Course of lectures on network technologies. The capabilities of modern switches on the organization of virtual networks

Embedding information about virtual network accessories in the transmitted frame. Virtual local networksBuilt on the basis of the IEEE 802.1q standard, use additional frame fields to store information about VLAN accessories when moving over the network. From the point of view of convenience and flexibility of settings, the VLAN standard IEEE 802.1q is better decision Compared to VLANs based on ports. Its main advantages:

1. flexibility and convenience in setting up and change - you can create the necessary VLAN combinations both within the same switch and throughout the network built on switches with support for IEEE 802.1q. The ability to add tags allows VLAN information to spread through a set of 802.1q-compatible switches according to one physical connection ( trunk, Trunk Link);
2. allows you to activate the algorithm of the binder tree (Spanning Tree) on all ports and work as usual. The Spanning Tree protocol turns out to be very useful for use in large networks built on multiple switches, and allows switches to automatically determine the tree configuration of the network connections at an arbitrary connection of ports. For normal switcher, the lack of closed routes online. These routes can be created by the administrator specifically for the formation of backup bonds or randomly occur, which is quite possible if the network has numerous connections, and the cable system is poorly structured or documented. Using the Spanning Tree Protocol, switches after building a network circuit block excess routes. Thus, the occurrence of loops on the network is automatically prevented;
3. the ability of VLAN IEEE 802.1q to add and extract tags from frame headers allows you to use switches and network devices that do not support IEEE 802.1q standard;
4. devices of different manufacturers supporting the standard can work together, regardless of any proprietary solution;
5. to tie subnets on network levelYou need a router or switch L3. However, for simpler cases, for example, to arrange access to the server from various VLANs, the router will not be required. You need to enable the switch port to which the server is connected, in all subnets, and the server's network adapter must support the IEEE 802.1q standard.

Fig. 6.5.

Some identification of IEEE 802.1q

• Tagging ("frame marking") - The process of adding information about 802.1Q VLAN accessories in the frame header.
• Untagging ("Removing a tag from frame") - The process of extracting information about 802.1Q VLAN belongings from the frame header.
• VLAN ID (VID) - VLAN identifier.
• PORT VLAN ID (PVID) - VLAN port identifier.
• INGRESS PORT ("Input Port") - A port of the switch to which frames come, and at the same time decide on belonging to VLAN.
• EGRESS PORT ("Output Port") - The port of the switch from which the frames are transmitted to other network devices, switches or workstations, and, accordingly, it should make a decision on marking.

Any switch port can be configured as tagged. (labeled) or as untagged. (unmarked). Function untagging. allows you to work with those network devices Virtual network that do not understand tags in the Ethernet frame header. Function tagging. Allows you to configure VLANs between multiple switches that support IEEE 802.1q standard.

Fig. 6.6.

Tag VLAN IEEE 802.1Q

The IEEE 802.1Q standard determines the changes in the Ethernet frame structure, allowing to transmit information about VLAN over the network. In fig. 6.7 shows the format of the tag 802.1q

Building VLAN. based on ports based only on adding for more information to the address tables and does not use the possibilities embed virtual network accessories in transmitted frame. Its main advantages:

1. Flexibility and convenience in setting up and change, the ability to add tags allows information about VLAN. spread through a set of 802.1q-compatible switches according to one physical connection ( trunk, Trunk Link);

2. Ability VLAN IEEE 802.1Q. Add and remove tags from frame headers allows you to use switches and network devices that do not support the standard IEEE 802.1q;

3. Devices of different manufacturers supporting the standard can work together, regardless of any branded solution;

4. To associate subnets on the network level, a router or switch L3 is required. However, for simpler cases, for example, to organize access to the server from various VLAN.The router will not need. You need to enable the switch port to which the server is connected, in all subnets, and the server's network adapter must support the standard IEEE 802.1q.

Some identification of IEEE 802.1q

· Tagging ("frame marking") - The process of adding information about the accessories to 802.1q VLAN. In the frame header.

· Untagging ("Removing a tag from frame") - The process of extracting information on 802.1q VLAN. From the frame header.

· VLAN ID (VID) - Identifier VLAN..

· PORT VLAN ID (PVID) - port identifier VLAN..

· INGRESS PORT ("Input Port") - the port of the switch to which shots come, and at the same time decides on belonging to VLAN..

· EGRESS PORT ("Output Port") - The port of the switch from which the frames are transmitted to other network devices, switches or workstations, and, accordingly, it should make a decision on marking.

Any switch port can be configured as tagged. (labeled) or as untagged. (unmarked). Function untagging. Allows you to work with the virtual network network devices that do not understand the tags in the Ethernet frame header. Function tagging. Allows you to configure VLAN. between multiple switches supporting standard IEEE 802.1q.

Figure - marked and unmarked ports VLAN.

Tag VLAN IEEE 802.1Q

Standard IEEE 802.1q Defines changes in the Ethernet frame structure, allowing you to transmit information about VLAN. over the network. In fig. 6.7 shows the format of the tag 802.1q VLAN.. The Ethernet frame added 32 bits (4 bytes), which increase its size to 1522 bytes. The first 2 bytes (Tag Protocol Identifier, TPID field) with a fixed value of 0x8100 define that the frame contains the 802.1Q protocol tag. The remaining 2 bytes contain the following information:

– Priority ("Priority")- 3 bits of the transmission priority field are encoded to eight priority levels (from 0 to 7, where 7 is the highest priority), which are used in Standard 802.1.

– Canonical Format. Indicator (CFI) - 1 bit of the canonical format indicator is reserved to refer to frames of other types of networks (Token Ring, FDDI) transmitted by Ethernet highway;

– Vid. (VLAN. Id) - 12-bit identifier VLAN. Determines which VLAN. Belong traffic. Since under the field Vid.12 bits are assigned, then you can set 4094 unique VLAN. (Vid. 0 I. Vid. 4095 reserved).

I wrote this article for after finally understood the ethernet frame format (2nd level of the OSI model) and figured out how traffic was labeled to belong to VLAN.

I remind you that the Ethernet standard (FasteTherNet) data transmission technology described in the IEEE 802.3 Committee standard. When data transfer in the environment, the data on a 2-level level is divided into frames (frames) and sent to the transmission medium. Frame format is quite simple:

Fragrance frame Fastethernet

1. Preamble. Seven bytes of data intended for synchronization. Each byte contains the same sequence: 10101010. This field is used to enable transceiver schemes to come to steady synchronism with received signals. Also, this field turns on the SFD byte (not shown) - the initial frame limiter, which has the form: 10101011. The appearance of this combination is an indication of the forthcoming reception of the frame.
2. Dest Mac. Recipient Hardware Address (Destination).
3. SRC MAC. Source hardware address (Source).
4. Type: Type of top-level protocol. 0x800 - IP, 0x806 - ARP, etc. Full list can see :
5. Data: Actually data frame. It can take from 0 to 1500 bytes, but if the data is less than 46 bytes, then a special addition field is used, it is not indicated here. Thus, we believe that the frame contains 46-1546 bytes. The addition field is used for effective definition Collisses.
6. FCS: CRC32 frame checksum. What is the checksum - explain, I hope not necessary. In general, it is rarely used, it is much easier to check the integrity of packages or fragments on a higher level protocols. Well, for whom it is new, I will tell you why the checksum is needed. When sending a frame, the transmitting station calculates a special algorithm value that is written to this field. In the value, all bits of the frame are taken into account. When receiving a frame, the receiving Party calculates this value again (excluding this field) and compares with the value in the field. If they are equal, it is believed that the frame is obtained without errors.

Ethernet ETHERNET frame in 802.1q.

As we can see, almost everything remains unchanged. The same preamble, byte SFD, source and recipient MAC addresses. And then - added 4 new bytes. This is the so-called tag VLAN.. The remaining fields are unchanged, and just slightly displaced. When passing traffic through the port, the tag simply is withdrawn and work is carried out at the usual level.

I will consider in more detail the VLAN tag:

Bits	Value
1-16	Tag Protocol Identifier. On the scheme it TPID. For 802.1q Always equal 0x810. That is, having met these 2 bytes. It can be concluded that the traffic is labeled.
17-19	Priority.. Traffic priority. These three bits are used by the standard 802.1p. To set the priority of traffic. These are the following two fields - TCI.
20	Canonic Format Indicator - indicator of the canonical format of the MAC address. If bit \u003d 0, - canonical. If \u003d 1, not canonical. Well, I mean Token Ring
21-32	Vlan Identifier. In these 12 bits, the number is encoded. VLAN.. It can take a value from 0 to 4094. Not all switches support such a quantity, and in truth to say, it is not always necessary. In our case, this value is equal 0xa.What does the 10th mean VLAN..

If you think about how virtual networks work, it comes to mind that the whole thing is not in the shipping car, but in the frame itself VLVS. If there was any way to identify the VLVs by the header of the frame, would have to disappear to view its contents. At least, in new networks THNA 802.11 or 802.16, it would be quite possible to just add a special header field. In fact, the frame identifier in Standard 802.16 is just something in this spirit. But what to do with an Ethernet - the dominant network, which has no "spare" fields that could be given under the identifier of the virtual network? The IEEE 802 Committee took over this issue in 1995. After long discussions, it was done impossible - changed the format of the header of the Ethernet frame!? The new format was published under the name of 802.1Q, in 1998. The WLVS flag was inserted in the frame header, which we will briefly consider now. It is clear that the introduction of changes to something already established, such as Ethernet, should be produced in some kind of non-trivial way. Stand up, for example, the following questions:

• 1. And what, now it will be necessary to throw several millions of already existing Ethernet network cards on the garbage?
• 2. If not, then who will deal with the generation of new personnel fields?
• 3. What happens to frames that already have the maximum size?

Of course, the 802 committee was also concerned about these issues, and the decision, in spite of everything, was found.

The idea is that in fact the VLVS fields are actually used only by bridges and switches, and not by users of users. So, let's say, the network does not really concern their presence in the channels coming from the terminal stations, until the frames reach bridges or switches. In such a way that work with virtual networks can be possible, bridges and switches should know about their existence, but this requirement is also clear. Now we exhibit another requirement: they should be aware of the existence of 802.1q. The appropriate equipment is already available. As for old network, Ethernet cards, they do not have to throw out. The 802.3 committee could not force people to change the field type on the length of the Length. You can imagine what would have the reaction to a statement that all existing Ethernet maps can be thrown away? Nevertheless, new models appear on the market, and there is hope that they will now be 802.1) - secure and will be able to correctly fill out the identification fields virtual networks.

If the sender does not generate a field of a sign of a virtual network, then who does it? The answer is: the first bridge met the bridge or switch, processing the frames of virtual networks, inserts this field, and the last - cuts it out. But how does he find out in which of the virtual networks to pass? the local network Traffic router

To do this, the first device that inserts the VLVS field can assign a virtual network number to the port, analyze the MAC address or (God forbid, of course) to highlight the contents of the data field. While everyone does not switch to Ethernet cards compatible with the 802.1Q standard, all that will be. It remains to hope that all network fees of gigabit Ethernet will adhere to the 802.1q standard, from the very beginning of their production, and thus all users of the gigabit Ethernet of this technology will automatically be available 802.1q features. As for the problem of personnel, the length of which exceeds 1518 bytes, then in the 802.1Q standard it is solved by increasing the limit to 1522 bytes. When transmitting data in the system, both devices that reduce the VLVS says to an even account of anything (for example, classic or fast Ethernet), as well as virtual networking equipment (for example, gigabit Ethernet). Here, the shaded symbols mean VLV-compatible devices, and empty squares are all others. For simplicity, we assume that all VLVS-compatible switches are compatible. If this is not the case, the first such VLV compatible switch will add a sign of a virtual network to the frame, based on the information taken from the MAC or IP address.

VLVS-compatible Ethernet network cards generate frames with flags (that is, frames of 802.1q standards), and further routing is done using these flags. To routing the switch, as before, should know which virtual networks are available on all ports. Information that the frame belongs to the gray virtual network, and by and large, does not say anything, since the switch still needs to know which ports are connected to the gray virtual network machines. Thus, the switch is needed a port matching table with virtual networks, from which one could also know whether the VLVS ports are compatible. When the usual, unsuspecting the existence of virtual networks, the computer sends a frame to a virtual network switch, the latter generates a new frame inserting the flag of the VLVS. It receives information for this flag from the sender's virtual network (it uses the port number, MAC or IP address to determine.) From now on, no one else is experiencing due to the fact that the sender is a machine that does not support standard 802.1q, In the same way, the switch wishing to deliver a frame with a flag to such a machine must bring it to the appropriate format. Now consider the following format 802.1q. The only change is a pair of 2-byte fields. The first is the name of the VLV protocol identifier. It always matters 0x8100. Since this number exceeds 1500, then all network cards Ethernet interpret it as "type", and not as "length." It is not known what a card incompatible with 802.1Q will make, so such frames, in theory, should not go to it in any way.

In the second double-byte field there are three invested fields. The main one is the VLVS identifier, which takes 12 younger bits. It contains that information due to which all these transformations of formats, in fact, were rowed: it is indicated which virtual network is belonging to the frame. The three-bit field priority has nothing to do with virtual networks. Simply a change in the format of an Ethernet frame is such a weekly ritual that takes three years and is executed by some hundreds of people. Why not leave the memory of yourself in the form of three additional bits, and even with such an attractive appointment. The priority field allows you to distinguish between the traffic with the tight requirements for the reality of the time scale, the traffic with the average requirements and traffic, for which the transmission time is not critical. This allows you to provide more high quality Services in Ethernet. It is also used when transmitting voice via Ethernet (although there is already a quarter of a century in the IP there is a similar field, and no one has been required to use it). The last bit, CFI (Canonical Format Indicator - a classic format indicator) should be called the company's egoism indicator. Initially, it was intended to show that the format of the MAC address was applied with direct order of bytes (or, respectively, with reverse order), but in the heat of discussions about it somehow forgotten. His presence now means that the data field contains a safe frame 802.5, which is looking for another 502.5 format network and in Ethernet got completely accidentally. That is, in fact, he simply uses Ethernet as a means of movement. All of this, of course, almost no means associated with the virtual networks discussed in this section. But the policies of the standardization committee is not very different from the usual policy: if you vote for the introduction of my bits in the format, then I vote for your bitch. As mentioned earlier, when a frame with a virtual network flag comes to a VLVs-compatible switch, the latter uses a virtual network identifier as a table index in which it is looking for, on what port to send a frame. But where does this table come from? If it is manually developed, it means returning to the starting point: manual configuration of switches. The whole charm of transparency of bridges is that they are configured automatically and do not require any interference from the outside. It would be very shameful to lose this property. Fortunately, bridges for virtual networks are also self-adjusted. The setting is made on the basis of the information contained in the flags of the incoming frames. If a frame marked as VLVs 4 comes to port 3, it means undoubtedly one of the machines connected to this port is in the virtual network 4. The 802.1Q standard explains how dynamic tables are built. At the same time, references are made to the corresponding parts of the Perlman algorithm (Perlman), which entered the standard 802.id. Before you complete the conversation about routing in virtual networks, you need to make another remark. Many users of Internet networks and Ethernet are fanatically tied to networks without establishing a compound and frantically oppose them to any systems that have at least a hint of a network level or data level. However, in virtual networks, one technical moment is like a very strongly reminds the installation of the connection. This is the fact that the work of the virtual network is not possible without the identifier used as a table index built into the switch in each frame. This table defines a further fully defined frame route. This is what happens in the connection-oriented networks. In systems without connecting the connection, the route is determined at the destination address, and there are no identifiers of specific lines through which the frame must pass.

The two approaches described are based only on adding additional information to the address tables and do not use the ability to embed information on the frame to the virtual network to the transmitted frame. Method of the VLAN organization based on labels - tags uses additional frame fields to store information about the frame affiliation when it moves between network switches.

The IEEE 802.1Q standard determines the changes in the Ethernet frame structure, allowing to transmit information about VLAN over the network.

From the point of view of convenience and flexibility of settings, VLANs based on labels is the best solution compared to the following approaches. Its main advantages:

· Flexibility and convenience in setting up and change - you can create the necessary VLAN combinations both within one switcher and throughout the network built on switches with support for the 802.1Q standard. The ability to add marks allows VLAN to spread through a set of 802.1q-compatible switches according to one physical connection.

· Allows you to activate the coating tree algorithm on all ports and work as usual. The Spanning Tree protocol turns out to be very useful for use in large networks built on multiple switches, and allows switches to automatically determine the tree configuration of the network connections at an arbitrary connection of ports. For normal operation of the switch, there is no closed routes on the network. These routes can be created by the administrator specifically for the formation of backup bonds or randomly occur, which is quite possible if the network has numerous connections, and the cable system is poorly structured or documented. Using the SPANNING TREE protocol, the switches after building a network circuit block excess routes are thus automatically prevented by the occurrence of loops on the network.

· The ability of VLAN 802.1q add and extract marks from packet headers allows VLAN to work with switches and network adapters of servers and workstations that do not recognize the tags.

· Devices of different manufacturers supporting the standard can work together, regardless of any proprietary solution.

· No need to apply routers. To associate subnets on the network level, it is enough to include the desired ports In several VLANs, which will ensure the possibility of trafficking. For example, to arrange access to the server from various VLANs, you need to enable the switch port to which the server is connected, in all subnets. The only restriction is the server's network adapter must support IEEE 802.1Q standard.

By virtue of these properties, VLANs on the basis of tags are used in practice much more often than other types of VLANs.

5.6. Spanning Tree Covering Tree Algorithm

One of the methods used to increase fault tolerance computer network, this is Spanning Tree Protocol (STP) - Protocol of the binder tree (IEEE 802.1d). Designed for a long time, in 1983, it still remains relevant. IN ethernet networks, switches support only tree connections, i.e. which do not contain loops. This means that the organization of alternative channels requires special protocols and technologies that go beyond the framework of the basic to which the Ethernet relates.

If several connections are created to provide redundancy between switches, then loops may occur. The loop involves the existence of several routes by intermediate networks, and the network with several routes between the source and the receiver is characterized by increased resistance to violations. Although the presence of excess communication channels is very useful, loops, however, create problems, the most relevant of which:

· Broadcast storms - Broadcasting frames will be infinitely transmitted by networks with hinges using the entire accessible network transmission bandwidth and blocking other frames in all segments.

· Multiple copies of frames - The switch can get several copies of one frame, simultaneously coming from several sections of the network. In this case, the switching table will not be able to determine the location of the device, because the switch will receive a frame into several ports. It may happen that the switch will not be able to send the frame at all, because will constantly update the switching table.

To solve these problems, a protocol of the binder tree was developed.

Algorithm Spanning Tree (STA) Allows switches to automatically determine the tree configuration of links in the network with an arbitrary port of ports.

Switches that support the STP protocol automatically create a tree-like configuration of bonds without hinges in a computer network. This configuration is called the covering tree - Spanning Tree (sometimes they are called the core). The covering tree configuration is built automatically with the exchange of service packages.

Calculating the binder tree occurs when the switch is turned on and when the topology changes. These calculations require a periodic exchange of information between the binder switches, which is achieved using special packets called the BPDU BPDU protocol data blocks (Bridge Protocol Data Unit).

BPDU packages contain basic information necessary to build a hinge network topology:

· Switch identifier, on the basis of which the root switch is selected

· Distance from switch source to root switch (root route cost)

· Port identifier

BPDU packets are placed in the channel-level frame data field, such as Ethernet frames. Switches exchange BPDU at equal time intervals (usually 1-4C). In the event of a switcher failure (which leads to a change in topology) adjacent switches, without receiving the BPDU package for a specified time, the binder of the binder tree is beginning.

Modern switches also support RAPID STP protocol (IEEE 802.1W), which has better time convergence compared to STP (less than 1 second). 802.1W back compatible with 802.1d.

Comparison of STP 802.1D and RSTP 802.1W protocols.

5.7. Port aggregation and creating high-speed network highways

Port Trunking- This is a combination of several physical channels (link aggregation) into one logical highway. Used to merge together several physical ports in order to form a high-speed data transmission channel and allows you to actively use redundant alternative links in local networks.

Unlike the STP protocol (Spanning Tree - protocol of the covering tree), during the aggregation of physical channels, all redundant bonds remain in working condition, and existing traffic is distributed between them to achieve the load balance. If you refuse one of the lines included in such a logical channel, traffic is distributed between the remaining lines.

The ports included in the aggregated channel are called members of the group. One ports in the group acts as a "binding". Since all members of the group in the aggregated channel must be configured to work in the same mode, all changes to the settings made relative to the "binding" port belong to all members of the group. Thus, to configure ports in the group, you only need to configure the "binding" port.

An important point When implementing port combining into an aggregated channel is the distribution of traffic on them. If one session packages are transmitted by different ports of the aggregated channel, then there may be a problem for more high level OSI protocol. For example, if two or more adjacent frames of one session become transmitted through different ports of the aggregated channel, then due to the different lengths of the queues in their buffers there may be a situation when due to uneven delay in the frame transfer, a later frame will ring its predecessor. Therefore, most implementations of aggregation mechanisms use static methods, and not dynamic distribution of frames by ports, i.e. Fastening a certain session of two nodes behind a specific port of the aggregated frame stream. In this case, all frames will pass through the same queue and their sequence will not change. Usually, with a static distribution, the port selection for a specific session is based on the selected port aggregation algorithm, i.e. Based on some signs of incoming packages. Depending on the information being used to identify the session, there are 6 port aggregation algorithms:

1. MAC-address of the source;

2. MAC-address of the destination;

3. MAC-address of the source and destination;

4. IP address of the source;

5. The IP address of the destination;

6. The IP address of the source and destination.

Aggregated communication lines can be organized with any other switch that supports point-point data streams by one port of the aggregated channel.

Channel Combine should be considered as a network setting option used mainly for switch switch connections or "switch - file server" that require higher transmission speed than a single link can provide. Also this feature can be used to increase reliability. important lines. In case of damage to the communication line, the combined channel is quickly reconfigured (no more than 1 s), and the risk of duplication and changes in the order of frames is insignificant.

Software Modern switches support two types of communication channels: static and dynamic. With static channel aggregation, all settings on switches are performed manually. Dynamic channel aggregation is based on the IEEE 802.3ad specification, which uses the LACP Aggregated LINK CONTROL (LINK Aggregation Control Protocol) protocol in order to check the channel configuration and send packets to each of the physical lines. In addition, the LACP protocol describes the mechanism for adding and selecting channels from a single communication line. To do this, when setting up an aggregated communication channel on switches, the corresponding ports of one switch must be configured as "active", and another switch as "passive". The "active" ports of LACP perform processing and distribution of its control frames. This allows devices supporting LACP, agree on the settings of the aggregated channel and be able to dynamically change the ports group, i.e. Add or exclude ports from it. "Passive" ports of processing management frames LACP are not performed.

The IEEE 802.3ad standard is applicable to all types of Ethernet channels, and it can be built with even multi-bit lines consisting of several Gigabit Ethernet channels.

5.8. Maintenance quality assurance (QoS)

Personnel priority (802.1R)

The construction of switches based networks allows you to use traffic prioritization, and do it regardless of network technology. This feature is a consequence of the fact that the switches buffer the frames before sending them to another port.

The switch usually leads for each input and output port not one, but several queues, and each queue has its own processing priority. In this case, the switch can be configured, for example, so as to transmit one low-priority package for every 10 high-priority packets.

Priority processing support can be particularly useful for applications that have different requirements for admissible personnel delays and bandwidth Networks for streaming frames.

The ability of the network to provide various levels of service requested by those or other network applications can be classified in three different categories:

· Unregistered data delivery (BEST EFFORT SERVICE). Ensuring the connectivity of the network nodes without a warranty of time and the very fact of delivery of packages to the destination point. In fact, non-gear delivery is not part of QoS, since there is no service quality assurance and package delivery guarantee.

· Differentiated service (Differentiated Service). Differentiated service involves the separation of traffic on classes based on quality quality requirements. Each class of traffic is differentiated and processed by the network in accordance with the QoS mechanisms specified for this class (faster it is processed, above the average bandwidth, below the average loss level). Such a service quality assurance scheme is often called COS (Class of Service). Differential service in itself does not imply ensure the guarantees of the services provided. In accordance with this circuit, traffic is distributed by classes, each of which has its own priority. This type of service is convenient to apply in networks with intensive traffic. In this case, it is important to ensure the separation of network administrative traffic from everything else and assign a priority to it, allowing you to be confident in the connectivity of the network nodes at any time.

· Guarante service (Guaranteed Service). Guaranteed service involves reservation of network resources in order to meet specific requirements To maintenance from traffic flows. In accordance with the guaranteed service, there is a pre-reservation of network resources throughout the traffic trajectory. For example, such schemes are used in technology global networks Frame relay and ATM or in the RSVP protocol for TCP / IP networks. However, there are no such protocols for switches, so they cannot give guarantees.

The main issue in priority processing of frames switches is the issue of appointing priority frames. Since not all channel layer protocols support the frame priority field, for example, the Ethernet frames are missing, the switch must use any additional mechanism for binding a frame with its priority. The most common way is to attribute the priority of the switch ports. In this method, the switch places the frame in the framework of the personnel of the relevant priority, depending on which port entered the switch into the switch. The way simple, but not flexible - if it is not a separate assembly to the port of the switch, and the segment, then all segment nodes receive the same priority.

It is more flexible to assign priorities to frames in accordance with the IEEE 802.1. This standard was developed in conjunction with 802.1Q. In both standards, a total additional header for Ethernet frames consisting of two bytes is provided. In this additional header that is inserted in front of the frame data field, 3 bits are used to indicate the priority of the frame. There is a protocol for which the end node can request from the switch one of the eight levels of the frame priority. If the network adapter does not support the 802.1p standard, the switch may assign the priorities of frames based on the frame of the frame entering the frame. Such marked frames will be serviced in accordance with their priority by all network switches, and not just the switch, which directly accepted the frame from the end node. When passing frame network adapternot supporting the 802.1p standard, the additional header must be removed.

Switches provide differentiated service, so package identification is needed, which will allow them to be attributed to the corresponding COS traffic class, including, as a rule, packages from different streams. The specified task is performed by classification.

Packet Classification (Packet Classification) It is a tool that allows you to attribute a package to one or another class of traffic depending on the values \u200b\u200bof one or more fields of the package.

In managed switches are used various methods Package classifications. The following parameters are listed, on the basis of which the package is identified:

· 802.1p priority class bits;

· TOS byte fields located in the header of the IP package and the Differentiated Service Code field (DSCP);

· Appointment address and source of the IP package;

· TCP / UDP port numbers.

Since high-priority packages must be processed before low priority, multiple COS priority queues are supported in switches. Frames, in accordance with their priority, can be placed in different queues. Different service mechanisms can be used to process prioritization queues:

· Strict priority queue (STRICT Priority Queuing, SPQ);

· Weighted cyclic algorithm (Weighted Round Robin, WRR).

In the first case (SPQ algorithm), packets that are in the priority queue begin to be transmitted first. At the same time, the more priority will empty the packets, there will be no packets from less priority queues. The second algorithm (WRR) eliminates this restriction, and also eliminates the lack of bandwidth for the queues with a low priority. In this case, for each queue of priorities is set maximum amount Packages that can be transmitted at a time and the maximum wait time through which the queue will again be able to transmit packages. Range of transmitted packets: from 0 to 255. Revolutionar time range: from 0 to 255.

5.9. Network access restriction