Question Managed ethernet switch with redundancy

[think2]

We are currently using MOXA 405A 5 port managed ethernet switches connected in a ring topology because we need redundancy / dual path between any two devices, we also need QOS. These switches are kind of expensive yet the manufacturer describes them as "entry level". What determines the price of an ethernet switch - is it the amount of memory or the amount of software features or the switching speed or what? Are there any cheaper managed ethernet switches that provide redundancy/ ring topology? They need to keep their configuration settings whilst powered down.

 

[sdifox]

What do you mean ring topology exactly? Draw a picture of your config. Also, 5 port unmanaged switches are not meant to be used as a business network solution.

 

[think2]

It's a ring - as in circle. It's a managed switch not an unmanaged switch and it's an industrial application not a business solution.

 

[sdifox]

It's a ring - as in circle. It's a managed switch not an unmanaged switch and it's an industrial application not a business solution.

Just a simple ascii picture.


Is it like this?

--- hub1 ---- machine1 --- hub2 --- machine2 --- back to hub1?

 

[think2]

yes, switch 1 - - - switch 2 - - - switch 3 - - - // - - - switch N - - - back to switch one.
If there's a single break anywhere, all switches are still connected.

 

[VirtualLarry]

Hmm. Make sure Spanning-Tree is enabled. *chuckle*.

 

[sdifox]

yes, switch 1 - - - switch 2 - - - switch 3 - - - // - - - switch N - - - back to switch one.
If there's a single break anywhere, all switches are still connected.

Can you give a sense of your needs and number of devices that need to be connected? Whole bunch of small managed switches is definitely more expensive than a few big switches.
Hub and spoke makes more sense to me.

Also do you have some sort of environmental constraints?

 

[think2]

Can be up to 64 switches, sometimes connected with fibre because it goes further. There is a requirement for temperature tolerance but I don't think it's anything special.

 

[sdifox]

Can be up to 64 switches, sometimes connected with fibre because it goes further. There is a requirement for temperature tolerance but I don't think it's anything special.

Any reason why it's not hub and spoke or clusters of hub and spoke? Wires are cheaper than switches. Do the networked devices need to talk to each other or do they just talk to server? Having to traverse a ring of switches is not ideal.

How much traffic does each "computer" generate and what is their latency envelope?

 

[think2]

Yes, all devices need to talk to all other devices. Hub and spoke doesn't provide redundancy. The MOXA switch has a "turbo ring" mode.

 

[think2]

oops, missed your second question. Traffic is fairly light, except when VOIP is happening but even then it's not too bad. Most traffic doesn't care about latency except for a homegrown version of NTP that we use for synchronization of audio.

 

[sdifox]

Yes, all devices need to talk to all other devices. Hub and spoke doesn't provide redundancy. The MOXA switch has a "turbo ring" mode.

How so? Nothing stops you from doing double tap hub and spoke. Unless you lose your central stack but that is rare.

Your machines don't have profinet hub built in?


<---- not an expert

Anything "Industrial" cost more than "business" because of environment hardening so you pay for that more than the actual nework performance. And for stuff like turbo ring. But if you don't need those things, you can put in "business" hardware.

 

[think2]

Not sure what you mean by double tap but the connecting device has only one ethernet port and any "duplicate wiring" would have to traverse physically separated paths. There would also be cable length problems I think.

I don't know what profinet is, except for what I can see on wikipedia.

 

[think2]

Thanks for the info on environment hardening. I'll check what the requirements are.

 

[sdifox]

Not sure what you mean by double tap but the connecting device has only one ethernet port and any "duplicate wiring" would have to traverse physically separated paths. There would also be cable length problems I think.

I don't know what profinet is, except for what I can see on wikipedia.

If each of the devices have only one ethernet port, you don't really have redundancy to that device. If you are not doing industrial control where timing is important then I fail to see how the ring of small managed switches pays off. What are you using the turbo ring feature for?


Even keeping with the same topology, what keeps you from using 32 8 port switch as opposed to 64 5 port switch?

Bottom line you are paying for gigabit prices but getting 100mbit if you don't need all the hardening or induatrial protocol.

 

[mxnerd]

Searched. Interesting.

MOXA proprietary Turbo Ring redundancy.

Moxa Redundancy Protocol

https://www.moxa.com/doc/man/Moxa_E_Series_Managed_Ethernet_Switch_Redudancy_Protocol_UM_v2.pdf

 

[sdifox]

Searched. Interesting.

MOXA proprietary Turbo Ring redundancy.

Moxa Redundancy Protocol

https://www.moxa.com/doc/man/Moxa_E_Series_Managed_Ethernet_Switch_Redudancy_Protocol_UM_v2.pdf

I worked with some weidmueller sensors in the early 90s. 68k assembly. Getting the cycle down to the token window was a nightmare.

They probably wouldn't have been happy to learn coop students were coding mine (as in digging into earth) sensors xd

 

[JackMDS]

Ring Network! "Wow", last time I dealt with one was 30 years ago, even then they were in its way to extinction.

It main use was by IBM (in a version called Token Ring) that was dominating whole the levels of Computers' related work at the time.

https://en.wikipedia.org/wiki/Token_ring

Well, for for ever reason some of the Rings are still in existence and because it so rare prices of the hardware involved with it is High.

So... an entry level Ethernet switch can be bellow $20 and Ring related switch is more expensive.

The fact that it call Entry level does Not make them equal just like a Smart car and a Ferrari both are called "Cars" it does not make the Ferrari cost like Smart car. (it is called false verbal analogies).

 

[mxnerd]

Moxa's Turbo Ring is proprietary. There is an ITU-T G.8032 Ethernet Ring standard , probably similar to what Moax has, I guess.

But G.8032 based switches not cheap either.

 

[think2]

If each of the devices have only one ethernet port, you don't really have redundancy to that device. If you are not doing industrial control where timing is important then I fail to see how the ring of small managed switches pays off. What are you using the turbo ring feature for?


Even keeping with the same topology, what keeps you from using 32 8 port switch as opposed to 64 5 port switch?

Bottom line you are paying for gigabit prices but getting 100mbit if you don't need all the hardening or induatrial protocol.

It's redundant as far as the installation standards requirement is concerned. The redundancy only applies to a fault in external wiring plus a total failure of one device isn't allowed to affect communication between other devices. The MOXA switch must be providing electrical isolation between ports. Reason for using turbo ring - I'm guessing it makes installation simpler than more general RSTP (spanning tree) - or maybe recovery from a fault is faster and more reliable. In general, data packets need to arrive in the order they are sent. Most messages are sent to all devices and we're (mostly) not using broadcast or multicast.

I don't understand the question about 32 versus 64 switches.

 

[sdifox]

It's redundant as far as the installation standards requirement is concerned. The redundancy only applies to a fault in external wiring plus a total failure of one device isn't allowed to affect communication between other devices. The MOXA switch must be providing electrical isolation between ports. Reason for using turbo ring - I'm guessing it makes installation simpler than more general RSTP (spanning tree) - or maybe recovery from a fault is faster and more reliable. In general, data packets need to arrive in the order they are sent. Most messages are sent to all devices and we're (mostly) not using broadcast or multicast.

I don't understand the question about 32 versus 64 switches.

1 8 port switch is cheaper than 2 5 port. That is all I was trying to get across. You do get faster recovery time with solution such as moxa. I just find it odd that you have voip running on ethernet/ip as well. I don't quite understand what you mean by packets arrive in order. UDP may arrive out of order but TCP handles that.

I doubt that switch electrically isolate each of the ports.

Interesting stuff. Care to share what you guys do that require this type of network?

 

[think2]

Regarding packet ordering I was just wondering what happens when a break on the ring occurs whether any packets that have already been sent out get re-routed to go in the other direction. I don't really want to say what we do, sorry.

 

[sdifox]

Regarding packet ordering I was just wondering what happens when a break on the ring occurs whether any packets that have already been sent out get re-routed to go in the other direction. I don't really want to say what we do, sorry.

TCP will request packet resend. Only difference is the downed link detection and path reconfiguration speed. That is why I was asking about how time sensitive it is. RSTP alone is like ren seconds, Turbo Ring claims 20ms.

Key pieces of info like vibraton, heat, accessiblity, physical space constraint, power supply, service level agreement all play a part in deciding what to get.