|
A broadcast is a data packet that is destined for multiple hosts. Broadcasts can occur at the data link layer and the network layer. Data-link broadcasts are sent to all hosts attached to a p
articular physical network. Network layer broadcasts are sent to all hosts attached to a particular logical network. The Transmission Control Protocol/Internet Protocol (TCP/IP) supports the following types of broadcast packets:
-
All ones—By setting the broadcast address to all ones (255.255.255.255), all hosts on the network receive the broadcast.
-
Network—By setting the broadcast address to a specific network number in the network portion of the IP address and setting all ones in the host portion of the broadcast address, all hosts on the specified network receive the broadcast. For example, when a broadcast packet is sent with the broadcast address of 131.108.255.255, all hosts on network number 131.108 receive the broadcast.
-
Subnet—By setting the broadcast address to a specific network number and a specific subnet number, all hosts on the specified subnet receive the broadcast. For example, when a broadcast packet is set with the broadcast address of 131.108.4.255, all hosts on subnet 4 of network 131.108 receive the broadcast.
Because broadcasts are recognized by all hosts, a significant goal of router configuration is to control unnecessary proliferation of broadcast packets. Cisco routers support two kinds of broadcasts: directed and flooded. A directed broadcast is a packet sent to a specific network or series of networks, whereas a flooded broadcast is a packet sent to every network. In IP internetworks, most broadcasts take the form of User Datagram Protocol (UDP) broadcasts.
Although current IP implementations use a broadcast address of all ones, the first IP implementations used a broadcast address of all zeros. Many of the early implementations do not recognize broadcast addresses of all ones and fail to respond to the broadcast correctly. Other early implementations forward broadcasts of all ones, which causes a serious network overload known as a broadcast storm. Implementations that exhibit these problems include systems based on versions of BSD UNIX prior to Version 4.3.
In the brokerage community, applications use UDP broadcasts to transport market data to the desktops of traders on the trading floor. This case study gives examples of how brokerages have implemented both directed and flooding broadcast schemes in an environment that consists of Cisco routers and Sun workstations. Figure 19-1 illustrates a typical topology. Note that the addresses in this network use a 10-bit netmask of 255.255.255.192.
Figure 19-1: Topology that requires UDP broadcast forwarding.

In Figure 19-1, UDP broadcasts must be forwarded from a source segment (Feed network) to many destination segments that are connected redundantly. Financial market data, provided, for example, by Reuters, enters the network through the Sun workstations connected to the Feed network and is disseminated to the TIC servers. The TIC servers are Sun workstations running Teknekron Information Cluster software. The Sun workstations on the trader networks subscribe to the TIC servers for the delivery of certain market data, which the TIC servers deliver by means of UDP broadcasts. The two routers in this network provide redundancy so that if one router becomes unavailable, the other router can assume the load of the failed router without intervention from an operator. The connection between each router and the Feed network is for network administration purposes only and does not carry user traffic.
Two different approaches can be used to configure Cisco routers for forwarding UDP broadcast traffic: IP helper addressing and UDP flooding. This case study analyzes the advantages and disadvantages of each approach.
Note Regardless of whether you implement IP helper addressing or UDP flooding, you must use the ip forward-protocol udp global configuration command to enable the UDP forwarding. By default, the ip forward-protocol udp command enables forwarding for ports associated with the following protocols: Trivial File Transfer Protocol, Domain Name System, Time service, NetBIOS Name Server, NetBIOS Datagram Server, Boot Protocol, and Terminal Access Controller Access Control System. To enable forwarding for other ports, you must specify them as arguments to the ip forward-protocol udp command.
IP helper addressing is a form of static addressing that uses directed broadcasts to forward local and all-nets broadcasts to desired destinations within the internetwork.
To configure helper addressing, you must specify the ip helper-address command on every interface on every router that receives a broadcast that needs to be forwarded. On Router A and Router B, IP helper addresses can be configured to move data from the TIC server network to the trader networks. IP helper addressing in not the optimal solution for this type of topology because each router receives unnecessary broadcasts from the other router, as shown in Figure 19-2.
Figure 19-2: Flow of UDP packets from routers to trader networks using IP helper addressing.

In this case, Router A receives each broadcast sent by Router B three times, one for each segment, and Router B receives each broadcast sent by Router A three times, one for each segment. When each broadcast is received, the router must analyze it and determine that the broadcast does not need to be forwarded. As more segments are added to the network, the routers become overloaded with unnecessary traffic, which must be analyzed and discarded.
When IP helper addressing is used in this type of topology, no more than one router can be configured to forward UDP broadcasts (unless the receiving applications can handle duplicate broadcasts). This is because duplicate packets arrive on the trader network. This restriction limits redundancy in the design and can be undesirable in some implementations.
To send UDP broadcasts bidirectionally in this type of topology, a second ip helper address command must be applied to every router interface that receives UDP broadcasts. As more segments and devices are added to the network, more ip helper address commands are required to reach them, so the administration of these routers becomes more complex over time. Note, too, that bidirectional traffic in this topology significantly impacts router performance.
Although IP helper addressing is well-suited to nonredundant, nonparallel topologies that do not require a mechanism for controlling broadcast loops, in view of these drawbacks, IP helper addressing does not work well in this topology. To improve performance, network designers considered several other alternatives:
-
Setting the broadcast address on the TIC servers to all ones (255.255.255.255)—This alternative was dismissed because the TIC servers have more than one interface, causing TIC broadcasts to be sent back onto the Feed network. In addition, some workstation implementations do not allow all ones broadcasts when multiple interfaces are present.
-
Setting the broadcast address of the TIC servers to the major net broadcast (164.53.0.0)—This alternative was dismissed because the Sun TCP/IP implementation does not allow the use of major net broadcast addresses when the network is subnetted.
-
Eliminating the subnets and letting the workstations use Address Resolution Protocol (ARP) to learn addresses—This alternative was dismissed because the TIC servers cannot quickly learn an alternative route in the event of a primary router failure.
With alternatives eliminated, the network designers turned to a simpler implementation that supports redundancy without duplicating packets and that ensures fast convergence and minimal loss of data when a router fails: UDP flooding.
UDP flooding uses the spanning tree algorithm to forward packets in a controlled manner. Bridging is enabled on each router interface f [1] [2] [3] 下一页
 |
频道声明:本频道的文章除部分特别声明禁止转载的专稿外,可以自由转载.但请务必注明出出处和原始作者 文章版权归本频道与文章作者所有.对于被频道转载文章的个人和网站,我们表示深深的谢意。
| 原始作者:佚名 |
录入时间:2006-9-29 2:41:51 |
| 信息来源:不详 |
投稿信箱:itqoo@126.com |
|
|
 |
|