Multi-layer Exchange integration of multiple applications

In this increasingly networked world, people are constantly using communication between people and devices in new and different ways. Some types of communication are already familiar, such as IP voice (VoIP), digital images, multicast, video-on-demand, peer-to-peer file sharing, remote video conferencing, and more. However, all of these applications have a common feature: the need for network bandwidth can be described as "insatiable".

In the long run, bandwidth itself is always inadequate. The smart "devices" behind the network infrastructure--switches and routers--must assume the difficult task of being intelligent to keep pace with bandwidth requirements. Applications like video and digital x-rays always require a larger, smarter "pipeline", while VoIP applications require low latency and consistent delivery rates. In the middle of the 90, with the decline of traditional switches, people began to race to develop faster, more intelligent switches and routers. A group of talented people in Silicon Valley have seen this market opportunity to invent networking hardware and compatible software based on a new concept called "Multilayer switched routing". These new "smart" switches/routers provide faster speed and shorter latency than software-based routers at the time, and can combine the capabilities of multiple network devices.

Historically, when demand for network bandwidth increases, network administrators have redesigned the network to avoid router bottlenecks. The server often bypasses the router and is reinstalled more closely away from the user. For example, a group of stock-trading workstations may be away from other devices in the company and placed with servers that provide real-time data entry for them. This is because the fewer devices that share network resources (such as routers), the more bandwidth each device can get. Traditionally, the closer the user is to the data, the faster they can get the data, because it avoids the bottleneck of the router.

In large enterprises, users are divided into smaller networks (subnets) that interconnect through routers. The basis of user partitioning is usually geographical, operational type of application, amount of data required, and security reasons. For example, the accounting department is often placed in its own group to protect the financial records of the company, not because of the bandwidth they use. VoIP phones are often placed in their own networks, so they can bypass the bottleneck of traditional routers.

When computers need to communicate with other computers that are not on their local network, they send packets to their nearest router in order to send packets to their own group. Routers provide connectivity and security boundaries between the company and the Internet, as well as connections between groups within a company (intranet).

Traditional routers are used only when absolutely necessary, such as connecting remote offices over a WAN, connecting to the Internet, and isolating the key, high-bandwidth-demanding groups in the company. Traditional routers were expensive (and still are), and there was no significant progress compared to the original design, using components similar to a standard PC and running proprietary software using multiple interface cards.

In contrast, multilayer switched routers focus all of these functions on a dedicated special application integrated circuit or ASIC. The ASIC is less expensive than traditional routers and is typically distributed across network ports. Today, a typical switch/router may include 50 ASIC in a single device that can support hundreds of interfaces. The new ASIC allows intelligent switches/routers to forward data at very fast speeds on all ports, regardless of the type of network traffic. They forward traffic at the actual interface speed (often called the line speed). Today, new switches/routers are available on the market for corporate local area networks (LANs) that can forward traffic at a single interface of gigabit bandwidth (OC-192) per second.

Because of the use of a central architecture, traditional routers often lack scalability. All packets arriving at the router must be sent to a single processing area. The more interfaces you have, the heavier the load on the system, resulting in excessive resource consumption. This limits the services that can be run on the network, such as VoIP.

When a router using a central blocking architecture needs to handle more traffic than its own capacity, it starts discarding packets. When a network application or computer cannot receive a response, they send more packets for the recovery session. This only makes the situation worse-because it can easily overload the cross session. In this case, an overloaded router develops its own illogical thinking, optionally discarding packets based on application, user priority, or network destination/source. It is clear that a new approach to dealing with traffic growth is needed.

Over the years, the speed of traditional routers has been achieved a lot of growth, but still not enough to keep up with the pace of many powerful applications. For example, they can forward nearly 1 million packets per second. Consider a single 1000 Gigabit Ethernet interface capable of sending 1,488,000 packets (PPS) per second, but simultaneously receiving packets at the speed of 1,488,000 pps, which means that 2000 gigabit Ethernet ports can easily overload the system. In contrast, multilayer switches/routers forward packets at wire speed. The switching ASIC exists in a distributed manner, allowing the entire system to deliver traffic efficiently. When you add more interface cards, the processing power of the system increases-because the associated logic and forwarding decisions are distributed across the device. Some of today's high-end switches/routers can forward 480 million packets per second.

These new switches/routers use a new network design and management model. In the implementation of wire-speed forwarding today, blocking points can be eliminated, users can be farther away from the data, and do not have to worry about performance degradation. The stock trader we mentioned in the previous example can now connect to a server or network data that is several floors or hundreds of miles away from itself, depending on the type of interface supported by the switch/router and the type of cable or fiber used. In addition, new IP and optimized Ethernet routers are easier to manage, and it takes only a little time for managers to keep their networks synchronized with new applications. Like the network Bigiron chassis series products, simple transmission of all the flow from the application, while the capacity and speed of the increase in addition to more modules can be.

To determine the type and number of network traffic, ASIC now has a new packet sampling technology to provide the control platform for the whole system flow monitoring. RFC 3176 or Sflow has now become an increasingly popular method for businesses and service providers to provide real-time views of the application traffic in the network, the bandwidth required for traffic, and the flow of traffic. Sflow allows large enterprises to better monitor the use of network resources across multiple departments. Universities can identify illegal wireless and wired applications in the network and Detect and suppress denial of service (DoS) attacks in a timely manner before network performance is compromised. RFC 3176 is now a fast-growing requirement for important businesses that are aware of security.