When connecting switches and hubs, the loop configuration should not be done.This occurs when two switches are directly connected with two LAN cables or when both ends of one cable are connected to the same switch.So why not make a loop, what happens when you make a loop?
When I make a loop configuration
What happens if you make a loop with two switches?Before explaining the reason, let's introduce from the results.Here, two switches have been prepared with several PCs connected.First, connect the switch between the switches with one LAN cable and confirm that communication is being performed.The access lamp on the front part of the switch flashes according to the communication.
However, in this state, when another LAN cable was connected, the access lamp was lit immediately afterwards.According to the traffic that flows between the switches, the bandwidth was always nearly 100 % (Fig. 1), and almost all of them were on a broad cast frame.It is a so -called "broad casttom" phenomenon that continues to flow in large quantities of broad cast frames.
Fig. 1 Traffic amount flows between the switches in a loop state
When I tried ping from one PC to another PC in this state, "Request TimeD OUT.Continued, and communication was not successful.Of course, you cannot transfer files between PCs or access to the Internet website by sharing Windows files.
The answer, "What happens when two switches are connected with two cables?" The broadcast storm occurs and communication is unable to communicate between nodes connected to the switch.Certainly, you shouldn't do it.
Photo 1 Two switches used for experiments.Since the STP is enabled in the initial state, the lower intelligent switch was disabled and experimented.
So why do you create a broad casttom when making a loop?The Broadcast Frame is an Ethernet frame with the destination MAC address "FF: FF: FF: FF: FF: FF", and is sent to all nodes belonging to the same subnet.Unlike the IP broadcast, the broadcast frame has no function to send it only in the subnet.However, since the router does not transfer the broadcast, it is eventually limited to the subnet.Broadcast frames are used in ARP, which solves the MAC address from the IP address, and the PC regularly offers a broad cast frame.
Then, how will the PC transmit when the PC sends a broad cast frame?In a normal network that is not a loop, it is sent to each node by one -way (Fig. 2).In other words, the switch transfers a broad -cast frame to all ports, but does not perform it to the source port.Therefore, the packet once sent will not return to the source.
Fig. 2 Flow of broadcast frame on normal network
On the other hand, the situation is different for a network where the switch is looped by two cables (Fig. 3).The broadcast frame sent by the switch from one cable returns from the other cable by the opposite switch.Then the switch transfers the frame again, which comes back again.This repetition occurs, causing a broad casttom.
Fig. 3 Broadcast frame that connects through a looped network
The IP has a "TTL (Time to Live)" mechanism that destroys packets via a certain number of routers.This is a mechanism that reduces the value of the TTL in the packet header by one by one via the router, and discards it when it becomes 0.This prevents a packet with unknown destinations from flowing on the Internet.
However, Ethernet has no such mechanism, and the broadcast frame is not automatically destroyed.Therefore, as long as the switch is looped, the frame keeps the switch forever.
The negative effect of the loop is not only the increase in traffic between switches.When the loop state, a large number of broad -cast frames can continue to be sent to the PC connected to the switch.Due to the load of this process, the operating speed of the PC may decrease.
In this experiment, when it became a loop state, one Windows 2000 -equipped PC did not even move the mouse cursor.I thought that the processing had stopped due to the hang -up of the OS, but as soon as the loop was eliminated, it began to work normally.PC freeze may not occur due to the performance of the CPU or network interface.However, even if it is operating, the CPU load is high, and it is certain that it will be difficult to use.
To prevent broad -cast storms, you may want to introduce an intelligent switch for STP (Spanning Tree Protocol).STP is a mechanism that eliminates the loop by blocking one route when there are multiple routes between the switches (Fig. 4).
Figure 4 STP (Spanning Tree Protocol) to prevent loops
STP also has a function to enable the other route when an enabled route occurs.Therefore, by creating multiple routes in advance, it can be used as a means of redundancy of the network.
STP -compatible intelligent switches are higher than normal non -intelligent switches.Therefore, it may be difficult to introduce it to the floor switch.However, if you use STP -compatible products only with the switch used for the back end, the trouble should be reduced.
This article re -edited the special feature in the December 2004 issue of Network Magazine.The content is in principle at the time of publication, and may be different from the present. |
