Designing Resilient Business Networks

Introduction

Computer networks are often complicated systems used to support a businesses day to day operations. It is the foundation for which businesses both access their internet based WAN connection as well as accessing common cloud computing application, designing a network is not as simple as purchasing the equipment and hooking it up, that is at least part of the process, networks are often designed around a set of fundamental components, that often initially begins with the hardware used. Often when deciding at the overview you determine a networks current demands, scalabiliy, resilency tolerance, and performance requirements, as well as the environment in which to deploy a network.

Power delivery

Power quality and cleanliness can often be an overlooked aspect of designing business networks and the protection of business operational capacity. Power redundancy can be used to protect many aspects of your business and those technologies are uinterruptable power supply units, they come in two form factors, Online and Line-interactive, there is also power conditioners which take in the 230V and clean up the output power on the sinewave via AVRs and other components and they are usually actually called an AVR unit.

• Online - The battery is permanently connected to the load, where the power delivered is scrubbed before it is sent to the device the battery is constantly on charge, so if there is a power outage there is a lack of transfer time, which means there is a much lower chance of a hiccup happening during that transfer effectively being "seamless" in the transfer process.

  
  

• Line-interactive - The battery is always being charged, but there is a bypass circuit intended to provide power, during the detection of the surge the excess energy is dumped into the respective suppression components and battery. Line-interactive UPS systems have a transfer delay that make not prevent a surge from being passed on, but the time for transfer is usually fairly low sometimes on the order of nanoseconds.

  
  

• Power conditioner / AVR - Sometimes a UPS might not be necessary in this case we use something called an AVR or power conditioner what it does is it cleans up the noise on the sinewave so it more closely approximates the sinecurve correct as well as managing the input voltage, so it reduces the effects of brownouts and spikes, this circuitry is often integrated into both Online and Line interactive UPS technology by default.

Physical Medium

Determining the type of cable that should be used is best determined by deciding on the level of resilience and speed the network needs, wired based cables such as those used for Ethernet are typically not as resilient as another medium known as fibre optic, howeer the cheapest for any business is definitely going to be an Ethernet network, when choosing the wired communication medium cost, interference and distance are the key considerations that need to be made when using wired networks At the core of every network the communication medium is what should be examined to determine what a business network requires typically this begins with the physical cabling a network may need. Different types of physical cabling have varying properties that make them suitable for the various tasks they may be used for, the most likely candidate used for building a business network is Ethernet cabling, build typically using CAT cable or Category cable usually the most basic being Cat5, the cabling is also which will be used to support the wireless communication for all devices in a business network.

Ethernet cables have many different variants, those variants are suitable for specific distances, types and their respective applications. Ethernet CAT cables are divided into two structural camps solid-core and stranded, solid core cable is more resilient to damage however if the goal is to bend the cable a lot it is a poor choice, the alternative to use is twisted pair stranded cable,

• Cat5 - Maximum speed 100Mbps you can use this speed class to connect devices that do not need extreme bandwidth

• Cat5e - Is often the most cost effective when it comes to purchasing cables,

• Cat6 - Supports up to 1000Mbps or 1Gbps good option if your workload needs increased resilience from noise

• Cat6a - An extension on existing Cat6 standard

• 
  

The CAT designation affects everything from maximum attainable speed, noise resistance, maximum allowable distance.

Network Interface cards (NICS)

When talking about device nodes we are talking about the endpoints of network connections which are the individual computers and servers. Depending on the node type determines the kind of network interface cards (NIC) clients are often the simplest to consider as the typically have an individual network interface that connects to the network and those network cards often come in 2 types and several subtypes, just like CAT cables are used to determine the speeds which can be achieved at the other end of the cable the network interface card is the modulation device that transmists and receives

• Standard Ethernet adapter Single-port - Using conventional Ethernet CAT cabling to transmit data over the network, comes in 3 speeds and two subtypes. the first being single port NICs.

  • 100Mbps - Depending on the workload costs can be saved by targeting this speed group but usually 100Mbps / 1Gbps network interface cards are both capable of 100Mbps and 1Gbps

    
    

  • 1Gbps - Most common and the cheapest often is used to support network devices which do not require extreme amounts of bandwidth

  
  

  • 2.5Gbps - Is slowly becoming mainstream many motherboards are now integrating 2.5Gbps link ports, in future it's likely this will replace all 1Gbps connection to client nodes.

  
  

  • 10Gbps - Often the fastest type can fit a total of 10Gbps network speed through the cable allowing you to provide a high speed connection to a device.

    
    

• Standard Ethernet adapter Multi-port - Are typically deployed on server systems, the reason being is they will have more than one 1Gbps port or more than one 10Gbps port dual can either bond network connectivity speeds or use the second port in what is known as 'failover configuration' which is as it says when connectivity is lost to one of the ports it quickly switches to the secondary port. Dual port 10Gbps NICs often run quite hot as well, switching 10Gbps on both sockets often generates a lot of heat.

  
  

• Small Formfactor Pluggable SFP Adapters share many traits with the multi-port ethernet network cards, that is they can have multiple ports, but they also have something called an "SFP" port or Small Formfactor Pluggable, these are little modules that go into an SFP port, the ports themselves are in usually three categories.

  
  

  • SFP - Maximum speed 1Gbps but this is not the entire story it also depends on the actual SFP interface, as they can come in multimode or single mode. Multimode fibre links are much cheaper for both the cabling and the actual module. Single mode is good for long distance links but they are most definitely more expensive than standard single mode.

    
    

  • Direct Attach Copper (DAC) - these often have similar traits to SFP modules but instead of being simply a module where you insert your fibre optic connectors it becomes a cable that is hard tethered to the other end of the cable, these typically support SFP+ and SFP.

    
    

  • SFP+ - Maximum speed 10Gbps, often network cards used by servers are often using more than one of these in the actual SFP+ slot, this allows a server to potentially have access to speeds upwards of 20Gbps or higher if the links are put into a linkbond.

    
    

  • QSFP - Often the most expensive out of all the SFP module type adapters and also the most expensive cabling, offers the highest speeds possible and typically requires the appropriate optical signal modulation to achieve these speeds as well as switches with QSFP.

    
    

• Speciality interface cards - This is class of system all it's own but these technologies such as HBAs and Infiniband technologies often are built with even higher speeds and lower latencies than both SFP and Ethernet combined. They are what they call external high speed device interconnects, they allow you to bridge multiple computers into a low latency network which might be sensitive to delays across the entire network, such an example might be scientific computing servers which share and update via MIMD (Multiple instruction multiple Data) distributed architectures, this is what we call the field of high performance device interconnect and it often requires both specialized switches and specialized network cards to operate, before Nvidia bought them one notable company was Mallanox responsible for developing the Infiniband high speed interconnect technology.

  
  
  
  

Wireless Medium

There is often the requirement that individuals in a business are able to move around the premises and retain connectivity but also there is an issue where it may also be necessary to reach far parts of the building where Ethernet connectivity is not a viable option in both of these cases a wireless network should be considered, Wireless networks are a foundational element of any business network and careful consideration of how that network works is often important wireless networks themselves can be deceptively simple, all you have to do is plug in a cable to the port on the AP / device and you have WiFi connectivity, however it's more complicated than that.

WiFi Networks

Are the foundational core of any wireless network when deploying a wireless network it can be as simple as having a single radio which is the entire network, to extremely complicated networks consisting of wired cabling and multiple access points all broadcasting similar networks, in the most simple case a standard router / access point is what can be suitable for a business network but it may not offer increased scalability or improved reliability over time.

• Types of WiFi devices found in homes and businesses

  
  

  • Standard router / switch / access point - combinations all the features needed for the WiFi network are built into device without needing independent devices like routers, switches and dedicated physical access points, in many small business cases

    
    

  • Access Points - Are a device that connects to a controller or managment hub, it allows the deployment of multiple points of connectivity improving the overall coverage of the entire network, as well as improving the speed.

    
    

Antenna types - Antenna choice can play a critical role in determining the effectiveness of a wireless network, antennas are engineered using different designed, but to determine the effectiveness of the antenna for a specific application depends on how the antenna is positioned within the building. Radiative patterns are usually communicated using polar E-plane and H-plane diagrams, this allows you to determine the rough horizontal projection of an antenna or the vertical projection, bare in mind the radiative pattern of an antenna when consiodered together is typically a vector field, which has high intensity at the intentional radiator another name for an antenna, replacing different antennas with higher gain improves the received signal, by increasing the perceived power at the end point of the antenna, an increase in gain increases the power receieved by increasing the focusing of the signal to an endpoint.

• Bipolar - Commonly what is found on general WiFi devices like home based and some smaller, SOHO - Small office Home Officer configurations, they are best positioned in an open area because they radiate in an intensity pattern from one side to the other in sort of a "bulb hourlgass like pattern" the way we describe this radiation pattern is using a

  
  

• Omni-directional - You can find this inside access points mounted to roof of buildings as they are designed to consistently radiate in all directions, they often are used in open spaces and not used in corners or edges of buildings.

  
  

• Unidirectional - Is the type of antenna that is used when projecting over a large area such an example might be a large conference room the antenna being in the top corner of the building and the "radiation pattern" which typically has a stated angle is defined by the shape of the antenna, often parabolic is used to create a unidirectional radiative field.

  
  

• Highly directional - These are usually the basis of high speed microwave links, one particular type is the 60Ghz P2P microwave links that might be used between building or other nodes, highly directional antennas might also be employed atop cell towers when they need to use a radio communication based relay system.

Satellite Uplinks

While not hugely common they have become more prevalent and have many considerations to deal with but the most common technology used is conventional geostationary and phased LOE satellites one over the other depends on the application, but satellites can provide a redundant path for your business in the event of a critical WAN failure such as the loss of connectivity to the NBN network or another hardline based WAN service.

Cellular Extension Networks - These are not typically anything particular fancy but usually repeaters installed ontop of buildings or other communications sites to boost the signal of the received cell signal, to maximise the coverage of cell service in a large building, repeaters will take the signal received from local celltowers and amplify the existing signal to improve coverage.

Switches

Come in two forms unmanaged and managed, for simplicity unmanaged switches make deploying extra-connections needed on a network and they can be used to reduce costs at locations which do not need a "fully managed switch" the difference between managed and unmanaged network switchs are they have the ability to configure every setting on a switch, that includes restricting login controls, to being able to setup link-aggregation, define DNS addresses, as well as many other options. They often form a critical part of the networks redundancy and resillience.

Logical layout

The physical connectivity of a network is not the only consideration for designing a reliable network the logical structure also forms a core foundation of that system, which provides a logical address set of states which is used to support both local communication and remote communication

Layer 2 (Datalink layer) Frames

At the lowest level of the OSI model in the basic logical layout is these structures for addressing individual machines called MAC addresses (Media Access Control addresses) they form the basis of addressing local machines only, consisting of a burned in address of the Vendor ID and the the OUI (Organizational Unique Identifier) all mac addresses are 48bits in length, and they form an important function in the network which is to address devices on a local network that are in a similar broadcast domain, all devices on the same ethernet network that has no actual logical segmentation, switches are designed to support local communication and "frames" are what are forwarded around a local network.

Layer 3 (Network) layer IP packets

Are intended to allow forwarding to remote networks, this is a bit complicated but to explain we start with a basic networ, packets are what is wrapped around the datalink layers frame the purpose is to forward the packet to a remote network, now remote network does not necessarily have to be an actual long distance connection but another network with a different address to the network who is trying to communicate across the network, this could be an access point, a WAN connection which may also be an internet connection or even a VPN connection, the simple way to define a remote network is to define it as a different IP address or IP subnet, those two words are key, IP addresses themselves come in two forms IPv4 commonly what I used currently and IPv6.

Internet Protocol Version 4 - Is still alive a well in many organizations, and has several classifications used to define their purpose some are globally routable, some are not, in the case of not, the non-globally routable addresses often consist of the private and public IP address spaces. The reason for these to exist was a technology introduced to allow the extension of the usable address space with IPv4 as there are many more devices in the IPv4 world than was intentionally created so private and public address spaces had to be created, in order to deal with the dwindlikng supply of IP addresses.

Layer 4 (Transport) Session tracking (Datagrams and Segments)

Scenario case study example: GeoSpatial Engineering

Is a business who does high density geospatial mapping of building and arial capture, this system consists of a small fleet of light aircraft used for mapping surfaces of terrain and existing buildings using LIDAR technology, the company does not use high speed data transfer on the planes as the data is likely to potentially be corrupted during transfer so they use arrays of disks aboard the lightweight aircraft.

When data arrives at the companies HQ there is servers "ingestion servers" they are the systems that take all the data to be filed and processed, the software they use allows for almost realtime playback of this data, and the workstations used by the actual engineers has extremely high speed connectivity into the network, this allows them to perform realtime analysis on this data from a large multiple-array spanning display system to visualize this information the workstation that runs that screen has an incredibly high speed connection a single connection 10Gbps however they are considering the current 10Gbps is not enough, so they are considering switching to high speed SFP based modules and networks. (not finished) the LIDAR visualization platform in this case is likely to benefit from using Infinbind as the bridge technology for the workstation used to present to both commercial and governent customers.

The company also has a WAN connection, this connection currently is a limited 1Gbps in future GeoSpatial Engineering wants to branch out into having more than one branch office and will likely need some kind of connectivity, they can likely keep their existing WAN connection which facilitates their internet and purchase a leased-line using metro-ethernet to bridge a high speed connection to their secondary office allowing for fast replication to the second site.