Network Basics EP 1: From the scratch

Hello there,

Shall I introduce you to basic networking terminology in an "immersive" way while more real things are discussed, or should you be provided with a list of terms you have to get familiar with as a homework, if you really want to gain knowledge in networking field? I know that you would like to experience with the hardware as soon as possible, so I don't want you to get bored too quickly. Because of that I also won't duplicate the content which you have plenty of on the Internet, so you won't find here full-blown descriptions of ISO OSI model or anything else, I will just briefly point them out and describe them so you can get familiar and google for more knowledge if you are interested. Some of the terms however might be difficult for me to introduce inline, so at the very end of some Network Basics posts you will find "things to google" section for further reading. Alright, after kinda long introduction, it's time for takeoff.

Generally you can learn networking basics (and even advanced stuff) in three ways:

• academic approach, where you just read and read to understand things over time,
• full-blown hands-on approach, where you build your hardware lab and do things based on vendor docs, networking guides and some deployment scenarios,
• virtual hands-on approach, where you play with virtualized switching/routing appliances

The first approach can give you a very high level of knowledge, but when it comes to touch the hardware for the very first time, you will most probably not feel comfortable. It also takes a lot of time to remember things when you just read, while discussion and experience lets you gain and keep knowledge in the most efficient way.

The second approach is what every boy likes - remember when you were a kid and had fun with puzzles, LEGO bricks, electronics or map editors for your favourite PC games? That's it - building something with a purpose behind, guided by just few rules or no rules at all, gives a lot of satisfaction when you see that it works, just like that dumb little LED lit up with a lemon. Switches however, might be really expensive if you expect high grade quality or if you want to get prepared to your new job where you will work with Extremes, HPEs, Ciscos, and other top vendors. You can buy ten years old switches on Ebay for a fraction of modern prices, but have in mind that such old devices are out of full support since long time and you might encounter some operating issues with complicated or sometimes even simple feature configurations. Vendors doesn't necessarily test all possible config variations as it would take centuries to be sure no bugs are there - while competition never sleeps!
The third approach is something I recommend you for the very beginning. You can get a virtual machine image of a switch firmware, sometimes even for free as with Extreme, and you can deploy those VMs in your virtualized lab that might be built upon popular hypervisor servers (e.g. free of charge VMware ESXi or Hyper-V) or with some network emulator, where free-of-charge GNS3 is the most common example. Things run pretty well, but have in mind that some features will not be working in such environment. Hardware-based features might be limited or not available at all, or you might have only control plane with some switches (that means you will be able to get some protocol up and running but you will not be able to test it by doing ping between connected PCs - or I should say virtual PCs). More than that, most probably you will be limited to something like 4 or 8 ports within such virtual machines, don't expect 24 or 48 which is the most typical port count with physical devices.

Well... not to wait too much, if you work for some reseller and have Extreme stuff right within sight, you're good to go. If not, I encourage you to at least try to create your your lab with EXOS VMs on top of ESXi or GNS3. EXOS VM site consits of relevant manuals but if you need any assistance, let me know!

While the stuff is being downloaded, you might wish to handle the very first words of network communications language (this is what you could get bored with if you dared to study it on university, while now perhaps you'll find this interesting). You can click on the links to get to Wiki articles or other pages and drill down right away:

• OSI model - it's a model that describes layers of systems (devices in a network) that are connected and would have to interoperate for some reason;
• there are seven layers, from which first four allow us to describe things pretty nice, then whatever, then Layer 7 which is called Application Layer (so the ultimate payload of network traffic, like your e-mails or zip file full of memes being sent to your Dropbox folder in the cloud);
• every layer protocol data unit (PDU) is built mainly from layer header and a payload, which is in fact higher layer PDU; so for instance, the Ethernet frame encapsulates the IP packet, which encapsulates UDP datagram;
• example of L1 device is a hub (obsolete), L2 - switch, L3 - router (or 'L3 switch'), there are also content-switches or load balancers which I won't to cover here;
• today however, this layer-to-device bindings are not much accurate, as modern switches or routers can often make decisions based on another layer header or even based on the ultimate payload;
• TCP/IP model - that one is more relevant according to what has been implemented in real world compared to OSI model; OSI model, however, is still used for theoretical explanations quite often;
• One of three types of transmission media used right now: copper cables (twisted pairs) that simply transmit electric signals between connected devices, the signal is then processed by internal hardware;
• they come with many different flavors (like UTP, STP, S/FTP and so on) that enabled the industry for different categories of those patchcords and thus different signal transmission capabilities (for example Cat 6 is the lowest possible category of a twisted pair to run 10 Gbps conversation between devices);
• when copper cables are mentioned, network devices are often mentioned to be equipped with "RJ45" ports to plug the cables in (8P8C in fact, but... treat it just as a fun fact, calling it RJ45 is not wrong because everybody understands that term and nobody cares to fix the world at this point);
• twisted pairs enable you for 10, 100 (for rather old devices, often for out-of-band management interfaces as well), 1000 Mbps or 10 and 40 Gbps connectivity so far; those are often referred as 1000BASE-T, 10GBASE-T and so on.
• Second type of transmission media: optical cables (fibers) that are made of freaking glass - thin as your hair!
• Transmission is based on laser diodes or lasers that emit light (extremely-high-frequency electromagnetic waves in fact) within one of the best suited optical frequency bands (based on the fiber characteristics), this gives really big theoretical and practical throughput possible that can go in thousands of gigabytes per second in a single fiber (quite recent world record was over 150 Tbps for a distance of over 1000 kilometers); it can also provide greater distance for point-to-point connections compared to copper cables (tens of kilometers vs up to 100 meters with modern transmission rates), but for short distance (tens of meters) and 1/10 Gigabit speeds copper patchcord is way cheaper, perhaps you would decide on the medium also based on delay requirements in your network;
• there are single-mode (SM) or multi-mode (MM) fibers, where SM is better for long distances; also Photonic-Crystal Fibers (PCF) exist, but I'm not sure if it's not too early to seek for those in enteprise networking industry;
• in common networks today we can utilize optical fibers for 1/10/40/100 Gbps connections (more complex naming appears, like 10GBASE-SR, 10GBASE-LR, 1000BASE-SX, 1000BASE-LX, 100GBASE-SR4 etc.), but keep in mind that the possible distances without any active amplifiers (like EDFA) drop down drastically with the speed - the higher the frequency, the smaller the energy, so it's more and more prone to the dispersion and attenuation, I think this is why it happens so;
• some trick sometimes used is to multiply the fibers within a single patchcord to get higher throughput by multiplying the amount of data that travels accross links at the same time; for example, 40GBASE-SR4 is 4 times 10 Gig multi-mode fiber for each direction, and 100GBASE is about having 10 times 10 Gig or 4 times 25 Gig; but... wait a moment... the industry already works heavily on 200 and 400 Gbps connectivity!
• with optical fibres there was much hassle with connectors historically, there is no RJ45, but most popular LC, MPO, SC, FC and other;
• but to make life harder, most of network devices enabled for optical connection doesn't have any interface ready to plug a patchord in, they have holes that can also be called ports or slots which can be used depending on your needs; those ports are designed to support transceivers (also called optical modules), i.e. small devices that convert electrical signal to optical one and the opposite; nowadays there are SFP, SFP+, QSFP+, QSFP28 modules available and most widely used, you might also find QSFP, SFP28, BiDi modules or mini-GBIC or XENPAK or few others;
• optical patchord consists normally of two fibers (each for one direction) but some vendors can provide you with BiDi modules where just single fiber is used, with different light wave frequencies for transmitting and receiving the traffic (so take care to use the same Tx on one side as Rx on another side);
• with QSFP+ and QSFP28 you can use so-called breakout cables that allow you to utilize single 40 or 100 Gbps port as four separate 10 or 25 Gbps ports (as long as the device allow for such interface split); you can find breakout cables that are ended with transceivers on both sides, or such with a transceiver on one side and LC (or other) connectors on other end(s);
• most probably outside enterprise networks (carrier backbones, transport networks) you might get in touch with WDM (CWDM, DWDM etc.) transmission;
• Third type of transmission media: radio frequencies in 2.4 GHz or 5 GHz bands used for wireless (WiFi) transmission described with IEEE 802.11 standard and some ammendments; I don't plan to expand this topic here right now as it deserves a separate article; if you wish to gain more knowledge on this, you might enjoy all the knowledge that Devin Akin shares with the world, in fact he is the wireless;
• With wired transmission you might also be interested in 2.5/5/25/50 Gbps transmission rates or with Fiber Channel that is mainly used for networks inside data center (some say it will be replaced, as everything, with Ethernet - thanks to iSCSI, but I'm not much into details to make any judgements here);
• Connections on close distances can also be made based on direct attach cables (I call them DAC in every case); they can be copper-based (DAC - passive, Active Copper Cable - ACC - active) or fiber-based (AOC - Active Optical Cable), in both cases ended with transceiver or transceiver-like plug. AOCs provide greater distances (like between rack) but still, it's just a fixed cable.
• Besides interconnecting your network nodes, servers, PCs and all the rest, it is also good to know about serial transmission console cables (DB9/COM/RS232C) and that there are straight-through and crossover UTPs (so you should take care what kind of cable a particular vendor uses and what RJ45->DB9 adapter should be used to be able to talk with a switch console port - but don't worry, typically switches are equipped with such cables/adapters out of the box);
• When you are building your enterprise network you place devices somewhere (typically in a rack, where each device takes some height referred as 1U, 2U etc.) and you connect the devices with structured cabling that has already been spanned across the building; the connections you use to create your network has a characteristics which is called topology; it can be bus topology, star topology, double star topology, ring, hierarchical, mesh (partial or full), leaf-spine and many others. They result in different traffic handling and different traffic patterns will benefit from this or that topology. In terms of interconnection you might also want to provide more links than needed to have every node reachable physically from any other node (like in mesh or ring or leaf-spine or else) for higher capacity or redundancy in case of failover; but redundant links create loops, and loops have to be handled properly so there won't be any issues with loop-caused congestions (we will discuss this next time);
• For redundancy also PSU (Power Supply Unit) might be doubled inside a network device, this is often called an RPS (Redundant Power System); same with fan modules that cool the chipsets for optimum operating temperature (airflow can be front-to-back, back-to-front or side-to-side, if any - some small bureau switches can be fanless); both power supplies and fan modules are hot-swappable in higher grade devices; sometimes UPS boxes are put into rack for keeping things operational when electricity in whole building goes down, but it doesn't operate indefinitely;

Phew... that escalated somehow but I hope I didn't miss things that might be useful in the future.

I would like to treat the entire "Network Basics" series as evolving material that get's better over time so all comments are appreciated. If you would like to point to topics that I should cover, also feel free to put them in comments. Thanks!

P.S. This "Network Basics" series is dedicated to friend of mine, who wanted to learn something new and change his career path.