Continuing my series as I work through the CCNA syllabus. The introduction to the series can be found here.
I will be pretty much following the CCNA Composite Exam Blueprint point for point. One post per bullet point. I’m using Version 11 (640-802).
Selecting Components to meet a specification
The bullet reads: Select the components required to meet a network specification
Again, this is pretty vague, and could easily encompass the whole of a CCIE/Network Engineer’s role! So, we’ll fall back to the general CCNA outline again. We are dealing with small office/Branch office networks of up to 100 devices.
This could quite easily be the shortest blog post in the series. “Well I know what the devices are from the first post, so now I just need to pick a combination for a given number of users.” Cool, done, off we go…
The devices we pick and choose for any particular role depend on a number of things. Particularly in a small network we have many options that will equally suffice. How do we pick between the options? In a large network, this actually becomes a little easier. For sanity if nothing else a large network has to be kept simple, and modular. This in itself means we just have to pick the correct module once, and we are done. In the small network, we have more ways we can skin the cat, whilst still being successful.
The Cisco product line also has a small bearing. Knowing what different varieties of switch, router and access point can accomplish is quite useful.
A good example of this is the smallest of small networks. Take a small office with two staff, and a shared printer. No matter what we will need a router, to access the Internet. We could also specify a switch or hub to connect the PCs and the printer, and an access point in case any users need or want a mobile device. We could include a firewall for security. Now we have four networking devices, twice as many as we have staff! So let’s look again.
Every router Cisco produce either comes with, or can have added to it, a 4 port switch module. This means we have less devices to worry about, and less connections to fail. Cisco routers are also capable of being firewall devices utilising access lists etc. So for such a small network a separate firewall is unlikely to be necessary. Finally, we have not been asked to provide wireless, so we should not include it. Even if that were part of the specification, we could again get the functionality from the correct router. We’ve reduced the number of devices down to one from four. The number of interconnects also reduces substantially, and finally, all of the interconnects are of the same type. This is the trick of this bullet. It isn’t just picking kit that will do the job. It is doing so intelligently and efficiently.
Just a minute, you are thinking, why is he talking about types of interconnect? Surely everyone uses MIDX Cat6e now adays? Well, no. The final part of this bullet encompasses the role of straight, crossover, serial and rollover cables. Perhaps I should have mentioned these in the first post
A brief interlude – Interconnects
For a CCNA we are primarily concerned with 4 types of interconnect. Straight and Crossover Cat5(6,5e,6e) cables, straight serial cables, and “rollover” cables.
The rollover cable is effectively an RJ45 (the terminator that Cat5(6,5e,6e) cables use) to 9 pin serial connector cable. This also slightly alters out the pinning of the serial cable. This connection is used to connect to routers via a serial cable in order to control them, but not for data transmission. Why a plain old serial cable is not sensible, I have no idea.
A serial cable is used for “high speed” (ha, yeah) data transmission primarily between routers. In the real world, I have only seen this on the ISP side equipment, and even then, more and more of this is coming to use Ethernet. As far as we are concerned, serial connections are for router to router connections.
A straight Cat 5 cable is used to connect devices of differing type. PC to switch, PC to router, Router to switch. It is not used to connect switch to switch (Except with the advent of the GBit spec, the ports are now designed to negotiate a data pin if a straight cable is used, so cross over cables aren’t really useful any more). Cat5e, Cat6, and Cat 6e are simply further advances to the specification designed to carry more data than the original Cat 5, but that doesn’t make a great deal of difference at CCNA level, and certainly not with less than GBit interfaces.
A cross over cable is there to connect like devices. Switch to switch or router to router. In CCNA router to router connections are almost always serial. A cross over cable is exactly the same as a straight cat 5, except the data and ground cables are “crossed” meaning that data at one end goes to ground at the over and vice versa. This makes little to no sense until you realise that in cat 5, there are two data pins. One is used to transmit and the other to receive. This allows the cable to be full duplex, i.e. capable of sending and receiving at the same time, without collisions. Obviously, if both listening pins are connected, and both transmit pins are connected, nothing will hear anything, and there will be lots of collisions! The hardware in PCs and routers is pinned differently to that in switches. The short version is, if you are connecting switches to switches, you want a crossover cable.
Back to business
So, lets end with a less contrived example. A small office with 30 staff. They have two servers and three divisions: Accounts, Engineering and QA. Each division shares a room, and has 10 staff.
Here we go with a router again. This is necessary to connect to the Internet. We could also utilise any switch ports in the router to connect to the servers, or we could have them on their own switch. We will come back to that decision. Each division gets it’s own switch. Most traffic will be within a division, or between the division switch and the router (be that to the Internet or the servers). The switches will be connected to the router, as we will see later the different divisions will be on different subnets, and so a router will be needed. We always want to ensure less hops (jumps from one device to the next) in a network, so we connect the switches to the router, rather than chaining them. This also has the advantage that it spreads the traffic over more connections, and cause less trouble if one connection fails.
Now back to the servers. We have three switches already. So if we have only four ports on the router, we must put another switch in for the servers. In most cases though, we could utilise an 8 port switch in the router, or even use a layer 3 switch. This will let us directly connect the servers to the routing device. This has the advantage that most traffic will be from clients to the servers, so reducing the hop count, or “circumference” of the network.
A final Note
Finally, hubs and switches. As I alluded to in the first post. We basically never use hubs now. A switch is *far* more efficient, with the reduced collision domains. *Never* include a hub in a design.