Energy Efficiency Made Simple Vol IV 2015

country-wideWAN (Wide Area Network) and then split this larger range into 10 subnetworks, one of which will be used for countrywide admin, with the nine remaining to be split amongst the nine provinces. A differ- ent group of engineers is in charge of each of the provincial networks. As long as each engineer does not use IP addresses from outside of his/her allocated range there will not be an overlap, so the different groups do not have to consult constantly with one another regarding the IP design. (Note that this does not take into account the company standard etc for the network design, but does give a basic example of how IP subnetting works). A single provincial network’s IP range may then be split into smaller subnets for different processes (VoIP, control, CCTV, monitoring etc.). This would generally be an acceptable level of subnetting to provide a stable and clean network. Splitting up these subnetworks much further may start to cause problems as more and more data has to be routed between the different subnets. Also it is important to note that each time a network is subnetted we lose two usable addresses (due to the way Ethernet works these addresses become special addresses for the subnet that cannot be assigned to actual end devices), another reason not to overly subnet the network, as eventually one may run out of addresses to be used for end devices. Once these major points have been covered, we should have a good foundation for planning other components and functionality on the network. We have an idea of the physical and logical topologies for the network, as well as the addressing (IP ranges) and VLANs that must be implemented. This means that we can start to plan some of the other parts of the network, such as the specific redundancy and various traffic control aspects. In some cases these may simply be an added functionality or a way of easily monitoring the network, in other cases they may be critical to the correct running of the network. Which parts should be considered critical and which are simply ‘added extras’ will depend on the network itself and the requirements thereof. Redundancy One of the most critical functionalities to be considered on any mission critical network is redundancy, which we have mentioned in passing in the course of this article multiple times. However, redundancy is important enough that we will go over it in more detail here. Redundan- cy means that the network caters for certain failures, such as cables breaks or hardware failure. In the event of a failure, the redundancy mechanisms in place will attempt to ‘recover’ the network to a point that communications are not interrupted, thus allowing the failure to be addressed without the need for the entire network to be brought to a standstill. These recoveries need to happen extremely quickly, especially in the context of communication networks for SMART grids. A loss of communication between sites, even for a second or two, can lead to devices automatically shutting off the flow of electricity on portions of the grid as a matter of safety. Cable redundancy There are many different types of redundancy; however, generally when people speak about redundancy relating to a communications network (without explicitly stating the type of redundancy) they are referring to cable redundancy. Before we look at the different cable

PC A

Switch A

Switch B

1

2

2

4

1

1

2

Switch D

Switch C

2

PC B

1

Figure 1: Basic ring topology without redundancy activated.

redundancy mechanisms available, let us look at why they are needed on an Ethernet network. Consider the basic network (see Figure 1 ) - assume this hypo- thetical network is not running any form of cable redundancy yet. If PC A needs to establish communications with PC B, A will start by sending a type of message called an ARP request (Address Resolution Protocol). This is an example of a broadcast packet, meaning that rather than a specific device/s as its final destination, this message will be broadcast or sent to every device on the logical network. Note there are other ways of generating broadcasts, as well as multicasts (which are similar to broadcast in regards to the requirement for redundancy). For this example we will just look at the most common broadcast, an ARP request, as we are more concerned that it is a broadcast rather than the actual details of the message being broadcast. PC A will send this broadcast to Switch A, which then propagates the message out of every port it has active except for the port on which it received the broadcast. So, Switch A sends to both Switch B and Switch D. The message received by Switch B is sent on to Switch C (again, it sends out of every port except that on which it received the broadcast). Switch C sends that message to PC B and to Switch D. PC B then responds (directly to PC A, so this message is not a broadcast) with its machine address, and unicast (one-to-one) communications can happen between the two PCs. Let us not forget that in the meantime, Switch C would have sent the broadcast on to Switch D, which then passes it back to Switch A, and the whole cycle begins again. We can see quite clearly, with this physical loop on the network, that our broadcasts will effectively continue circling the network indefinitely. As more and more broadcasts are introduced to the network, this erroneous traffic keeps building up, to a point where the communications links and devices are so busy transmitting this ‘waste’ information that normal and critical data is not able to ‘fit’ on the network. In fact these broadcast storms, as they are known, can be so severe as to hang up end devices such as PLCs and older PCs or servers as all of their processing power is taken up having to inspect the packets generated by the broadcast storm. The basic point described is that: Ethernet does not like loops! The simplest solution is to remove one of the cables, thus breaking the loop and interrupting the broadcast storm. In this case each device on the network only receives the broadcast packet once and then it is done, rather than storming around the network indefinitely.

56

ENERGY EFFICIENCY MADE SIMPLE 2015