Linux system NBMA and BMA exchange method introduction

There are two different exchange methods in Linux system, namely NBMA and BMA. Don't look at the two exchanges just a letter, but it's completely different, even against each other. This article will introduce the exchange of these two networks in the Linux system.

1.NBMA exchange

While everyone is connected to a link, but that does not mean that the data frame is free to go anywhere. In the NBMA network, the control is on the switch, and the end host can only reach the target host through the switch. That is, only the switch is configured with the data frame from this port. Some data can pass the test. When the policy is sent from that port, the data frame can pass. For NBMA networks, the implementation of the switch is more complicated because it has a lot of protocol layer logic built in, rather than a passive learning. If we give a few examples of NBMA, it might be better.

For NBMA, the most typical ones are frame relay and ATM. The widely used technology is ATM. We know that ATM is very complicated and more complicated than Ethernet. I think it is more complicated than IP. For ATM, one of the important concepts is the virtual circuit and the virtual channel. The virtual circuit must be established before communication. The work of the ATM switch is to exchange the virtual path to transmit the data frame in the configured virtual channel. A specific data frame. You can only follow a virtual channel and will not reach other virtual channels. The configuration of the virtual channel of the ATM is performed on the switch, that is, the switch controls how the data frame is forwarded, and the end host can only passively accept the arrangement.

The strength of NBMA is its controllability, unlike BMA. For WAN, BMA is rarely used, and most of them use NBMA. Because the forwarding strategy of WAN is quite complicated, under the guidance of BGP, the link layer is generally a specific target forwarding, and no random broadcast can occur. The mandatory policy to each AS, if any broadcast occurs, will make the WAN no longer secure, and the security here may be highly correlated with politics and national security.

2.BMA exchange

In fact, the BMA network is not a switched network or a switch. The logical topology of the BMA network is always a bus topology. The concept of introducing a switch in the BMA network is completely complete. It is to optimize the network. Taking Ethernet as an example, the original Ethernet was a bus type. Later, HUB was introduced for easy expansion, and a bridge was used to connect to heterogeneous Ethernet. Later, Ethernet switches were widely used in high-performance and low-cost twisted pairs. The application of this fact has driven the wind down, and the bus-based coaxial line gradually withdrew. The biggest advantage of Ethernet switches is their learning function. This kind of learning is a kind of passive learning. Although there is no secret on the BMA network, due to the existence of the learning switch, the secret is limited to the mutually agreed communication parties. Throughout history, we can see that the Ethernet switch is a latecomer, although it really subverts the Ethernet architecture, so that later Gigabit, 10 Gigabit Ethernet is only compatible with the traditional 802.3 for compatibility, it Still can't subvert the essence of Ethernet, what is this essence? This is the ARP protocol, which is a broadcast that confirms that Ethernet is still the original broadcast network. Any communication has to use the arp protocol for address resolution (regardless of the static configuration of the arp mapping), this address resolution and subsequent unicast communication gives the switch a learning opportunity to know a certain MAC address to the switch, Which port can be reached, so the switch can filter out unnecessary broadcasts. This kind of learning is passive, even if there is no such learning mechanism, the data packet can reach the target, because the implementation of the BMA network is on the end host, not the switch. For Ethernet, the default data frame will reach the same chain. Any host on the road, the host will filter out the data frame whose target MAC is not the local machine. This is the essence of BMA - that is, it is implemented on the host, the switch only helps, and the switch is powerless for judgment (such as MAC/port mapping entry aging). It will return to the essence of BMA, that is, use broadcast to broadcast data frames on each port.

IPv6 cancels the arp broadcast, and to some extent is completely destroying the foundation of the nature of Ethernet broadcast. Even so, a data frame can change the destination MAC to a broadcast address. Before the standard is changed, the switch sees that the data frame will still be forwarded on all ports when the switch is still in the blind state.

The above is the introduction of the two network exchange methods of Linux system. NBMA and BMA can be said to have their own advantages and can make up for each other.