Moving on to Network Terminology, there's a couple kinds of packet routing. Two primary types of packet routing and one of them is called store and forward switching. It predates computers and it dates back to the days of point-to-point teleprinter equipment when there was a human being that would type a message into a teleprinter equipment and it would be transmitted across some wires to a receiver where it would be printed out and the receiving station, a human being would read it, and see where it was intended to go. Go over to a specific console and type the message and send it on its way. I think I have a picture of the old teleprinter machines here, so typewriter. The message was sent from point-to-point and human beings were in the middle doing the forwarding. So, the message would be transmitted in whole and then get sent to the next point-to-point. In some cases, it was stored on punch paper tape. I remember one of one first jobs first had I worked in a machine company and our CNC controlled blades and mills and so forth and the programmer wrote the program for the CNC machine would punch the program onto these paper tapes and or nylon tapes over about a three centimeters in width. He would punch that out on a machine in the Engineering office and then he would carry it down to the floor and feed it into the actual milling machine and then the milling machine would start up that job and melt the material. Like I said, so humans had read the address and then forward it onto the recipient. In modern use, it means an entire packet is received and reliable transport TCP/IP and so forth have a notion of APNIC protocol and the packet is covered with a cyclic redundancy check. So, the entire packet is received. Generally, it's hardware that checks that CRC says this is good. I've got this whole packet, it's good to go hardware looks at the header information has the destination address and decides which outbound switch port it needs to send it to. This can be used in delay-tolerant applications or where intermittent communication is acceptable. My opinion it's not terribly applicable in real-time systems and there's a better way from a latency standpoint. That it's called Cut Through Switching. Cut Through Switching, the beginning of the packet is being received and the hardware looks at the address and looks up in a table the routing information and decides which outbound port it needs to be routed to and starts the transmission out that port before the CRC value is received at the end of the packet. It achieves a much lower latency. So, if the CRC check at the end fails, there has to be a mechanism and all these protocols have some mechanism of poisoning or marking. There's different terminology that's used, but the last handful of symbols at the end of the packet has to be marked bad in some way so that the receiver at the other end of it even though they started to receive it, realizes that when that end set of symbols gets there that the packet was an error and they may generate a knack back to the sender and the sender can then can retransmit it. But it a big boon, a big help a biggest system and reducing overall system latency. Such applicable to real time systems where you're concerned about latency. API is a term that means Deep Packet Inspection. I'm not an expert on this but I know I have at least enough understanding to grasp that it's not a cursory look there's dedicated pieces of hardware that can perform deep packet inspection, as we'll see in some of the upcoming short videos here. That means taking a look inside what's inside the packets. TCP/IP network layer model has a series of layers. Layer 1 is referred to as a Network Access layer, and this is the slayer defines how data's physically sent. Layer 2 has the Internet layer and it packs data into packets known as datagrams. Contains the source and the destination IP addresses. This layer and the protocol is responsible for doing the routing, making sure that the packet gets to its destination. The transport layer enables devices at the source and the destination to carry on a conversation and then defines a level of service and connection status. Are we talking are we not talking. You can get into quality of service means and all that but that's I think that's a little beyond where I want to go. Layer 4, application layer. This provides the APIs and the protocols application programs are going to write an application programmer wants to use TCP/IP. Set of APIs for building packets and sending them out Internet connection or some outbound port. MIMO, Multiple Input Multiple Output. M2M, machine to machine. In this class, I think I alluded to in the first class. When we see that in this class, this is non TCP/IP communication. So, it's some other protocol but not TCP/IP. An example in there Bluetooth. IoT or IIoT, this uses TCP/IP based communication. How many are familiar with the OSI model layer of communications? You probably at least if you've had a networking course you've seen this before, so this is old news it has been around for a long long time and many protocols have been built that will pick or combine layers from the OSI model, depending on what the goals of the particular protocol, the application or the solution group is interested in solving. So, in TCP/IP they had the Network Access layer, and then the Internet layer, and Transport layer, and the Application layer. So, you can see how these layers that I listed previously mapped to the OSI layer model just for reference. Network Topologies. There's a physical topology and that refers to the cabling, or wireless connections, the node locations, the interconnections between nodes. The physical topology is driven by the level of control or fault tolerance and costs that are required. Logical topology refers to the way the data is passed through a network from one node to another. Got some examples from our good friend Wikipedia here is a star topology. No more explanation is needed there. Rings are very popular. When I first started working in chip design back in the 80s, we use mentor workstation and they ran on was called token ring. So, this was a notion of an outbound port on the workstation and an inbound port to a packet would come into a workstation in hardware and software, and there would look at the packet and see if the packet was for this, at its destination yet and if not it would send it on its way. Then there's mash or partial meshes, full-mesh networks and we'll see some of that coming up. This is becoming very popular in the industrial space because of its inherent fault tolerance, we will see more on that coming up. Trimble engineering is also doing some cool things with mesh networking out in the field. Token ring is physically a star topology but operates as a ring. Most of our homes and Wi-Fi networks are physically and logically a star right. We've got a router and then we've got all our stuff up and think I listed them here. So, I've got a Hub/Router/Gateway. Cell phones are hooked up, our computers are hooked up, all kinds of devices are hooked up. Refrigerators hooked up, microwave is hooked up, Xbox is hooked up and all these devices hooked up in a star topology. This next drawing is important. You'll see it on an exam, be able to draw this picture. I've had a number of meetings with engineers arrow that deploy Industrial Internet of things and I went out for lunch one day one I said, "What do the students need to come away with in terms of what the overall hardware structure of IIoT system looks like any without destroying? And say, "Here, your students need to learn this." So I recreated it here. So, we have a bunch of sensors and actuators communicating by some protocol Machine to Machine on TCP/IP, and then go to a gateway so think of this as potentially a single board. Computer that communicates with these sensors and actuators wirelessly okay. So, there will be a CPU, there'd be memory and it could be a whole bunch of these in the case of the power plant we saw that there were 15,000 just in unit three at that power plant over 15,000 sensors. There's a number of gateways that enable these sensors to connect towards a data center. At this point, the communication switches from machine to machine and becomes TCP/IP. So, we're talking IoT at this point here or IIoT. This Gateways can talk to other gateways. This might be a wired connection or a wireless connection, I showed it as a wireless connection. It's expensive to run cables everywhere in an deployment when you can communicate wirelessly you're going to save cost. Cost is important for the bottom line of the business. These are rapidly becoming wireless communication. This could be wireless as well eventually these gateways. One or more of these gateways ties into the in-house network, either wirelessly or via a hard Internet cable. The in-house network provides access to storage, and servers, and daily driver computers of the employees at the business. That then think about back to the picture of the platform. So, part of the platform view the software view is here as wirelessly communicating to cloud and fog services where additional servers and storage remain available. The last piece is cell service providers allow us with our cell phones to communicate to a cloud service providers to check up on what's going on, look at data analytics runs. Also communicate with the factory or the physical plant, cyber-physical, system find out what's going on there. So, we've got this sort of three-way communication thing taking place. This was the picture that he drew and he wanted you to be able to take away. So, you go to a job interview, important to understand these terminologies and this topology.
