Outline

• Cell Technologies
• Packet Switching
• In-Class Exercise

Introduction

• Last day, we began to talk about wireless networks.
• We talked about Wifi:
  • The various 802.11 standards, (now WiFi 6, 7...).
  • Collision avoidance for WiFi nodes using MACA (RTS/CTS messages).
  • WiFi distribution systems
  • WiFi frame format.
• We then talked about Bluetooth. We looked at its notions of profiles, how it uses bandwidth, the spread spectrum techniques it uses, and device discovery.
• We then started talking about Access Networks (a network that lets one connect to the broader internet).
• We discussed PONs as a type of access network, and we began talking about cellular as an access network.
• We start today by reviewing what we said about cellular networks at the end of last day before continuing our discussion.

Cell Technologies

• Cell Phone technologies can be used to get mobile computers to communicate.
• The frequency bands for cell phones vary slightly for different places around the world. Typically and traditionally, it operates in the 700MHz to 2400 MHZ range. Newer 6GHz, and 24 GHz ranges are also opening up.
• Another issue is that there are a fair number of incompatible cell phone standards and a variety of legacy protocols which are supported by different carriers to different extents.
• Cell phones also rely on the use of base stations, often called Broadband Base Units (BBUs) (also known officially as Evolved Node B (eNodeB or eNB -- NodeB is what the radio unit used to called)).
• Mobile devices that connect to a BBU are referred to a User Equipment (UE).
• A set of BBUs are anchored at an Evolved Packed Core (EPC) hosted at a central office (often via a PONs). The area served by an EPC is called a Radio Access Network or RAN.
• The EPC typically consists of a MME (Mobile Management Entity that tracks movement of UEs), an HSS (home subscriber server, a database of subscriber info), and a Session and Packet gateway pair (to process and forward packets to and from the RAN and the internet)
• The geographic area served by one BBU station is called a cell.
• UEs are typically in range to communicate with several BBUs, however, at any given time they communicate with and are under the control of only one. If that BBU senses the signal strength is weakening, it will try to choose a BBU to handoff the UE to that is more like to have a stronger signal strength.

Generations Cell Technologies (1-3)

There have been several different generations of cell phone technologies:

1. (1G) was analog and didn't support data communications to any degree (it did have something later called D-AMPS).
2. (2G) and (2.5G) were digital technologies such as GSM (global system for mobile communication). North America was a late adopter of this technology so several other competing standard also existed. The different 2G technologies either combined FDM and TDM (this is what GSM does) or used a technique called CDMA (code division multiple access). The latter uses different chipping sequences for each sender. GSM used 8Kbps compressed voice and had a data part of its standard called GPRS (general packet radio service). Speeds were up to 30-70Kbps or a little higher using EDGE (what iPhone 1 used (2007)).
3. (3G) makes use of a variant of CDMA called wideband-CDMA (w-CDMA). It can communicate at speeds up to 1.92 Mbps (what iPhone 3G used (2008))

Generations Cell Technologies (4 and beyond)

• (3.something G) consists of several technologies some of which promote themselves as 4g. These were defined by the 3GPP (3G partnership project -- this partnership continues to define 4g and 5g standards). It includes family of standards such as HSPA and LTE, 3GPP2 family of standards such as EV-DO, and the IEEE family of standards such as Wi-Max.
• (4G) again several competing technologies. Came to include LTE-A, HSPA+, and WiMAX Release 2. Came into use between 2009-2012. (what iPhone 5 supports) Can support 150Mbps down / 50Mbps up. According to Wikipedia in 2021, it represented 58% of the mobile telecommunications market.
• Unlike WiFi whose contention mechanism is based on light load assumptions, cellular networks assume high utilization. For this reason LTE and its sucessors use a reservation strategy to allocate spectrum - Orthogonal Frequency Division Multiple Access (OFDMA - also in WiFi 6). 12 orthogonal subcarrier frequencies (15kHz wide) are used (each modulated independently). Data can be sent on behalf of multiple users by scheduling blocks of subcarrier frequency/ time intervals to that user's data. These resource elements (REs) consist of using one subcarrier frequency for the time need to transmit one OFDMA symbol. (The data in a symbol depends on the modulation rate, 4-bits using 16-QAM, 6, for 64 QAM). The scheduler operates at the granularity of 84 REs at a time (one Physical Resource Block). Signal quality feedback from the UE to the BBU is sent using a Channel Quality Indicator (CQI) mechanism which affects how this scheduling is done.
• (5G) (Supported by iPhone 12). Unlike 4G, 5G can operate at different frequencies using different choices for the subcarrier frequency ranges and time intervals. I.e., for different frequency ranges it uses different sized REs. For example in the sub 1GHz band it has two waveforms: one using 15Khz and 0.5ms, the other using 30kHz and 0.25 ms for it REs. 5G also supports a variety of waveforms at the 1GHz and 6GHz frequencies and a second group at the millimeter wave range (24GHz-71GHz -- more line of sight, so beamforming techniques needed - use range of frequencies to locate UE then use higher frequencies to target). Data speeds can be up to 1Gbps.
• (6g) is currently under development. It will likely involve similar ideas but operate at yet higher frequencies.

Packet Switching

• Devices that interconnect links of the same type are called switches or sometimes Layer 2 switches.
• We are now going to look at how packet switching is done in networks.
• One problem a switch must solve is which line to forward a request out on.
• We will also look at how to handle when switches become congested.
• We will look at LAN switching which evolved out of Ethernet Bridging.
• We will discuss switching in ATMs.

Switches

• We said we were interested in how switches know which line to forward on and how switches handle congestion.
• A switch is multi-input, multioutput device, which transfers packets from an input to one or more outputs.
• In terms of the topologies for networks we've seen so far, we have point-to-point links, Bus (Ethernet), and rings. Switches add to this a star topology.
• Some advantages of the star topology related to switches include:
  • Even keeping the degree of the switch vertex small (even as small degree 3), by connecting several switches, we can build a large network with a complicated structure.
  • We can connect switches to each other and to hosts using long-haul, point-to-point links, which means that we can build networks which span large geographic distances.
  • Adding a new host to the network by connecting it to a switch does not necessarily reduce the performance of hosts already connected to the network. So switched networks are considered more scalable than things like a strict Ethernet approach.

Forwarding

• A switch is connected to a set of links. Each link uses a data link layer protocol to communicate with the node or nodes on the other end of the link.
• As an example, a switch might have a T1, a T3, and a STS-1 line coming into it. So the data link protocol is not necessarily the same for each line.
• Where a link comes into a switch is called a port. Each port usually has a number associated with it. i.e., port 1, port 2, etc. of the switch.
• A switch needs to be able to receive data on one link and be able to send it out of the correct outgoing link. This job is called forwarding or switching.
• There are two common approaches to solving the switching decision problem: a datagram (also called connectionless) approach and a virtual circuit (also-called connection-oriented) approach.
• There is also a less common approach called source routing.
• MAC addresses are used to uniquely identify all switches/hosts in a switched network. That is why Ethernet cards have to have a unique address.

Datagrams

• In this approach, to decide how a switch forwards a packet, the switch consults a forwarding table that it has inside its memory.
• This table consists of two columns one for the destination host and the other for the port to use as the outgoing line. For example, for switch 2 above, the table would look like:

  Destination	Port
  A	3
  B	3
  C	1
  D	0
  E	0

In-Class Exercise

• Make a similar table to the last slide for switch 3.
• Please post your solution to the Feb 22 In-Class Exercise Thread.

More on the Datagram Approach

• In the datagram approach, we can imagine that the network operator configures all the cables and sets up the the routing table statically.
• For small networks this is possible, but for large, dynamically evolving networks, this is close to impossible. This harder problem is called routing, and we will talk about routing tables later in the semester.
• Connectionless approaches have the following characteristics:
  • A host can send a packet anywhere any time. This contrasts with connect-oriented approaches where a "connection-state" might need to be established first.
  • When a host sends a packet, the host does not know if the network is capable of delivering it or if the receiver is even alive.
  • Each packet is forwarded independently of the previous packets. Thus, if a forwarding table gets changed between two packets being sent, then the two packets might take different routes through the network.
  • A switch or link failure might not have any serious effect if it is possible to find an alternate route around the failure.

Virtual Circuit Switching

• In this approach, a virtual circuit/connection (VC) is set up from the source host to the destination host. Once this connection is established, data is then sent on this connection.
• This is a two stage process: (1) connection setup and (2) data transfer.
• We will now look at the harder of these, connection setup...

Connection Setup -- VC Table Entries

• In this stage, it is necessary to establish a "connection state" in each of the switches between the source and destination.
• The connection state for a connection consists of an entry in a VC table in each switch through which the connection passes.
• This entry contains:
  • A virtual circuit identifier (VCI) that uniquely identifies the connection at this switch, and which will be carried inside the header of the packet that belongs to this connection.
  • An incoming interface (i.e., incoming port) on which packets for this VC arrive at the switch.
  • An outgoing interface (i.e., outgoing port) in which packets for this VC leave this switch
  • A potentially different VCI will be used for outgoing packets.
• Given a packet, a switch, by looking at the VCI in its header, can find the corresponding entry in its VC table, and then use the correct outgoing line. Before sending on this outgoing line the VCI in the header of the packet is set for the next switch.
• The VCI is only significant for one link in the circuit path. That's why each link we change the VCI.

Connection Setup -- Setup algorithms

• There are two main approaches to setting up connection states in the switches.
  1. The network administrator manually sets up the connection link for each switch in the circuit. That is, a permanent virtual circuit (PVC) is established.
  2. A host can send signalling messages into the network to cause the state to be established. This establishes a switched virtual circuit (SVC).
• Let's look at the second approach. To do signalling, host A sends a message into the network, say to switch 1, with which it is connected. This message contains the complete destination address of host B to which A wants to send data.
• We assume the network is smart enough to be able to propagate the signalling message all the way to Host B. Each switch the message goes through sets up a VC entry. The task of assigning a VCI value for the incoming port through which message was sent is up to the switch. The outgoing VCI is at this point empty and will be determined by the next switch in the route.
• Note VCI values might be significantly smaller than the 6 bytes used for Ethernet address.
• Once the signalling packet reaches B, an acknowledgement packet is sent back to A. This packet traverses the connection just established back to A. As it does the outgoing VCI's of the in between switches are set up based on the value in the acknowledgement packet.

More on Virtual Circuit Switching

• Once the connection is established for the SVC, data can be transferred along this connection just as in the PVC approach.
• When A no longer needs the connection, it can tear down the connection by sending a tear down message to the first switch along this connection.
• This switch would remove the VC table entry for the connection and send along the tear-down message to the next switch in the connection...
• There are several things to note about virtual switching:
  • There is at least one RTT before data can be sent.
  • Although the connection request message needs to have the full address of B, thereafter, each packet from A only needs the much smaller VCI of the next switch in the link.
  • If any switch in the connection fails, the whole connection fails, and a new connection needs to be established.
  • You still need some kind of routing algorithm to send the connection message in the first place. We will talk about how to do this when we cover Chapter 4.
  • If a connection can be established then Host A knows Host B is alive and is willing and able to receive data from A.

Resource Allocation

• During the connection setup phase, it is also possible to allocate resources to the service.
• As an example of this, consider what is done in an X.25 network (a connection-oriented network that used to be used by the telephone companies):
  1. Buffers are allocated to each virtual circuit when the circuit is initialized.
  2. The sliding window protocol is run between each pair of nodes in the VC, and this is augmented with flow control so that the sender on each link does not send too fast and overrun the buffers on the receiving node.
  3. The circuit is rejected by a given node if not enough buffers are available at that node when the connection request message is processed.
• So in X.25, each node is ensured of having the buffers needed to queue packets that arrive on that circuit. This is called hop-by-hop flow control.

Final Remarks on Virtual Circuits

• A datagram network in contrast to a virtual circuit network has no connection establishment phase. Each packet that arrives, regardless from whom, competes for buffer space with all other packets.
• In the virtual circuit model, one could imagine providing each circuit with its own level of Quality of Service. We will talk about QoS again when we describe the transport layer.
• In addition to X.25, some popular virtual circuit technologies are frame relay and ATM.
• Frame relay is somewhat lighter weight than X.25. Maybe network service providers provide Frame Relay PVC services which can be used to construct Virtual Private Networks, which we'll discuss when we cover Ch 4.