 100% then use a baby switch with 1g in and 10g out to your 10g switch. 

 

Perhaps Lumin @wklie might want to test this out? 

[SQFIRST]

On 1/31/2023 at 9:29 AM, jabbr said:

The essential question:

 

1) if the network segment is “perfectly isolating” do upstream devices matter eg cables, servers, power supplies?

2) does everything really matter?

3) if so, how? How might effects cross a network segment that doesn’t pass differential mode, common mode nor phase mode noise?

 

Prior to just few days ago, before I tried the 10g link in my network, my answers to these questions would have been different. After introducing just one single complete 10g connection in my chain, I can confirm that:

1) No. 

2) To the extent that the device modifies the audio signal, such as cpu changes, dsp etc. For regular usage I am seeing proper isolation to be fully effective.

3) Still early days for me but so far nothing. Will report back if I hit something new.

 

Thank you @jabbr for your help.

[jabbr]

3 minutes ago, SQFIRST said:

 

Prior to just few days ago, before I tried the 10g link in my network, my answers to these questions would have been different. After introducing just one single complete 10g connection in my chain, I can confirm that:

1) No. 

2) To the extent that the device modifies the audio signal, such as cpu changes, dsp etc. For regular usage I am seeing proper isolation to be fully effective.

3) Still early days for me but so far nothing. Will report back if I hit something new.

 

Thank you @jabbr for your help.

 

Interesting ... it would be great if other folks can confirm this. I can't test this anymore because my servers are all connected to the network by 100G fiberoptic, so as you can see from my above network I have multiple hops... I honestly can't hear USB reclocking anymore

[SQFIRST]

@jabbr Your case is different as your network is already at 100gb and possibly climbing, but hope you aren't suggesting that you didn't test what you have been advocating? :) 

 

For myself, the main takeaway, from this recent revival of the isolation topic was that if you are only working with 1gb speed devices, chances are that there will be some noise that creeps into the audio. For complete network isolation, adding a 10g (or higher) connection will be necessary. The method by which the 10g is added is, of course, flexible as there are FMCs and switches available. 

 

It will be nice to hear from others on their experience and also from those that already are using high speed networking for any isolation insights.

 

 

 

[SQFIRST]

Additional finding from my setup: I am seeing that any direct use of 1000base-x (1g fiber) within the network isolation segment introduces some noise. For complete isolation I had to relocate any 1g fiber to outside of the isolation segment or just use copper 1000base-t. My personal conclusion (not an expert one) is that for effective network isolation one needs to completely avoid 1000base-x and use any combination of 1000base-t and/or 10gbe (or higher). Usage of 1g fiber outside of the audio network isolation segment does not matter.

[jabbr]

4 hours ago, SQFIRST said:

Additional finding from my setup: I am seeing that any direct use of 1000base-x (1g fiber) within the network isolation segment introduces some noise. For complete isolation I had to relocate any 1g fiber to outside of the isolation segment or just use copper 1000base-t. My personal conclusion (not an expert one) is that for effective network isolation one needs to completely avoid 1000base-x and use any combination of 1000base-t and/or 10gbe (or higher). Usage of 1g fiber outside of the audio network isolation segment does not matter.

 

Are you saying that you prefer the copper over the fiberoptic input to your Lumin? I'm not sure that's the general conclusion from what I've read.

[SQFIRST]

10 minutes ago, jabbr said:

Are you saying that you prefer the copper over the fiberoptic input to your Lumin? I'm not sure that's the general conclusion from what I've read.

Yes. It is now.

 

Optical was my preferred connection until my recent and effective network isolation change, that required removing all 1g fiber to achieve the isolation. I am sure this is setup specific. But for my particular set of components, I just could not get proper isolation using 1000base-x.

 

[SQFIRST]

On 2/6/2023 at 8:08 AM, SQFIRST said:

Additional finding from my setup: I am seeing that any direct use of 1000base-x (1g fiber) within the network isolation segment introduces some noise. For complete isolation I had to relocate any 1g fiber to outside of the isolation segment or just use copper 1000base-t. My personal conclusion (not an expert one) is that for effective network isolation one needs to completely avoid 1000base-x and use any combination of 1000base-t and/or 10gbe (or higher). Usage of 1g fiber outside of the audio network isolation segment does not matter.

 

Need to follow up and report that this was a faulty grounding issue causing the noise and not the 1000base-x as I mentioned above. While overall my setup does benefit from reduced connections it was not necessary for the isolation. My conclusion was off.

If anyone can tell me how to update a post I will appreciate as I could not figure out how to do that.

[SQFIRST]

On 2/3/2023 at 12:30 AM, wklie said:

On 2/2/2023 at 2:34 PM, taipan254 said:

Can you run your 10G Fiber directly into your Lumin and just run it at 1G speed?

 

I think there's a high probability that this won't work.

 

On 2/3/2023 at 9:46 AM, jabbr said:

There are 10/1g hybrid SPF(+) that plug into the switch and communicate with 1g SFP modules on the endpoint. The protocol runs at 1000base-X.

 

On 2/3/2023 at 10:48 AM, jabbr said:

On 2/3/2023 at 10:39 AM, SQFIRST said:

Would the 10g related isolation still apply when connecting down at 1000base-x using just one 10g switch?

When the switch receives the packets over the 10Gbase-X protocol then it is required to reduce jitter and noise. Typically when the switch receives packets using the 1000base-X protocol, it will go through the same circuitry as 10G but that's not guaranteed.

Sharing additional findings from my setup regarding the above premise as originally asked by @taipan254 

For the Mikrotik CRS305, which is what I am using, I tested out the scenario by using dual rate SFP+ transceivers (Finisar FTLX1471D3BCV) and it does indeed, use the 10GBase-x protocol when operating at 1G speed. One can achieve 10G isolation using a single 10G device with dual rate SFP+ modules. The CRS305 is a managed switch and it is bit tricky to make the rate setting work. I realize this post is better suited for the optical networking thread but wanted to close out the prior discussion here.