MetaRouter performance/througput another ROS ?

1st - I am new to Mikrotik Metrarouter (I just built my first on).

Question:
- If I use something like a Mikrotik RouterBOARD 952Ui-5ac2nD
- If the hosting Mikrotik is only hosting a second MetaRouter (another ROS)
- Where the hosting Mikrotik has no work load - other than a management IP address and some bridges for the hosted MetaRouter ROS
- Where the hosted second MetaRouter image is performing NAT (Live IP to customer NAT network - and a DHCP server and DHCP client)
What kind of throughput should I expect ?
If the only real work the physical Mikrotik is performing is running a single MetaRouter ROS image, will the hosted MetaRouter ROS perform around 80 percent as fast as the physical Mikrotik ?

What I am considering is the following:
Something like a RouterBOARD 952Ui-5ac2nD with a Mikrotik SFP GPON module - and a private management vlan IP address for management
Where this acts like a type of SubH Out-Of-Band-Management
Where the MetaRouter hosted ROS is a full NAT device for a residential home network (NAT - WAN is DHCP client - LAN is a dhcp server.

How reliable is something like this?
Is there much of a performance hit when doing this?

EDIT: another question
How well do 2 to 4 MetraRouter images work running all at the same time? Lets say I have Fiber *1-Gig or GPON) to the Mikrotik and I have a multi-dwelling customer location (2 or 4 homes). Where each dwelling is running their own MetrRouter hosted ROS or MetaRouter hosted OpenWRT image (DHCP client - NAT - DHCP server for each home). Where I have only a single physical Mikrotik hosting 4 MetaRouter virtual customer home networks.



North Idaho Tom Jones

I have never attempted to measure it in terms of throughput performance, but there is definitely a performance hit with even a single MetaROUTER guest, and even if that guest is doing zero work. And this performance hit significantly affects the host, too, because it turns out that if you are running a MetaROUTER, the "host" RouterOS instance is itself also virtualized (while given special/privileged direct access to hardware).

It does not seem to decrease linearly with the number of running guests. The very first running guest causes the biggest performance hit, and subsequent guests only cause minimal performance degradation (unless they are actively doing something / hogging the CPU of course).

You can see a post of mine from a few years back, in the context of a larger discussion about MetaROUTER performance and stability, where this is clearly demonstrated, and MikroTik acknowledged that this is just the way it works and that it is "working as designed": viewtopic.php?f=15&t=60746&p=371013#p371013

It wouldn't surprise me if certain tasks were affected more than others. Perhaps forwarding packets is not as heavy a task and although you might see a slowdown, it might not be that huge. It's probably worth your time to bench and measure it with the specific configuration you have in mind. In another situation, we had to split the duties we had delegated to a MetaROUTER to a separate device entirely, because lookups to the host's DNS cache were taking forever and causing CPU to spike to 100% with only moderate DNS load as long as there was a MetaROUTER guest running on the device. Once we shut it down, everything was fine.

MetaROUTER has/had a lot of promise, but still has a ways to go and is in dire need of some TLC at this point.

-- Nathan

P.S. -- Greetings from Moscow, ID, your neighbors to the south