1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Draft Networking 101 (Sticky Candidate)

Discussion in 'Tips & Resources Collaboration' started by schwefel, Nov 13, 2011.

Thread Status:
Not open for further replies.
  1. schwefel

    schwefel Cool Member

    Jan 3, 2009
    *This primer is written with a focus on the DirecTV integration to a home network, but most it will equally apply to Dish Network and generic MoCA deployments, as well.

    *I am still relatively new to MoCA (and by extension, DECA). As I learn more I will correct/add to the document in those areas.


    A network consists of two general classes of devices, hosts and infrastructure. Hosts are the end devices like a PC, a STB, a video game system and the like. Hosts are the devices that “use” a network by generating and/or receiving data. Infrastructure is the devices that hosts connect to. Yes, this is a bit of an over simplification, but it is not inaccurate. Also, connecting 2 hosts with just a cable and no infrastructure devices is still a network. It may not be hugely useful, but it is still a network.

    I am not going to spend a lot of time on hosts as most everyone knows what they are already or will easily be able to deduce what a host is based on the examples given in the previous paragraph. The focus of this write-up is on the network infrastructure.

    Network infrastructure is everything in a network that is not a host. The cabling, routers, switches, hubs, wireless access points, converts, etc are all parts of the network infrastructure. Understanding what each of this infrastructure is and how it operates will allow you to design (or redesign) your network to meet your needs.

    Important Terms and Concepts

    MAC The MAC address is the Media Access Control address of a networked device. Every network device has a MAC address. It is globally unique and therefore uniquely identifies every device in the world. It is also known as the Layer 2 address. MAC addresses are assigned to a host/device at the time of manufacture. A MAC address is commonly written as 01:23:45:67:89:AB or CDEF.0123.4567.

    Segment – A network segment is where all of the device share the “network connection” and can see each other directly. A mostly accurate analogy would, if computers were people and everyone is in the same room. The room would be the segment.

    IP Address – An Internet Protocol address is how computers around the world uniquely identify each other. IP addresses are hierarchal in nature and are locally administered. An IP address, combined with the subnet mask, defines the local network that a host/device exists on. If the local IP address and remote IP address are on different network, the data must be sent to a router to be sent to the remote network. An IP address is expressed in the form of

    RFC 1918 defines the reserved private address space. These are the only addresses that are officially allowed to be “duplicated.” Theses addresses are defined as private because they are not allowed on the Internet and require the use of NAT or PAT to allow hosts with these IPs to “talk” to the internet. While NAT and PAT are beyond the scope of this document, PAT is basically an IP multiplexing protocol. It allows many private IP address to use one public IP address, assigned from the ISP.

    QoS – Quality of Service is feature that selectively prioritizes/deprioritizes network traffic. The priority adjustment of the traffic is done so that time and/or delay sensitive traffic (voice/video) is sent as quickly as possible and non-delay/time sensitive traffic (web/file downloads) is held up a bit.

    If the utilization of a network segment, shared or dedicated, is too high, the effectiveness of QoS will decline. For most networking technologies this is around 75%. At around 90% utilization, QoS is no longer effective and may actually be detrimental.

    Subnet Mask – The subnet mask is used by a host/device to determine what hosts are part/not part of the local network. The subnet mask looks similar to an IP address, but it is not the same. On a home network the subnet mask will commonly be It is actually a binary mask that is applied to the binary representation of the IP address. Using an IP address of with a subnet mask of will tell the host/device that all IP address starting with 10.1.2 are on the same local network and all other address are on remote networks. (This is an extreme over simplification of subnetting, however a more in depth discussion is not needed for 99% of the home networks out there.)

    PoE – Power over Ethernet is way to supply power to network devices from a switch over standard Ethernet cables. It is very similar to how the DirecTV Power Inserter works. Typically IP phones, wireless access point and IP cameras use PoE. There are also devices called PoE splitters or taps that allow you “peel” off the power from the cable and power injectors or mid-span devices that allow you put PoE on an Ethernet cable individually. Historically PoE has been limited to a max of 15w of -48VDC and as of 2009 up to 25w. Currently, Cisco is selling pre-standard devices that can supply up 50w.

    There are two common types of PoE. The first is Cisco Pre-Standard PoE (not the 50w version above), which only works with Cisco devices. Most all Cisco PoE devices can deliver and/or utilize the Pre-Standard PoE. The second is the official IEEE 802.3af PoE standard. Make sure you know what your devices support

    IP Address Assignment

    All devices connected to a network MUST have an IP address assigned. Without an IP address the device is unable to talk to other devices. IP address assignment may be manually set (static) or automatically assigned (dynamic).

    Static IP address assignment required the administrator (that is you) of the device to manually configure the settings. In a small network this is generally not a problem. Care must be taken to not assign the same address to multiple devices or they will not work. Additionally, if there are changes to the network (changing the IP addressing range, a new ISP, etc) it may require the administrator go to every device and adjust the settings.

    There are a few classes of devices that should have static addresses, though. Printers, wireless access point, network based storage devices, the broadband router and similar should be statically assigned. It is undesirable for the addresses of these devices to change. It is also possible to use the dynamic method to statically assign an address. This is typically referred to as a reservation, but it may vary based on the device that is assigning the addresses.

    Dynamically assigned addresses are the generally preferred method to assign an address to a device. The protocol for this is called Dynamic Host Configuration Protocol (DHCP). Virtually every broadband router on the market has a DHCP server built in to it. The DHCP server keeps track of what addresses have been assigned so that it will not assign the same address twice. The DHCP server also is configure with what is called a “lease”. The lease is the amount of time that, after the address is assigned, it is guaranteed not to be assigned to another device. DHCP assigned addresses may have their leases renewed an infinite number of time. A device will request an address renewal every ½ * (lease time).

    The DHCP server is configured with all the same information as a device that has a static address is, except instead of a single IP address, it is configured with a range of addresses. This range is known as the DHCP pool. The overall configuration, the pool and other setting associated with the pool is known as the DHCP scope. In addition to all but eliminating administrative overhead, the other major benefit of DHCP is that additional hosts may be added without doing any configuration of them.

    If a device shows an IP address of 169.254.X.X, that means that it was unable to obtain a DHCP supplied address for whatever reason. Reasons may include that the pool has been exhausted, the DHCP server is no longer functioning or a network problem of some sort.


    All of the data the hosts send must be transported in some way. In networking jargon, the transport medium is know is the physical (PHY) layer or, in the OSI model, Layer 1. There are multiple physical mediums that may be used to transport the data and each of those mediums may have multiple specific configurations. I am not going to go in to every permutation, as that would literally require a book, but I will discuss the ones that I think are most relevant to the DBSTalk readers.
    The three main classes of transport are open air, copper and fiber optics.

    Open Air
    Open air is wireless. This may be Wi-Fi or some other type of wireless. While many types of wireless exist I will focus almost exclusively on Wi-Fi, except when discussing networking non-connected buildings together. Data will travel in open air at (very, very near) the speed of light.
    Wi-Fi is generally considered a short distance transport, typically measured in 10’s of feet. Walls, floors, etc will reduce/block the Wi-Fi signal and result in lower data rates. Specialized antennas (and other wireless technologies) can enable to use of Wi-Fi over the distance of miles. (Some wireless technologies even allow for non line of sight (NLOS) usage over those long distances.)

    Copper is the tried and trusted workhorse of the networking world. It comes in many different flavors, twisted pair, coax, twinax and others. Only twisted pair and coax will be discussed as the others are for very specific applications or are obsolete. Data will travel through copper at approximately 2/3 the speed of light.

    Twisted pair cable comes in various categories or “Cats”. The only cables you are ever likely to run in to are Cat3, Cat5/5e, Cat6 and Cat7. Cat3 is suitable ONLY for telephone connections and 10mb or 100mb Ethernet. Cat5 and Cat5e are, for all practical purposes, identical. You will likely only be able to find Cat5e. Cat5e is what you should use if you are doing any cabling project. It will support Gigabit Ethernet. Do not waste your money on Cat6 or Cat7 unless you plan on running 10 Gigabit Ethernet to your hosts and have (the VERY expensive) network infrastructure to support it.

    Coax was the medium that Ethernet was originally designed around. While it is still possible to purchase network cards that are designed for coax, it is considered an obsolete technology and only supports speeds of up to 10mb. MoCA/DECA use coax as the network transport medium as well. It is, however, not Ethernet. It is a pseudo-proprietary broadband modulation scheme.

    Copper is generally considered a medium distance transport. The various IEEE 802.3 specifications generally permit a maximum distance for twisted pair of 100m between any two hosts/devices. (If you are using a patch panel and/or wall plates, the length of the patch cables used between the patch panel and/or wall plate must be taken in to consideration as it counts towards the 100m maximum.)

    Fiber Optics
    Fiber optic cables are like copper, except they are better at just about everything. Everything that is, except for price. They are relatively expensive for both the cable, termination and modules. Fiber cables come in two basic flavors, multi-mode and single-mode. Data will travel in fiber optics is (very near) the speed of light.

    Fiber optic cables are generally considered a long distance transport. Distance limitations for fiber range from just over .5km for multi-mode and up to 70km for single-mode. Usage of fiber for a home network is limited, except in limited circumstances.


    For the purposes of this document, there are two types of data that we will be talking about, frames and packets.

    A packet is the unit of interesting data that is being sent from one host to another.

    A frame is the unit of data that is sent down the transport. A frame contains, among other things, the source and destination MAC addresses. MAC addresses are how a network connected device determines if it should process the frame or ignore it. If the destination MAC address matches the devices MAC address it gets processed, otherwise it is ignored. Packets are encapsulated in the frame as the payload. Frames only exist at Layer 2.

    Frames also have a minimum and maximum size. The minimum size is 64 bytes and the total maximum size including headers and payload is 1538 (or 1542) bytes. Gigabit Ethernet supports jumbo frames, up 9000 bytes. If a packet is being sent causes the frame to exceeds the maximum size, it will be fragmented and sent as two (or more) frames.

    Network Devices

    Network devices are the “active” parts of the network. They actually process in some way the data that flows through them. These devices are what make the network actually work. Without them, you could only have a network with two nodes. Network devices come in two classes, business and consumer. Both classes will be discussed

    Firewalls are network security devices. They are the bouncers between you and the internet. While network security, in general and firewalls, specifically, are beyond the scope of this document, you need to have one. Virtually all consumer class “broadband routers” have firewall capabilities and you likely already have one. Some of the more advanced consumer class firewalls have VPN capabilities that allow you to securely connect to your home network when you are not at home.

    A network hub is nothing more than a repeater. Everything that comes in to a hub is sent back out on EVERY other port on the hub. It does not matter if the host is interested in that data or not, if hosts are connected to a hub, they will receive a copy of every frame sent. Hubs are half-duplex, as such device cannot send and receive data at the same time.

    Hubs are a “shared media” technology and everything connected to the hub is connected to the same Ethernet segment. That means, from a practical standpoint, it is as if every host or device connected to a hub is connected to the same physical cable, even if each host/device has a separate cable connecting to the hub. Every host/device that is connected to hub, no matter how many hubs are in between, is on the same network segment.

    If a host/device wants to send something over the network, it waits until it sees no traffic on network before it sends its traffic. If, which happens all the time, more than one host/device tries to send data on the network at the same time, a collision occurs. When a collision is detected, all transmitting hosts/devices wait a random period of time before they attempt to resend. The more hosts/devices on a segment, the more likely collisions are. Collisions decrease overall throughput. If the number of collisions is excessive, it will be very noticeable. Hubs are considered an obsolete technology and, with the advent of cheap switches, have been supplanted. Hubs, if incorrectly deployed, can severely degrade the performance of a network.

    Switches are basically hubs on steroids. If networking was the movie Twins, Danny DeVito would be a hub and Arnold would be a switch. (I am not implying that Arnold takes or has ever taken steroids.) Switches are Layer 2 devices and only understand frames and MAC addresses. A switch only deals with the local network. (There are numerous exceptions to this, but for the purposes of this document they are not relavent.)

    A switch is basically like an old time telephone switchboard operator. Instead of looking up people, it looks up MAC addresses and instead of connecting the calling and the called parties with a plug cord, it connects sending and receiving hosts’ ports (and only those ports) with a virtual cord. In theory, with a 12 port Gigabit Ethernet switch, each host could talk to one other host at full gigabit speeds, for a total bandwidth through the switch of 6 gigabits. A switch also creates a separate Ethernet segment on each port.

    A switch works by looking at each frame that arrives on a port and placing the source MAC address in a table. Now, when the switch sees a frame arrive, it looks at the destination MAC address and looks up that address in the table it created. When it finds the entry for the destination MAC address it looks up the corresponding port, then it directs that frame to only go out that one port. If the switch receives a frame with a destination MAC address that it does not yet have a table entry, it floods it out EVERY port, just like a hub. The idea behind this is that the destination will respond and the switch can then add it to its table and further frames will only go out that one port. This same flooding behavior occurs when a broadcast is sent, as a broadcast is intended for EVERY host/device on the network. Think of it like a store that has phones at each register and an overhead paging system. If you know that the person you want to speak with is at a certain register, you call that register. If you do not know where the person is, you use the overhead paging system to send a broadcast message. In the future you will know what number to call that person at and not have to page them again.
    Switches, depending on their configuration, also have the capability of accepting frames on one type of transport and sending it out on a different type of transport. This feature is typically only found in business class switches.

    DANGER WILLROBINSON, DANGER!!! Never, ever, connect two switches (or hubs) to each other with more than one cable directly or indirectly. (More on this below)

    A router does nothing more than connect two (or more) separate networks together so that they can talk to each other. The internet is composed of thousands and thousands of routers. Without going too deep in routers and routing, routers will have a “map” (routing table) of how to get to different networks. This map can be very detailed or very simple. In the most simple, it will have one entry that says “For all networks other than the ones that you are connected to, go <here>.” This is also know as the “default route”, “default gateway” or “gateway of last resort”. In networking parlance, routers and gateways are synonymous. Routers also convert from one type of access protocol (Ethernet, cable, DSL, T1, etc) to another. For 99% of the home networks out there, the only router present or needed is the one on their cable/DSL/whatever connection. Routers only speak and understand IP addresses.

    Wireless Access Points are devices that allow Wi-Fi enable hosts/devices to connect to wired network. WAPs are basically a hub (yes, a hub) for wireless devices. The standards for Wi-Fi are specified in IEEE 802.11. There are currently four 802.11 standards, a, b, g and n. The original specification only defined 802.11a and 802.11b. 802.11a stood for “advanced”, operates on the 5GHz band and supports data rates up to 54mbps and 802.11b stood for “basic”, operates on the 2.4GHz band and supports data rates up to 11mbps. Eventually 802.11g and 802.11n were standardized. 802.11g operates on the same band as 802.11b, but offers data rates up 54mbps. 802.11n operates on both the 2.4GHz and 5GHz band and offers data rates up 150mbps

    Most access points implement both b/g standards on one device, though b/g/n devices are becoming more popular. 802.11a devices are still produced, but are not as common as the high frequencies it uses are more attenuated when passing through walls and the like. 802.11b only hosts/devices should not be used in an b/g environment as it will cause all b/g hosts/devices to operate in 802.11b mode only.

    Most consumer class WAPs come with fixed internal omni-directional antennas, though external detachable antennas can be found easily. Warning, not all WAPs with external antennas are detachable. Attaching a different antenna allows the user to select an antenna profile that meets different coverage profiles.

    Wi-Fi is a shared medium, just like a hub. However, it (usually) performs better than one. No two wireless devices may transmit at the same time. When there are a small number of wireless devices, this is not an issue.
    Most WAPs can be configured to transmit on one of several frequency bands, also known as channel. (The specific frequencies and channels available depend on your country. In the USA, channels 1 – 11 are available.) In order to achieve the best possible performance with Wi-Fi, adjacent WAP should not be on the same or overlapping channels. If there is only one WAP deployed and there are no close neighbors, any channel will work. If there are multiple WAPs in use and/or there are neighbors close by who also have wireless, more care will need to be taken. In the US the non-overlapping channels are 1, 6 and 11. Channels may be reused as long as the WAPs using the same channel are non-adjacent. This rule is also in effect for neighbors WAPs, as they can interfere as well.

    Wireless Bridges
    Similar to, but different from WAPs are wireless bridges and point to point wireless. These devices are typically designed to connect different buildings together over short or large distances. These types of devices are beyond the scope of this document and will not be discussed in detail, though their applications will be noted.

    Media Converters
    Media converters convert from one type of transport to different type of transport. Going from copper to fiber requires a media converter. A wireless access point is a type of media converter. A MoCA/DECA adaptor is also a media converter.

    Most media converters do nothing more than convert the media, though some are specialized and have additional functionality.

    (To preface, I am still getting up to speed on this technology. I will correct/update this and related sections as I learn more on the topic.)

    Multimedia over Coax Adaptors (and the DirecTV Ethernet Coax Adaptor, which is the same/substantially the same as MoCA, except for the frequencies used on the coax) is a specialized media converter. A MoCA device uses the existing coax cable plant in the home as the network transport. It connects to the coax cable and has an Ethernet port on it to connect it to the main network or host device.

    The MoCA segment, like WAPs, is a shared medium. [I do not know as of now if MoCA supports full or half duplex operation.] The 1.0 spec supports 100mb aggregate bandwidth, with 1.1 supporting 175mb and 2.0 supporting 400mb or 800mb. DECA, although at version 1.1, conforms to the MoCA 1.0 spec, I believe. [Still researching] As MoCA is a shared medium all nodes on a MoCA segment share the bandwidth of that segment. Based on specification data, average latency on a MoCA network is 3.5ms (4.5 if more than 8 nodes are connected), all frames are encrypted and QoS is used. Specifics on the encryption and QoS are still being investigated.

    One important point of note, referring to MoCA/DECA as cloud is incorrect. It is a network segment. [This is taken directly from Wikipedia] Cloud computing is the delivery of computing as a service rather than a product, whereby shared resources, software, and information are provided to computers and other devices as a utility (like the electricity grid) over a network (typically the Internet). With a cloud, you have no control over or knowledge of which specific resource you are accessing. An argument could be made, however, that the HR34 coupled RVU enabled STBs or TVs are a “cloud”, but that would be a stretch. The DirecTV VOD service would describe a cloud much more accurately.


    All networks will have delay. The laws of physics state the nothing can move fast than the speed of light. Photons and electrons are something, so they are bound by this law. As the speed of light is slower than instantaneous, it means that everything takes time. The amount of time may or may not be perceptable to us humans, but it exists.

    Business class devices will typically exhibit less delay than consumer class devices. Regardless of which class of equipment you use, daisy-chaining devices together will increase the overall delay. The use of hubs will also increase delay.

    Processing Delay
    Processing delay is the delay that is induced by the device examining the frame. Whenever an active network device receives a frame, it has to do several things to it. If it is a switch, it has to examine the destination MAC address and look up to see which port that MAC is on. If it is a router and/or itself is the destination, it has to reassemble the packet if it was fragmented and then examine the destination IP address, then it can determine where the packet needs to go. Device load (not necessarily the amount of data) and the quality of the device is the determining factor. In a car/highway analogy, this would be determining where you want to go, more or less.

    Business class devices, as a rule, introduce less processing delay than consumer class device. However, with the added configurability and functionality of business class devices, the opposite can easily be true.

    Queuing Delay
    Queuing delay is nothing more than waiting in line for your turn to go. Queuing delay varies directly on and is almost exclusively a function of, traffic load. Generally this is minimal. However, if a lot of data is being pushed through the port, queuing delay can be significant. This would be analogous to waiting in line on a metered on ramp.

    Business class devices will almost always introduce less queuing delay than consumer class devices. Additionally, business class devices may have other ways to mitigate queuing delay for time/delay sensitive data.

    Serialization Delay
    Serialization delay, unlike the first two, does not vary based on device load. It only varies inversely based on link speed. A gig Ethernet link will have 1/10th the serialization delay of a fast Ethernet link. Think of this as the acceleration of your car.

    Propagation Delay
    Propagation delay is the amount of time it takes for the frame to go from one end of the link to the other. Propagation delay varies directly based on the length of the link. This is due to the speed of light and it determines the speed limit. It is a hard speed limit that is dependent on the specific material of the transport medium. In copper it is ~2/3 the speed of light, in fiber it is just below the speed of light. Using copper as the link medium, the propagation delay is about 0.000000015 sec/10 ft.
    It does not matter if you are using Gigabit Ethernet or Fast Ethernet, propagation delay will ALWAYS be constant for a give distance and transport. Gigabit Ethernet will transmit more data for a given time period, but it will still take the same amount of time to for the first bits to go from one device to another.

    Business vs Consumer Equipment

    (My knowledge and experience with business class network devices is almost exclusively Cisco and those are the products that I will reference. That is not to say that other companies products are not comparable and/or, in some cases, superior.)

    As a general delineation between business class and consumer class equipment is the ability manage equipment. I am not talking about a limited GUI configuration interface, I am talking about getting in to advanced configuration, settings and monitoring. That does not mean that there are not highly configurable consumer class devices or that all business class devices are highly configurable.

    For the purposes of this document I will look at the differences between business and consumer switches and access points.

    Most all consumer switches have no management interface on them whatsoever and have a minimal number of ports (4 – 8 seeming to be the most common). You just plug your hosts in to them and away you go. They are truly plug-and-play. Most “broadband” routers have a built in switch. Some consumer switches are also starting to support IEEE 802.3af PoE as well, but their selection is limited.
    Business class switches usually have 12, 24 or 48 ports and, while they will perform basic switch tasks right out of the box, they generally must be configured. Additionally they may support QoS, loop prevention, multiple media types and PoE. Generally business switches will process the frames faster and introduce less delay than their consumer counterparts.

    If you are looking to buy a business class switch, stay away from the “small business” line. Not that they are not any good, but you are likely buying the switch because you want more control.
    If you are going to buy a business class switch, eBay is your best bet. Craigslist is hit or miss, with a lot of sellers overvaluing the equipment they are selling.

    As a general statement, I would suggest the Cisco WS-C3550-24-PWR. This is a 24 port, 10/100 PoE switch that also has two GBIC (A GBIC Gigabit Ethernet module that allows you to choose between copper or fiber.) ports. It has good QoS capabilities and is very configurable. You can find then on eBay for between $100 and $150. It is an end of sale model and only supports Cisco Pre-Standard PoE, but it works very nicely. It is also a Layer 3 switch. (A Layer 3 switch can be thought of as a combination switch (with all the ports) and router (able to route between different IP networks). 99% of the home networks out there will not need the Layer 3 capabilities, but that may change in the future.

    If you insist on something more current and/or capable of supporting multiple Gigabit Ethernet ports, there are the 3560 and 3750 series. These are going to me more expensive, but they have more horsepower and support both Cisco Pre-Standard PoE and 802.3af PoE. Most home network users will likely want stay away from the 3750 as it has features that 99.999% of home users would never need and is crazy expensive. (List price for a 48 port, PoE Gig 3750 is around $8,000 after the standard Cisco discounts are applied. List is about $15,000.)

    Most consumer WAPs have a limited configuration interface. They are also limited on their antenna selection and usually only support a single SSID (wireless network ID).

    Business class WAPs, however, usually support a myriad of antenna choices, multiple SSID, advanced security protocols and other features. The biggest advantage is that all but the oldest WAPs support PoE and most of them support Cisco Pre-Standard PoE. If you connect the WAP to a PoE switch, you do not need a nearby power outlet or a wall wart. You could neatly mount it to the wall, using the WAP to cover up the hole/box in the wall for the Ethernet cable. Alternatively, you mount it in a recessed wall and hang picture over it and make it completely invisible.

    Putting it all together

    So, now that we have gone through all the various pieces, parts and definitions, where does it leave us? An informed decision about the design, re-design, retrofit or upgrade of the home network may now occur.

    Simple, fast
    The ideal home network will have all Ethernet cables running back to a single location with a single switch. Everything will connect to that one switch. All hosts will be connected via wired connection unless they are laptops or only support wireless (i.e. Nintendo Wii). A business class switch will be used and QoS will be applied to prioritize data coming from the DVRs.

    In the above scenario, host to host round trip delay will be sub-millisecond. Yes, as load increases, there will some delay, but unless that load is very high it will not increase delay appreciably. Input lag will be virtually non-existent.

    Simple, not quite as fast
    Now, if you are in a situation where you cannot easily run a cable (or more cables) to a location that has a lot of hosts, there are a couple of options. The preferred would be to get a second (quality consumer or another business class) switch and install it near all of the hosts. Run all of the Ethernet cables to this second switch and use the one cable you do have to connect the second switch to the main switch. Yes, they will be sharing the bandwidth of that one link, but you could use a Gigabit connection between the two switches.

    The other alternative is to connect all the devices that support Wi-Fi and only use Ethernet for those devices that require it. This is less than ideal, but your usage patterns will determine if that is acceptable.
    You still want to use a business class switch as your main switch and run as many of the cables to it as Possible.

    What you do not want to do
    Unless you ABSOLUTELY have to, you do not want to daisy chain switches to get all of your devices connected. Do not go from the Main Switch to Switch A to Switch B to Switch C. Not only does this add more points of failure, the delay will add up significantly and the devices on Switch C, if accessing something connected to the Main Switch will have the share the one link from Main Switch to Switch A with all of the device on Switch A and Switch B and Switch C. It is just a really bad idea and a very poor network design.

    Outbuildings add some additional complexity to the solution, but nothing that cannot easily be addressed. If the outbuildings are closer than 80m – 90m, you can use standard copper network cable. (Yes, the spec says 100m, but you need to leave that 10m – 20m for the patch cables. Remember the total cable length between any two network connected devices on copper must be no more than 100m.)

    If the total length would be greater than 100m, you will want to use fiber. If you do need to use fiber, contact a local structured cabling vendor. Most will have short spools laying around and you can get a good deal. Make sure you order more than you think you will need so you have freedom in where you run it. If the distance between buildings is less than 500m (max total length per spec is 550m) go for whichever type of fiber is cheaper. If it is greater than 500m, you will need single-mode. You will also want a professional to do the fiber termination. I have seen DIY kits out there. Do not use them. It does not turn out well. If you using a business class switch, get the appropriate gig modules, otherwise buy media converters.

    If you are connecting via copper or fiber you need to determine how you are going to run your cable. Overhead is the easiest, but can be ugly and could potentially be damaged by weather or tall vehicles. Underground is the ideal. If you are already trenching to the building for some purpose, lay down fiber or copper at the same time. (If you are using copper and the trench is for electrical service, the cables must be separated by at least 18” to avoid inductive interference.) *** Before you run any copper cables between separate buildings contact a certified electrician. Extremely dangerous situations may occur if NEC code and/or local ordinances are not followed ***

    If neither copper nor fiber is practical, the only other option is point to point wireless. Generally this will require a clear line of sight between the two buildings. The options and equipment is so varied and numerous that it is beyond the scope of this document. There are many good resources on the Internet to help you make a decision.


    There is a very good reason why DirecTV has DECA as the only “supported” configuration for MRV. Supportability. The customer service agents/service techs only need to know how to troubleshoot DECA. As long as the DECA segment is working MRV will work. They do not need to know the dos/don’ts, the ins/outs, the good/bad practices of networking. They only need to know DECA. That is very good reason.

    However, DECA is a very new technology. It is still maturing. Yes Wi-Fi is barely older, but the installed base, and by extension the amount of engineering dedicated to it, is much, MUCH higher. DECA also adds additional delay and complexity to the network. It is a shared medium. It is another set of devices on the network that could die. There are many reasons. The best may be that Ethernet, in its various and evolving standards is almost 40 years old. It is mature, it is stable and there are billions of ports of it deployed throughout the world.

    All the receivers that do not have built-in DECA transceivers are still plain old Ethernet devices. All that DECA is doing is creating a coax network for the data to run it. It is still split outside the box in to regular Ethernet. Yes, newer devices (H25) are moving exclusively to built-in DECA transceivers, but there are ways to mitigate that without having a full blown DECA install for every other receiver. (This will be the subject of a forthcoming document.)

    Special Bonus Section

    Remember the warning above about not connecting two switches together with more than one cable? There is one specific case where you can, which will be discussed after why you must, otherwise, never do so.
    When a switch receives a broadcast frame, it sends it out every interface, except the one it received it on. As long as there is one and only one path from a given switch to any other switch, this is not a problem. However, let’s say that there are 3 switches (A, B and C), connected in a triangle: A2 --- B1, B2 --- C1 and C2 --- A1. If a host connected to A3 sends a broadcast, A will send that broadcast out A1 and A2. B1 will receive the broadcast and sent it out B2. C2 will also receive the broadcast and send it out C1. Now, B2 receives a broadcast from C1 and C1 receives a broadcast from B2. Both B and C send it to A. A turns around and sends it out and B and C get the broadcast again. Now, only the connections between the switches were mentioned, but every time one of the switches receives a broadcast, it is resent out EVERY connected port to every device that is connected. Yes, even the DECA segment.

    The scenario above will repeat infinitely until the one of the links between the switches is disconnected or one of the switches stops processing packets. In certain loop situations, the number of broadcasts can multiply, going from 1 to millions or billions in very short order. This is called a broadcast storm. It is ugly. Entire datacenters have been brought to their knees by this. Supposedly consumer grade switches fried themselves because the load caused the chips to overheat. People have been fired for this. The worst that will happen at home is a cheap switch bites the dust and/or you have “weird” network problems/slowness/rebooting of switches.

    Virtually all business class switches are able to detect these loops and shut down the ports. That ability can be turned off and frequently is on ports that PCs connect to. Connecting multiple cables between two switches will not increase your bandwidth. At best it will be disabled and only re-enabled if the other cable has a fault. At worst you will have a broadcast storm.

    The only time you can connect multiple cables between two switches is if you are using a business class switch that supports IEEE 802.11ab (Link Aggregation) or Cisco’s EtherChannel. With either of these protocols, you can aggregate up to 8 links together to form one virtual, fault tolerant link.
  2. Draconis

    Draconis New Member

    Mar 16, 2007
    Las Vegas, NV
  3. Kevin F

    Kevin F Hall Of Fame

    May 9, 2010
    Impressive sticky. I read most of it and found it to be very informative. I vote for this to be a sticky.

    Very few typos found, no big deal.

  4. veryoldschool

    veryoldschool Lifetime Achiever Staff Member Super Moderator DBSTalk Club

    Dec 9, 2006
    While I'm not a network type, I am a "DECA type", and cloud does explain the RF paths between nodes, as there are many of them and it's hard to know exactly "where" the data is at any one point in time.
    The current DECA is version 1.1.

    "Good network people" don't tend to make good "RF people" and the opposite is true too.

    To truly "know how" DECA performs one first need to know how the current 50 MHz bandwidth is being utilized, which to date, only DirecTV knows.
  5. dsw2112

    dsw2112 Always Searching

    Jun 12, 2009
    There is good information for many in your write-up, but to be fair, parts are a bit editorialized. My background is in RF & electronics, but I work alongside IT types more and more. Most despise MOCA (and that's ok) but keeping with the facts (and less with opinion) will make a better sticky (especially in a D* forum.)
  6. veryoldschool

    veryoldschool Lifetime Achiever Staff Member Super Moderator DBSTalk Club

    Dec 9, 2006
    Not to, in any way, knock the effort here, but with ethernet being totally unsupported by DirecTV, I'm not sure of the need for this much info.
  7. Draconis

    Draconis New Member

    Mar 16, 2007
    Las Vegas, NV
    I took the time to read over the document and have a few suggestions and grammatical corrections. Otherwise this was very well thought out. Thank you for the time you took in putting this together.

    Corrections / suggestions / notes are in green.

  8. schwefel

    schwefel Cool Member

    Jan 3, 2009
    I thought I got all of them. Like I said it is a draft. Looking for comment/suggestions before I "publish" it.

    I am going to have to disagree on the cloud terminology again. Even if each receiver keeps open a connection to every other receiver, there will only be n(n-1) / 2 total connections. Based on my sniffs of the the pure Ethernet network, that is not the case. The data always come from the DVR.

    If are watching recorded show and then watch it again later, it will come from the same DVR, every time. However, if you use the Amazon EC2 cloud, your "server" will be on physical server one moment and possibly on a different physical server they next.

    I cannot disagree on the RF people/network people comment.

    DECA, based off of everything I can find is very similar, if not functionally identical to MoCA. Yes, only DirectTV knows all of the tech specs, but I find it hard to believe that they would build their own/complete re-engineer solution that is so fundamentally similar to MoCA. I suspect that this is an "embrace and extend" technology.

    Fair enough. On the next edit I will clean that aspect of it up.

    There seem to be a lot people out there using the "unsupported" configuration. DirecTV will, obviously, not support them so someone has to. :evilgrin:
  9. schwefel

    schwefel Cool Member

    Jan 3, 2009
    Thank you. Yeh, I have go through and red pen all of the "correctly spelled" incorrect words. I wanted to get it out bofore the work week burried me alive.

    As for the wireless devices, I was not aware of that. I will def look in to it and add it to the doc.
  10. Draconis

    Draconis New Member

    Mar 16, 2007
    Las Vegas, NV
    Another note, Whole-Home is not supported over a WiFi connection. The 2.4 GHz spectrum can be pretty polluted by other devices. The main reason for the CCK-W is to get the Internet onto the DECA network.
  11. schwefel

    schwefel Cool Member

    Jan 3, 2009
    Actually, polluted would be an improvement. That is what you get when you us unlicensed spectrum. Microwaves. Cordless phones. Baby monitors. Car alarms. Bluetooth. Who knows what else.
  12. veryoldschool

    veryoldschool Lifetime Achiever Staff Member Super Moderator DBSTalk Club

    Dec 9, 2006
    This is fine, but also where your networking background is limiting your understanding of the RF "cloud". Cloud may have more than one meaning, and in this case does explain the RF part [which is a bit like an electron cloud], where it may not by the time you break it out to ethernet.

    MoCA and DECA are very much the same, so if you can find out about how MoCA utilizes their [v 1.1] 50 MHz bandwidth it would crossover. DirecTV is part of the "alliance" [the A in MoCA].
  13. Rtm

    Rtm Godfather

    Oct 18, 2011
    Amazing thread x10
  14. DarkLogix

    DarkLogix Godfather

    Oct 20, 2011
    One thing I would add is on teh considerations for outbuildings, if they share a common electrical ground then copper can be acceptable, but if not then is not up to building code and you may fry something.

    The issue here is that although ground is considered zero volts there may be a potential differance between two differant grounds and that can be enough to cause damage, if you use coax and put a grounding block on both ends then it "may" be acceptable.

    In short if you're running copper between 2 buildings for communications find an electrician.

    I say this because
    1. I've measured the voltage across 2 buildings before
    2. My cisco instructor ran into this issue back before fiber was common (when wiring up the college and fried some transceivers)
  15. RunnerFL

    RunnerFL Well-Known Member

    Jan 4, 2006
    No one has to. Having no one support them is the definition of "unsupported".
  16. veryoldschool

    veryoldschool Lifetime Achiever Staff Member Super Moderator DBSTalk Club

    Dec 9, 2006
  17. DarkLogix

    DarkLogix Godfather

    Oct 20, 2011
    Its more of officialy "unsupported" is D* won't help you, but forums are for people to help people so its my opnion that we should help (ie uspport) them if we so choose, though not be required to do so.
  18. veryoldschool

    veryoldschool Lifetime Achiever Staff Member Super Moderator DBSTalk Club

    Dec 9, 2006
    While this is true, it really goes back to this was never to be used for MRV, but as we [DBSTalk] tested it for a long time in beta, we were able to leverage them to open a loophole for us.
    "Going forward" I would suggest moving to DECA as ethernet ports are being deleted from newer equipment [ie the H25].
  19. DarkLogix

    DarkLogix Godfather

    Oct 20, 2011
    for 98% of all residential customers Deca will do them just fine
    for the top 2%(ie large homes) they will likely have someone that can sort it out for them or will come here and get info.
  20. Stuart Sweet

    Stuart Sweet The Shadow Knows!

    Jun 18, 2006
    Guys, meet schwefel! He's done some excellent work here, and we'd like to include this in our new tips and resources docs.
Thread Status:
Not open for further replies.

Share This Page