Ethernet and TCPIP are already regularly used as internal buses in large
NAS systems. What about taking this model to embedded devices for IoT?
Backplane = Ethernet hub/switch - ala D-link GO-SW-5E (5VDC powered - roughly $15) (or even a Mikrotik 750 router)
CPU with direct Ethernet = SoC ARM or x86 based compute and database engine - Intel based (Galileo,
Minnowboard),
Beagle Bone,
86duino,
Raspberry Pi...
Bus controllers = Lantronix
xPico classic.
The last part is the key piece. It can even in some cases be the CPU too. Platform specification for IoT often gets into nitty gritty I/O capabilities. This sometimes diverges into onboard or outboard coprocessors like TI Sitara's PRU, serially connected Atmel or PIC microcontrollers, special function chips like advanced application specific UARTs or communications units (think oddities like ARCNET, HomePNA, unusual RF, etc.)... and sometimes even to FPGAs. Usually these pathways take one to very specific/special specifications.
Ethernet is a natural actually - having:
- Isolation.
- Variable bandwidth.
- Wide "driver" support.
- TCPIP support - for web GUI and standard protocols/frameworks like XML, REST and layerable security and filtering.
- Various forms of pre-existing transparent, web configurable, "driver-less", pass-through devices.
Ethernet was formerly too expensive. Look at the
price of xPico. And Ethernet suffers from being thought of as a LAN bus (distances of 1-100M). Instead think of it like CAN, I2C or SPI.
Bus power? One can go:
- Power over Ethernet 802.3af on the high end.
- Passive Power over Ethernet (with 3V-12V over short distances).
- 5V - USB being the current best flavor of 5VDC - Common EPS.
Alternatively, think of the xPico as a replacement for a PLX PCI9030 or a PLX PEX8112
or various forms of "bridge as IP cores" to be fabbed on demand. Without the fabbing, minimums, BGA assembly, and so forth.
