Assign a specific IP address to each switch port

[surferb]

It's a very unorthodox concept in networking to try to assign an IP address based on switch port. Why would you want to do that? Here are some typical designs:
1. use "static" addresses on devices. Manually intensive, but simple. You can track what devices will have what address. Common for PLCs.
2. use an assignment pool (DHCP or BOOTP). You don't care who gets what address. Sharing is done by other means (NETBIOS names, etc) or not done at all (Internet cafe, etc).
3. use DHCP with hardware (MAC address) reservations. This allows you to centrally manage addresses.
4. there are many schemes if the purpose is security. MAC filtering based on port, and many others.

 

[TurpoUrpo]

Yeah, I wonder too why in earth anyone would want to have IP address for every port in switch. It sounds like OP does not really know what IP address is.

 

[lostcontrol]

I doubt that the OP doesn't know what IP address's are.

There has been some good discussion here, and the merits and non-merits discussed.

The OP may of been looking from a maintenance point of view. i know I have had to talk on-site techs through config of drive IP setups, due to a few reasons. If it was 'plug n play', then those calls would not of been warranted.

The devices themselves could still have a 'table' allocated, so that this is all documented etc, and debugging/fault finding is a tad easier.

I think there are merits to it, but would I do it, not sure

I think if this was implemented, using no.3 in SurferB's post would be the best method, but what if the type of device changed?

 

[todster]

It's a very unorthodox concept in networking to try to assign an IP address based on switch port. Why would you want to do that?

Why would it be unorthodox? What if I want cabinet 1 to be xxx.xxx.xxx.101, cabinet 2 to be 102, BCR 1 to be 201, BCR 2 to be 202 ...etc..? I can't really use a MAC address to define a location.

 

[The Plc Kid]

It's a very unorthodox concept in networking to try to assign an IP address based on switch port. Why would you want to do that? Here are some typical designs:
1. use "static" addresses on devices. Manually intensive, but simple. You can track what devices will have what address. Common for PLCs.
2. use an assignment pool (DHCP or BOOTP). You don't care who gets what address. Sharing is done by other means (NETBIOS names, etc) or not done at all (Internet cafe, etc).
3. use DHCP with hardware (MAC address) reservations. This allows you to centrally manage addresses.
4. there are many schemes if the purpose is security. MAC filtering based on port, and many others.

Nathan

I have seen recently that a lot of new equipment is being designed this way. It is just a means to make component replacement easier for maintenance techs with little or no Computer / Networking skills. All they have to know is cable # xxx is connected to port #xxxx is for takeup roll drive or drive # xxxxx. Plug it in out of the box it has dhcp on by default gets it's address from the port and you are ready.

So in a way it is like static you can have a map of what devices connect to a switch and what device is on each port as well as what it's ip address is just like static the only difference is when a drive or adapter needs to be replced there is no need to connect to the component and set a static address.

Here is a link to the switch we use with a description of how it works. http://literature.rockwellautomation.com/idc/groups/literature/documents/pp/enet-pp004_-en-e.pdf

 

[ianingram]

We recently spoke to our local AB guy about problems we had setting up a 1794-AENT module over a "long" distance (about 60 metres)

He asked what type of switch we were using. When we told him that it was an unmanaged switch, he just shook his head and said that for any ethernet remote I/O, you should always use a managed switch (such as the Stratix 6000 mentioned earlier).

After a bit of RTFM, his remark started to make sense....

 

[alwalker]

In effect, setting a switch up like this is still DHCP, you are just making the pool of IP addresses size 1.

 

[The Plc Kid]

In effect, setting a switch up like this is still DHCP, you are just making the pool of IP addresses size 1.

Correct

Plus it is port specific.

 

[kamenges]

Nathan, you're projecting. While your suggestions make sense in a typical PC networking structure that generally won't work in a "normal" plc system. The devices don't have the built-in capability to support the scheme.

We all currently do #1 with the attendant issues with setting IP addresses when components are replaced. #2 isn't an option with most I/O adapters since you can't set a host name. And even if you could, it isn't any easier to set a host name than it is to set an IP address. Most control networks aren't so dynamic that they require shifting IP address ranges. #3 has no advantage when you consider component replacement other than the modifications all occur at a central point. #4, while important, isn't germaine the the conversation.

Given the general state of devices on the market today this method seems to me to be the easiest to recover from when you lose a device.

Keith

 

[surferb]

The approach is unorthodox in terms of networking because it couples your OSI layers 1-3 (physical with IP) without providing any obvious benefit. I didn't say that there's no reason to do it - that's what I was asking. From the responses, I got that if you have an unconfigured spare PLC or Ethernet module that was used to replace a dead unit, you wouldn't have to configure any network settings. That's fair - the technician will still have to download the appropriate PLC program, settings, etc.

In "conventional" networking terms, if your cabinets had multiple devices, it would make sense to me to use a scheme like this xxx.xxx.101.yyy, xxx.xxx.102.yyy, etc. It really depends how many devices you have, how you want to segment traffic, etc. The reason for using the more significant bits to distinguish location is when you need to subnet/combine networks it's possible to do by just changing the subnet mask without changing addresses for your nodes.

If each cabinet does indeed have a single node, then you can use a MAC address to mark that location because there is only one.

Why would it be unorthodox? What if I want cabinet 1 to be xxx.xxx.xxx.101, cabinet 2 to be 102, BCR 1 to be 201, BCR 2 to be 202 ...etc..? I can't really use a MAC address to define a location.

 

[todster]

Nathan, you're projecting. While your suggestions make sense in a typical PC networking structure that generally won't work in a "normal" plc system.

In reality it might not even make sense in a PC network. You could design an entire facility and assign IP's to all the drops and MDF's ahead of time. I can tell you exactly where each and every computer and device is located by its IP. A good example of usefullness, a virus outbreak. Appears to be originating from xxxxxxxxxx MAC which is Ip ..... spreadsheet says that is in the left side of room 101 as you enter the door, the secondary IP for Lucy's desk. Looks like Lucy snuck her laptop in and plugged it into the network. Sure hope Lucy has her resume up to date. Much easier than visiting every computer in a 2 million square foot building to see what it's MAC is. I watched an Army IT expert (he proudly stated he was an expert)trudging thru the snow at -30 going to multiple buildings to locate a strange MAC. Still using soup can and fishline technology.
Like everything, it does have it's down side. Like when they move your office and you have the laptop out at a control cabinet and you want to shuffle a file to your desktop. You're traversing subnets and netbios is disabled so you have to use your IP. Then scratch your head when you can't find it on your desktop, ! Not that it's ever happened to me.....

 

[surferb]

Typically PLCs are configured with static IP addresses. I know. I've worked with some that initially use BOOTP (not even DHCP) for the settings and then remember that address. The best setup often uses a "hybrid" approach.

#2 wasn't for PLCs, but often for workstations (office PCs, HMI stations, etc) that aren't running any server services - they initiate all the talking.

#3 (DHCP with MAC reservation) does have advantages and disadvantages. Besides central administration, it's often simpler to set an IP address by DHCP than connecting to the device directly via serial/etc - that's the point of this whole thread. Disadvantages include that the DHCP server must be up and that it requires a more technical skillset to manage. You can pre-stage MAC addresses of spare hardware.

#4 was to give insight into what factors usually drive DHCP reservations, which is what this topic is about. And security will become increasingly important - just watch.

I do see the value in the simplest possible recover-ability - that's what I hadn't considered, and why I asked the question to begin with. I've taken those 3AM calls from networking-illiterate maintenance guys servicing remote locations and wished that the recovery process were simpler.

Nathan, you're projecting. While your suggestions make sense in a typical PC networking structure that generally won't work in a "normal" plc system. The devices don't have the built-in capability to support the scheme.

We all currently do #1 with the attendant issues with setting IP addresses when components are replaced. #2 isn't an option with most I/O adapters since you can't set a host name. And even if you could, it isn't any easier to set a host name than it is to set an IP address. Most control networks aren't so dynamic that they require shifting IP address ranges. #3 has no advantage when you consider component replacement other than the modifications all occur at a central point. #4, while important, isn't germaine the the conversation.

Given the general state of devices on the market today this method seems to me to be the easiest to recover from when you lose a device.

Keith

 

[surferb]

Point taken...

Point taken...

however,
if you're looking for xxxxx MAC in a Cisco environment, you use the show mac-address-table command. Didn't want to get fired with Lucy? Implement port security, or better yet 802.1X. You could even catch her for trying after her port is automatically shut off. We're getting pretty far off topic now - I've yet to see an industrial facility that really takes security seriously - but I've worked on a lot of networks with workable security settings. It sounds like your Army IT expert didn't know what he was doing.

In reality it might not even make sense in a PC network. You could design an entire facility and assign IP's to all the drops and MDF's ahead of time. I can tell you exactly where each and every computer and device is located by its IP. A good example of usefullness, a virus outbreak. Appears to be originating from xxxxxxxxxx MAC which is Ip ..... spreadsheet says that is in the left side of room 101 as you enter the door, the secondary IP for Lucy's desk. Looks like Lucy snuck her laptop in and plugged it into the network. Sure hope Lucy has her resume up to date. Much easier than visiting every computer in a 2 million square foot building to see what it's MAC is. I watched an Army IT expert (he proudly stated he was an expert)trudging thru the snow at -30 going to multiple buildings to locate a strange MAC. Still using soup can and fishline technology.
Like everything, it does have it's down side. Like when they move your office and you have the laptop out at a control cabinet and you want to shuffle a file to your desktop. You're traversing subnets and netbios is disabled so you have to use your IP. Then scratch your head when you can't find it on your desktop, ! Not that it's ever happened to me.....

 

[The Plc Kid]

And what if the switch goes down and has to be replaced .......

On the stratix switches and most other managed switches remove the CF Flash card and insert into the new swith.

 

[Clay B.]

On the stratix switches and most other managed switches remove the CF Flash card and insert into the new swith.

Hirshman makes a switch with the same features. I like the Stratix on AB systems since it can play on AB's special protocals, but on a non AB system I really like Hirshman and they cost less across the board.

Related to this topic. I use the addressing function of the switch to assign my ethernet address to my flex I/O. makes changing the Anet card almost painless (key word almost).