Network Devices: An Overview (2023)

11.1.1 Summary

A switch (sometimes called a switching hub) refers to a networking device that filters and forwards data packets on a network. A switch is typically a multiport device (it can have 48 or more ports;Figure 11.2(B) shows some switches), which means that it is an active element that works at layer 2 of the OSI model. Unlike a standard hub, which simply replicates what it receives, a switching hub stores the Media Access Control (MAC) addresses of devices connected to it. When the switch receives a data packet, it forwards it directly to the receiving device looking for the MAC address. The switch analyzes the frames arriving on its input ports and filters the data to focus only on the correct ports; As a result, it can act as both a port when filtering and a hub when handling connections. A network switch can use the full performance potential of a network connection for each device and is therefore preferable to a standard hub.

Here are some discussions of how hubs and switches perform their functions in networks.

(1) A network hub or repeater is a very simple transmission device. They do not manage the traffic passing through them and all packets that enter any port are forwarded to all other ports except the gateway. When two or more nodes try to send packets at the same time, this is called a collision, and the network nodes must go through a routine to resolve the conflict. The procedure is dictated by the Ethernet Carrier Sense Multiple Access with Collision Detection (CSMA/CD) protocol. Each Ethernet adapter has a receiver and a transmitter. If adapters didn't have to listen for collisions with their receivers, they could transmit data simultaneously with reception (full duplex). Since they must operate in half duplex mode (data flow in one direction at a time) and a hub relays data from one node to all nodes, the maximum bandwidth is shared by all nodes connected to the hub.

A core network behaves like a shared medium; h only one device can successfully transmit at a time, and each host is still responsible for collision detection and retransmission. Some hubs have special stacking ports so they can be combined to allow for more hubs than simply daisy chaining via Ethernet cables, but even then a large Fast Ethernet network will likely require switches to overcome the daisy chain limitations of hubs. It is possible to connect multiple hubs to centralize a larger number of machines; This is sometimes known as a daisy chain. To do this, the hubs are connected with crossover cables, a type of cable that connects inputs and outputs on one side with those on the other side. Hubs usually have a special port called an uplink to connect two hubs with a patch cable. There are also hubs that can automatically cross or uncross their ports depending on whether they are connected to a host or a hub.

(2) A network switch typically includes a set of input ports for receiving packets arriving on the buses, a set of output ports for forwarding packets on the buses, and a switching fabric, such as a crosspoint switch, for routing packets each other . inbound switch port to the outbound switch ports that should forward them. The network switch's input and output ports often include buffers to hold packets until they can be forwarded through the switch fabric or onto a network bus. Buffering an output port allows it to receive data faster than it forwards it, at least until the buffer is full. If the buffer is full, received data will be lost. A network switch port usually uses one or more SDRAMs to implement its memory buffer due to its low cost. Some incoming switch ports include protocol processors to convert each incoming packet into a sequence of uniformly sized cells. The input port stores the cells in its buffer until it can forward them through the switch fabric to one of thedeparture ports. Each outgoing switch port, in turn, stores the cells received by the switch fabric in its memory buffer and forwards them to another protocol processor, which reassembles them into a packet for forwarding to a data bus.

The traffic manager on a network switch forwards packets of the same stream in the order it receives them, but may preferentially forward packets with high priority. An outgoing switch port may contain a router to store cells received through the switch fabric in its buffer and then forward them to another protocol processor. The output port protocol processor reassembles each cell into a packet and forwards the packet over a network bus. A network switch output port can forward a packet to another network switch or to a network node on a channel selected from one of several different network buses, and the traffic manager of a network switch output port decodes the packet number stream identifier (FIN) to determine which output bus or bus channel should carry the packet out the port. The gateway router can also decode the FIN of a packet to determine a packet's priority and minimum and maximum forwarding rate. The network switch also includes an address translation system that maps a destination network address to an output port that can forward the packet to that network address. When a gateway receives an incoming packet, it stores the packet, reads its destination network address, consults the address translation system to determine which output port to forward the packet, and then sends a routing request to the network switching system.

Hubs and switches are used to divide networks into multiple subnets. For example, when a factory dynamically exchanges large amounts of data over the network, its traffic slows down the network for other users. To solve this problem, two switches can be used, connecting the computers of a plant to a network while the other computers connect to another. The two switches can then be connected to the router that sits between the internal network and the Internet. The floor's traffic is only seen by the computers on that network, but when they need to connect to a computer on the other network, the data is sent through the intermediate router.

In short, switches and hubs provide industrial users with much of the functionality that in the past could only be provided by separate proprietary control networks. The elimination ofCollisions connecting each node to a switched port, along with the hub's ability to prevent control and office traffic from inadvertently interacting while still using a physical network, allow industrial users to appreciate open architecture and enjoy high bandwidth. massive Ethernet speed without compromising the integrity of your control traffic.

introduction

With the explosive growth of datanetwork equipmentthe need arose to support many different serial protocols with a single port. The problem faced by interface designers is getting circuits for each serial protocol to share the same pins without introducing conflicts. The main source of frustration is that each serial protocol requires a different line termination that cannot be easily or cheaply changed.

With the introduction of the LTC1343 and LTC1344, a fully software selectable serial interface port is possible with an inexpensive DB-25 connector. The chips form a serial interface port that supports V.28 (RS232), V.35, V.36, RS449, EIA-530, EIA-530A or X.21 protocols in DTE or DCE mode and obedient NET1 and NET2 . The port is powered by a single 5V source and supports an echo clock and loopback configuration that helps eliminate tie-in logic between the serial driver and line transceivers.

Each LTC1343 contains four drivers and four receivers, and the LTC1344 contains six switchable resistor cables. The first LTC1343 is connected to the clock and data lines along with the LL (local loopback) and TM (test mode) diagnostic signals. The second LTC1343 is connected to the control signal lines along with the RL diagnostic signal (remote loopback). Single-ended controller and receiver can be separated to support RI (Ring Indication) signal. The LTC1344's selectable line terminations connect to high-speed data and clock signals only. If the interface protocol is changed via the digital mode select pins (not shown), the drivers and receivers are automatically reconfigured and the appropriate line terminations are connected.

4.3.5 Consolidation in the storage and fabric layers

Consolidation of data storage andnetwork devicescan lead to significant savings in space and energy consumption. Some vendors argue that by deploying multiprotocol network devices, the network fabric can be consolidated onto fewer resources, reducing footprint, power consumption, and cooling requirements.⁹In addition, the increasing use of blade servers and the migration of virtual machines are driving the use of networked storage, allowing storage efficiency to be improved through consolidation.[35].

Storage equipment manufacturers often provide multiple consolidated solutions under the unified storage banner. Traditionally, enterprise storage has used different storage systems for each storage function. A solution can be provided for online network attached storage,another for backup and archiving, while a third is used for secondary or near-line storage. These devices can use different technologies and protocols. With the goal of minimizing costs by reducing space and power requirements, unified storage solutions often support multiple protocols and provide transparent and consistent access to a pool of storage, regardless of the storage tier on which the data resides.[38](eg NetApps Data ONTAP, EMC Celerra unified storage platforms). Software systems are used to migrate data across different tiers of storage as per their reuse patterns.

1.4 Application architectures

Within the data center, an application architecture describes how functions are distributed across servers to provide end services to data center users. In the early 1960s, when computing power was expensive, application processing was done on a large centralized computer. Users interacted with this system through so-called "dumb terminals" (keyboards and screens without their own computational power). Data networks were limited to modems that communicated over the public telephone network between dumb terminals and large computers, usually at very low speeds (perhaps 10 to 56 kbit/s). With the development of microprocessors came the first personal computers in the mid-1970s. Although they could be configured to emulate a terminal connection to a larger computer, many new applications appeared that were compatible with desktop or laptop computers. As more and more data was distributed to individual users' computers, the need for an efficient file sharing mechanism quickly became apparent. Copying files to the hard disk, manually moving the hard disk to another computer and reinstalling the files has become very complicated for most users. This led to the development of Local Area Networks (LANs) in the late 1980s and early 1990s, which allowed computers to share files.

In addition to hardware architecture, data centers also use various software architectures. A detailed discussion of software architecture is beyond the scope of this chapter, although we provide some examples for good measure. Modern cloud computing systems can utilize a service-oriented architecture, which is actually an application development methodology that creates solutions through the integration of one or more web services.[3]. Each web service is treated as a subroutine or function call dedicated to performing a specific task (for example, processing a credit card or checking airline departure times). These Web services are treated as remote procedure calls or application program interfaces (APIs). Cloud computing environments can build software-as-a-service or platform-as-a-service offerings based on these approaches. Software architectures can include agile development methodologies, specifically the collaborative approach between developers, IT professionals and QA known as DevOps. These approaches emphasize principles such as fail fast and rely on other continuous improvement development processes such as the Deming Cycle (also known as Plan, Do, Check, Act or PDCA).[22].

These clusters can be large or small (up to 1000 servers) and organized into subgroups with interprocessor communication between the subgroups. An updated list of the world's top 500 supercomputers[23]It is maintained to provide an overview of the server, storage, and network connections currently in use. Most cluster networks are based on variations of Ethernet, although other proprietary network structures are used. Topologies can include variations of a hypercube, a torus, a hypertree, and full or partial meshes (to provide the same latency and shortest paths to all compute nodes). HPC networks can include four-way or eight-way equal cost multipathing (ECMP) and hash-based distributed forwarding of Layer 3 and Layer 4 ports.

1.4.12Chapter 13. Inexpensive optical amplifiers (Bruce Nyman and Greg Cowle)

Reducing the cost of optical amplifiers is becoming more and more important due to constant pressure on the price of optical amplifiers.network equipment, due to changes in network architectures and applications that are extending the reach of repeaters beyond the mainline line booster repeaters, and because the dominant EDFA technology is not as cost-saving through integration as other technologies such as semiconductors.

This chapter examines the issues associated with reducing the cost of optical amplifiers, with a focus on single-stage optical amplifiers as they are widely used in the highest volume and most cost sensitive applications such as: Channel gain for high speed channels with advanced modulation format, cable television distribution (CATV) amplifiers, and ASE sources for WDM Passive Optical Networks (PONs). Alternative technologies for low-cost amplifiers are discussed, such as semiconductor optical amplifiers and erbium-doped waveguide amplifiers (EDWA). EDFAs, which are the dominant technology, are composed of several components with different characteristics and are based on different technologies. Challenges and opportunities for reducing the cost of key components of EDFAs and the cost of labor to assemble EDFAs are discussed. EDWAs offer cost-saving opportunities by integrating the capabilities of many of the components required for optical amplifiers. However, the lower power-to-pump signal conversion efficiency in erbium-doped planar waveguides compared to erbium-doped fibers poses an obstacle to commercial realization of the potential cost advantages of EDWAs. A more recent approach is the PLC erbium-doped fiber amplifier, where many of the passive devices they are based on are integrated into a PLC, but the gain is provided by an erbium-doped fiber. This approach combines the cost benefits of PLC integration with theperformance and pumping efficiency of erbium-doped fibers and is particularly advantageous for complex amplifier architectures that require many optical components.

8.2 Who needs a web server?

14.2.1 Server Level Control

At the server level, many control variables are available for IT, power and cooling management. "Server" in this case means all IT equipment, including computer servers, storage drives andnetwork equipment. Compute resources such as B. CPU cycles, memory capacity, memory access bandwidth, and network bandwidth are all local resources that can be dynamically adjusted, especially in a virtualized environment. Power control can be done on either the demand side or the supply side, even at the server level. Server power consumption can be controlled through active management of the workload hosted on the server, such as admission control, load balancing, and workload migration or consolidation. On the other hand, power consumption can be adjusted through physical control variables such as dynamic voltage and frequency scaling (DVFS) and through on/off state control.[515],[516],[517],[518],[519],[520],[521],[522],[523]. DVFS is already implemented in many operating systems, for example the "CPU governors" in Linux systems. CPU usage is normally controlled by the DVFS controller, which adjusts power consumption to match the fluctuating workload.

Previous work has focused on how to handle the tradeoff between power consumption and IT performance. Varma et al.[524]discuss a control theory approach to DVFS. Cho and others.[515]Discuss a control algorithm that varies the clock speed of a microprocessor and the clock speed of memory. Leverich et al.[525]propose a control approachto reduce static power consumption of chips in a server through dynamic per-core wake-up power control.

Server-level cooling control is usually implemented by actively tuning the server fan to cool the servers.[526]. Similar to power management, the thermal state of servers (for example, processor temperatures) can also be affected by active load or performance control. As an example, Mutapcic et al.[520]focus on maximizing the processing capabilities of a multicore processor subject to specific thermal constraints. In another example, Cohen et al.[519]propose strategies to control the power consumption of a processor via DVFS to impose certain restrictions on chip temperature and workload execution.

17.5 Application example: WAN TE

All Tier 1 and many Tier 2 Internet Service Providers (ISPs) use some form of TE in their WAN infrastructures today. Providing greater determinism and better utilization of network resources are the main goals of TE, and MPLS TE networks are the preferred solution, mainly because normal IP networks cannot provide the same service and older methods of providing ATM TE networks or Frame Relay are no longer used.

But MPLS TE networks are expensive and complex; and do not provide network operators with the level of determinism, optimization and flexibility they require[14]. Note that in MPLS-TE, the reserved bandwidth of a tunnel is generally an estimate made by the network operator of the tunnel's potential usage. But traffic matrices change over time in unpredictable ways. Therefore, the reservation of a given tunnel can differ greatly from its actual usage at a given time, resulting in an unoptimized network.

Router manufacturers try to get around this problem with mechanisms like automatic bandwidth, but at best it's local throttling. Each router only knows the tunnels it creates and, to some extent, the tunnels that pass through it. For all other tunnels, the router only knows the aggregated bandwidth reserved by those tunnels on the links. In other words, although the router creates a map that provides the global state of the TE link, it only has a local view of the state of the tunnel (or TE-LSP state). As a result, local optimizations performed by various decision makers (tunnel headers) result in significant changes in the network.

Another option is to use a PCE. PCE can optimize all tunnels globally as it has a complete view of the tunnel and connection status. But PCE is an offline tool. HePCE calculation results are difficult to implement in live networks. Headend routers must be locked down individually and CLI scripts must be executed with care to avoid misconfigurations. This process is cumbersome enough to try less often (i.e. once a month).

With SDN and OpenFlow, we can get the best of both approaches by bringing the PCE tool "inline". We take advantage of the global optimization that the PCE tool provides, and then have the results of those optimizations directly and dynamically update the forwarding state (as routers can do). The net effect is that due to frequent (perhaps daily or hourly) throttling, the network operator can run a network with a much higher load without modifying the network and the operational problems of CLI scripts due to the programmatic interface of the platform. SDN TE and online. PCE -Application.

17.6 Conclusion

All six parts of this book series have highlighted the importance of fiber optic technologies in advanced telecommunications systems. Combined with advances in electronic switching and signal processing, as well as advanced protocols and network management systems, today's telecommunications network has achieved high levels of capacity, range, reliability, flexibility and affordability for users. For many years, capacity and cost considerations were the main driving forces behind advances in telecommunications. But recently, concerns about rising power consumption of telecom networks have brought energy efficiency into the mix of device vendors and network operators.

We identified several reasons for this recent increase in interest in energy efficiency. first hownetwork equipmentAs the capacity of optical transceivers and network switches and routers increases, the density of active devices must increase to maintain an acceptably small equipment size. This increasing density of equipment has created challenges related to heat dissipation from equipment racks. Improved thermal engineering can help alleviate some of these issues, but ultimately it is necessary to improve the energy efficiency of active devices. Second, the cost of ownership (OpEx) associated with device power consumption is becoming an increasingly important part of overall OpEx for network operators. In the past, energy costs were such a small part of a carrier's total cost of ownership that many operators paid little or no attention to energy. But that is now rapidly changing. Third, network energy consumption has a small but growing impact on global greenhouse gas emissions as the telecom network continues to expand to meet growing demand for new services and applications and to adapt. to a growing user base.

We also provide a detailed analysis of the energy consumption of the various major elements of a telecommunications system, including switching and transport. Fundamental power ratios were used to describe a lower bound for a minimum grid power consumption, based on technologies that are practical today and expected for the next decade. This result was discussed in terms of grid energy consumption estimates based on technology projections for commercial systems. The four orders of magnitude that separate these trends depend not only on the efficiency of the technology, but also on the many functional and performance requirements of today's business systems. Advances in energy-efficient telecommunications therefore require a combination of technological improvements with new, intelligent, service-aware features that can achieve significant performance or functionality with lower power consumption.

2.3 Summary