Introduction to performance metrics for routers

Speaking of the performance of routers, may I first give you a brief introduction of what is a router, what routers are used to do!

A router is a network device that connects multiple networks or segments to "translate" data information between different networks or segments so that they can "read" each other's data to form a larger network.

Routers have two typical functions, namely data channel function and control function. The data channel function includes forwarding decision, backplane forwarding and output link scheduling, which is usually accomplished by specific hardware, and the control function is generally implemented by software, including information exchange, System configuration and system management between neighboring routers.

For many years, the development of the router has a volt. In the middle of the 90, traditional routers became the bottleneck restricting the development of Internet. Instead, the ATM switch becomes the core of the IP backbone, and the router becomes a supporting role. Into the late 90 's, the Internet expanded further, the flow of traffic every six months, the ATM network has become a bottleneck, the router comeback, Gbps routing switch in 1997, people began to use Gbps routing switches to replace ATM switches, architecture to router as the core backbone network.

I believe you have a basic understanding of the concept of routers, then we will say the core router 10 performance indicators:

The system switching ability and processing ability of high speed routers is an important embodiment of their ability to be different from the general routers. At present, the high speed router's backplane switching capacity should reach more than 40Gbps, and the system, even if the oc-192/stm-64 interface is temporarily not provided, it must be in the future without the existing interface cards and general parts upgrades to support the interface. In terms of device processing capability, when the system is running at full load, all interfaces should be able to handle short packets at wire speed, such as 40-byte, 64-byte, at the same time, the exchange matrix of high-speed routers should be able to handle all the exchange of all interfaces without blocking, regardless of traffic type.

One metric: Throughput

Throughput is the packet forwarding capability of routers. The throughput is related to the number of router ports, port rates, packet lengths, packet types, routing calculation patterns (distribution or centralization), and test methods, which generally refer to the processor's ability to process packets. The packet forwarding capability of a high-speed router is at least 20Mpps above. Throughput includes two main aspects:

1. Overall throughput

The whole machine refers to the package forwarding capability of the equipment, which is an important index of the equipment performance. The router's job is to select the path based on the IP header or MPLS tag, so the performance metric refers to the number of forwarding packets per second. Overall throughput is usually less than the sum of all port throughput of the router.

2. Port throughput

Port throughput refers to the port packet forwarding capability, which is the packet forwarding capability of a router on a port. Typically two identical rate test interfaces are used. Generic test interfaces may relate to interface locations and relationships, such as the throughput of tests between ports on the same card may differ from the throughput values between ports on different cards.
Index Two: Routing table capability

Routers usually rely on the established and maintained routing tables to determine the forwarding of packets. The routing table capability refers to the limit of the number of routing table entries that are accommodated within the routing table. Because routers that perform BGP on the internet usually have hundreds of thousands of routing table entries, the project is also an important manifestation of router capabilities. In general, a high-speed router should be able to support at least 250,000 routes, averaging at least 2 paths for each destination address, and the system must support at least 25 BGP peers and at least 50 IGP neighbors.

Index three: Back plate ability

The back panel refers to the physical path between the input and the output port. Backplane capability is the internal implementation of the router, the traditional router uses shared backplane, but as a high-performance router will inevitably encounter congestion problems, and secondly, it is difficult to design a high-speed shared bus, so the existing high-speed routers generally use interchangeable backplane design. Backplane capabilities can be reflected in router throughput, and backplane capabilities are usually greater than the values calculated based on throughput and test Baochang. But the backplane ability can only be reflected in the design, generally can not test.

Index four: Packet loss rate

Packet loss rate refers to the proportion of packets that cannot be forwarded due to lack of resources in a stable continuous load. Packet loss rates are often used to measure the performance of routers as they overload their work. Packet loss rate is related to packet length and packet sending frequency, in some environments, can be combined with routing jitter or a large number of routes to test simulation.

Index Five: Delay

Delay refers to the time interval at which the first bit of a packet enters the router to the last bit from the router. The time interval is the processing time of the router that stores the forwarding mode. Time delay is related to packet length and link rate, and is usually tested in the range of router port throughput. Delay has a large impact on network performance, as a high-speed router, in the worst case, requires 1518 bytes and the following IP Bao Shiyan are less than 1ms.

Indicator Number SIX: back-to-back frames

Back-to-back frames are the number of packets sent at the minimum frame interval that do not cause packet loss. This metric is used to test router caching capabilities. A router with a linear speed full duplex forwarding capability, the index value is infinitely large.

Index VII: Delay jitter

Delay jitter refers to the time delay variation. The data service is insensitive to delay jitter, so this index is not usually used as an important index to measure high speed routers. For other services, such as voice and video services, which are on IP, it is necessary to test this index.
Indicator Eight: Quality of service capability

1. Queue management mechanism

The queue management control mechanism usually refers to the router congestion management mechanism and the queue scheduling algorithm. The common methods are red, wred, WRR, DRR, WFQ, wf2q and so on.

Queuing policy:

Support Fair queuing algorithm.

Support weighted fair queuing algorithm. The algorithm gives each queue a right (weight) that determines the link bandwidth that the queue can enjoy. In this way, real-time business can indeed get the required performance, inelastic business flows can be isolated from the normal (Best-effort) business flows.

In the management of input/output queues, the method of virtual output queue should be adopted.

Congestion Control:

Congestion control mechanisms such as WFQ and red must be supported.

A mechanism must be supported by which a higher drop priority should be marked for traffic that does not conform to its business-level Cir/burst contract, which should be higher than the drop priority that satisfies the contract's traffic and the best possible traffic.

In an Exchange environment where output queue contention is possible, an effective method must be provided to eliminate head congestion.

2. Number of port hardware queues

Typically, the priority supported by the router is guaranteed by the port hardware queue. The priority level in each queue is controlled by the queue scheduling algorithm.

Index Nine: Network management

Network management refers to the operation of centralized management of network resources by network administrator, including configuration management, accounting management, performance management, error management and security management. The degree of network management supported by the device embodies the manageability and maintainability of the equipment, and is usually managed using the SNMPV2 protocol. The granularity of network management indicates the granularity of router administration, such as management to port, to network segment, to IP address, to MAC address and so on. Managing granularity can affect router forwarding capabilities.
Metric Ten: Reliability and Availability

1. Redundancy of equipment

Redundancy can include interface redundancy, card redundancy, power redundancy, system board redundancy, clock board redundancy, equipment redundancy, and so on. Redundancy is used to ensure the reliability and availability of equipment, and the design of redundancy should compromise the reliability requirements of equipment and investment. Routers can guarantee the redundancy of routers through protocols such as VRRP.

2. Hot-swappable components

Because routers typically require 24 hours of work, replacing a part should not affect the router's work. Parts Hot plug is the guarantee of 24-hour router work.

3. No fault working time

This indicator indicates the time of the equipment failure to work according to the statistical method. Generally unable to test, can be calculated through the fault-free working time of major devices or the work of a large number of identical equipment.

4. Internal clock accuracy

Routers that have an ATM port to do circuit emulation or POS ports usually need to be synchronized. When using the internal clock, its precision will affect the BER.

In the High speed router technical specification, the high speed router's reliability and the reliability stipulation should meet the following requirements:

The ① system should meet or exceed 99.999% availability.

② no fault continuous working time: mtbf>10 million hours.

③ recovery time: System failure recovery time < mins.

④ system should have automatic protection switching function. Primary standby switching time should be less than 50ms.

⑤SDH and ATM interfaces should have automatic protection switching, the switching time should be less than 50ms.

⑥ requires high reliability and high stability of the equipment. Main components such as main processor, main memory, Exchange matrix, power supply, bus arbiter and management interface should have hot backup redundancy. The line card requires a M+n backup and provides remote test diagnostic capabilities. A power failure can keep the connection valid.

The ⑦ system must not have a single point of failure.