Thanks everyone for your responses.
I'll look deeper into the managed switch chips and try to educate myself some more on IP and Ethernet.
Maybe you shouldn't be using an STM32F4, rather get an more powerful* linux-capable system where you can easily use multiple ethernet/cellular connections with OTA updates of the essential libs and functional remote management capabilities.
On the other hand, how big is the team you're working with?
We have given some thought to using a more powerful Linux capable system. We only have a team of two. Myself and my colleague who has only really done some windows PC type programming (self-taught).
We currently have a system that uses a 8 bit Atmel microcontroller which communicates with the traffic management centre over a serial connection. I wrote the software for this system and it's been working well now for some time.
Now we need to add the ethernet connectivity. I've not done any programming with an operating system before so I thought the easiest step up for me was to progress to the more powerful STM32F4 without using any operating system. I've basically got a lot of the screen driving capability ported over to the STM32F4. A lot of the reason for taking this step first was to get myself familiar with the STM32 ARM chip. We need to keep the serial connectivity so I wanted to port that across as well. Then I wanted to get the IP side up and running and was hoping to use Lwip to do that. Maybe I'm totally crazy to attempt that without an operating system but it just seems the less steep learning curve for me.
My offsider does not have any embedded programming or linux experience so I'm a bit concerned we might hit a dead end if we choose to go with an embedded linux system instead. If we go that path there wouldn't be much I could do to help him. Maybe we will end up working separately on both solutions in parallel.
Any suggestions much appreciated.