In article one we discussed a variety of different remote contribution models, as well as examining some of the different backhaul options. But the network design of hardware devices such as hubs, switches and routers makes a huge difference to network efficiency. With a variety of topologies capable of moving data around the network, understanding network theory is key as every iteration will affect how efficient and robust the network is, as well as how much latency it introduces.

These are three main devices used for building your networks. Knowing how they work is key to delivering the best performance as each serves a different purpose:

• Hubs are for connecting small groups of adjacent devices.
• Switches are for connecting up to several hundred devices together.
• Routers are for connecting sub-networks together.

Hubs - How Do They Work?

A hub is the simplest kind of device for connecting multiple nearby Ethernet devices together. It is useful in a Local Area Network because there is no configuration necessary other than connecting the nodes with twisted pair cables.

The major downside of a hub is that all of the traffic appears on every wire.

Network capacity is squandered by delivering traffic to devices that won’t use it. This is not ideal for connecting large numbers of devices.

A hub is most useful for connecting several adjacent devices together. For example a couple of personal workstation computers and a printer. The hub can then be connected to the rest of the LAN using only a single cable instead of one for each device.

Switches - How Do They Work?

Network switches evolved as a more efficient replacement for Hubs. They connect two Ethernet ports directly together. The traffic between those nodes is not visible to the other nodes on that switch. These are well suited to larger networks of IP devices; Several hundred can be connected on a large-scale sub-net. Small and medium sized offices can inexpensively deploy a 32 channel 10G switch that is stackable when expansion is needed.

Switches can be equipped with additional features to manage the traffic. This can prevent individual ports from connecting to others. Meanwhile, connections can be inhibited at certain times of the day. They can also detect when hubs are attached to a port as expanders. Local systems administration policy rules can allow or deny such connections.

Unmanaged switches can be installed as direct replacements for hubs. They make more efficient use of network capacity and work without needing any configuration.

Routers - How Do They Work?

Routers are quite different to hubs and switches. They bridge between a Local Area Network (LAN) and a Wide Area Network (WAN). Typically a boundary router connects your private network to the Internet via an Internet Service Provider (ISP). Routers are cascaded to create nested sub-networks. Firewalls can be implemented at this point.

On the LAN side of the connection, an IP address space is created independently of the network on the WAN side. The Router uses a Network Address Translation (NAT) protocol to map the LAN IP addresses to the WAN IP addresses.

A netmask value determines the class of addressing used for the LAN sub-net. This determines how many devices can be configured. The leading bit values of the netmask are set to 1 to indicate the network ID. This defines the base IP address and is the same for all nodes.

There are a lot of other possible netmask values for special uses and setting it up is a complex topic in its own right.

There are some special IP address values reserved for admin and management purposes. From the LAN side, the Router has its own reserved unique IP address. Recall from earlier discussions that an IPv4 address is composed of four octets (groups of 8-bits). In a class C address space, a router IP address is typically set to:

Routers often support a web-based administrator user interface to edit their configuration settings which can be accessed with a web browser.

A router can also have a Domain Name Service (DNS) resolver defined for converting human readable domain names to IP addresses. If the DNS server is defined in the router, then it can act on behalf of the rest of the network nodes and cache the lookups to improve performance. Alternatively, the DNS configuration can be defined locally for each member node if necessary.

The IP configuration of nodes can be defined statically or they can inherit it from the router. A router would set up a pool of available IP addresses and vend them with the Dynamic Host Configuration Protocol (DHCP). Because there are a limited number of addresses set aside like this, eventually they run out and new devices cannot then connect in this way until one of the addresses is relinquished by another node. A static configuration defines the IP address, netmask, gateway (router) address and DNS servers manually. This process permanently reserves the IP address solely for that device.

Users should be wary not to use the same IP address on more than one device as it can cause instability across the entire sub-net.

Routers sometimes have an embedded hub on their Local Area Network (LAN) side. This would allow several switches to be connected in parallel.

Diagnosing router problems is facilitated by the ping and traceroute tools. The traceroute tool echoes a line of feedback for each router it encounters between the client node and the target being tested. Internal router security countermeasures at any point may prevent traceroute from completing the round-trip. Each router it passes through is described as a hop. A higher number of hops indicates a longer virtual distance in Internet terms. It is useful to remember that Internet geography is not like the real physical world:

More hops = higher latency

Examining the output of the whois and nslookup tools helps to diagnose DNS problems.

Network Topologies

Choosing the optimum network architecture in the remote site and production data center is important to maximize the performance of all remote contributions.

Here is a list of some alternative topologies:

• Point-to-point.
• High capacity multi-link point-to-point.
• Bus arrangement.
• Daisy chain.
• Ring.
• Star.
• Tree.
• Mesh.
• Partial mesh.
• Hybrid.

Point-to-point Topology

This is the simplest topology. It describes high-capacity connectivity between two endpoints. Connecting to a shared cloud solution via a high bandwidth point-to-point connection would bring a remote outside broadcast unit very close to the production data center in virtualized networking terms.

Backhauling describes a point-to-point network topology.

These are the principal benefits:

• Most efficient bandwidth allocation.
• Only two connected endpoints.
• No routing needed.
• It just requires a backhaul connection.
• Very low latency.
• Scalable to whatever is needed.

High-capacity Multi-link Point-to-point Topology

The risk associated with running a single point-to-point link is that of dropouts and disconnection. That risk can be managed by multiplexing across several physical links. This increases the available capacity but some complexity is introduced at the receiving end to rearrange packets that might arrive by different routes and at different times.

The benefits are:

• Higher capacity.
• Redundancy is improved, making it more reliable.
• Less risk of signal loss.

Bus Arrangement Topology

Here, data is transmitted to all nodes connected to the bus. This is not an optimal solution for backhauling but it is useful at each end of the link. Workstations can communicate with one another in the OB pool and at the production data center.

A bus is implemented using a hub.  A switch is preferable to a hub as it eliminates redundant traffic that reduces the net available capacity. A benefit is that adding new devices is very easy.

Daisy Chain Topology

This arrangement sends the data to the first node in the chain. That then passes the traffic on to the next. This topology works well for scenarios where all the devices need to see all of the data.

MADI and Dante protocol audio streams routed to personal monitor controllers use a daisy chain technique. This can distribute up to 40 channels of audio so that each musician can select and mix their fold-back individually. All of the audio packets are needed at every node.

The downside is that if a single node fails, then all the downstream nodes beyond it lose their data feed as well.

Ring Topology

This is a special case of the daisy chain topology with the ends joined together. It requires high-speed network connectivity to perform well. Arguably this is somewhat inefficient and latency is compromised by the packets potentially needing to travel round the entire ring to reach the furthest node. Historically this has been used in audio studios with a token ring protocol. This would not be optimal for distributing video.

Star Topology

This is a logical view of how switched Ethernet networks communicate.

A hub might still be useful at one of the endpoints of the star to aggregate several adjacent devices without needing additional wiring.

Tree Topology

The central WAN (Wide Area Network) bus connects multiple routers together. Each router manages a sub-net on its LAN (Local Area Network) side. The router can pass traffic from nodes on its LAN side to other routers on its WAN side.

Domestic broadband connections are managed like this. Within a large enterprise, different sub-nets can separate business functions such as Manufacturing, R&D and Finance. Switches would be used on the LAN side of each sub-net although they are not shown here:

Mesh Topology

A mesh topology is useful where extremely high-reliability connections are required, and these are typically in high-end broadcast or military installations. Every node is connected to all of the others. Wiring this is complex because each node requires multiple Ethernet connections to reach the other nodes. Maintenance is also difficult and adding new nodes is cumbersome.

In addition, physically wiring a mesh is expensive and complex. It is being replaced by modern wireless mesh solutions.

Partial Mesh Topology

A partial mesh might be a solution to simplifying the wiring of a mesh at the expense of direct connectivity between all of the nodes.

Every node is reachable but some are indirectly connected and messages need to pass through other nodes to reach the intended destination. Clearly, that will introduce latency.

Hybrid Approaches To Network Topology

Combining the individual topologies in a hybrid arrangement can deliver the benefits of each where they are needed, and can avoid their disadvantages by using a different technique to join them together.

For example, this is how we would connect a tree topology at the OB site to another tree topology at the production center by using a point-to-point backhaul link between them.

Choosing the correct local area network topology at each end of the link can take advantage of that point-to-point high capacity and distribute the connectivity to the individual devices used by the OB crew. A hierarchical topology offers the best attributes of both solutions.