Features of the architecture of Matrix N Series network devices.

logo11d 4 1

Features of the architecture of Matrix N series network devices.

Features of the architecture of Matrix N series network devices

Features of the architecture of Matrix N series network devices

Today it is difficult to surprise a sophisticated reader with some new network device. An attempt to read the announcement of a new product usually ends with the procedure of recalculating zeros in the column «performance», and if you manage not to get confused (in today's times, this figure can be very impressive), any normal user wonders what he will do with them, these zeros in such quantities. The «muddy wave» of delights in terms of ensuring information security that came from the Americans, as usual, can lead even an engineer knowledgeable in the subject into a state of stupor. In order to understand the functional difference in high-end devices from leading manufacturers such as Nortel Networks, Cisco Systems, Enterasys Networks, etc., you need to listen to a fair amount of in-depth courses on network technologies, devoting a significant part of your life to this, which is not available to everyone and not always possible.

Therefore, it seems reasonable to «go back to the roots» and talk about more accessible matters, at least at the level of organizational principles and basic logics.

A typical high-end network device, how is it organized? It is very similar to a freight train. A mandatory component of such a device is the control module. It does not matter what it is called — Supervisor, Control Module or something else. This is the locomotive that pulls the cars. A smart, powerful control module, it is the one that ensures the operation of the device, it looks into network packets, extracts and analyzes their contents, decides what to do with this packet: drop it or send it somewhere, reconstructing it accordingly.

Port or line modules — again, it doesn't matter what they're called — these are wagons. It doesn't matter what type of ports they have, their job is essentially to coordinate encodings and formats, to bring the received packets into a state suitable for transmission to the control module, which will «sort them out». After all, the wagon doesn't care what exactly is in it: coal or slippers, what's important is that each of them has a towing device that provides an interface to the locomotive, whose job it is to pull everyone who is attached to it.

Simple and clear, but not everything simple is brilliant. There is one «catch»: a locomotive, no matter how powerful it is, moves the slower, the more cars are attached to it. The more port modules are installed in the device, the smaller part of the control module's power falls on each port. Let's do some simple arithmetic. Let's take the data on the current control module of the flagship chassis X, a very well-known and respected manufacturer Y. Having broken through the marketing layers, we will record the IP routing performance of 90,000,000 (I'm simplifying the calculation procedure, 90 million) packets per second. On the same website, we will find information that this manufacturer X, without a doubt, offers 32-port gigabit cards. The last piece of information we need for our calculations is the theoretical maximum performance of a Gigabit Ethernet port. A 1000Base-X port operating according to standards is capable of producing no more than 1,488,000 packets per second. These figures can be found in any textbook on networking technologies. Now the arithmetic begins: we divide the performance of the control module 90 million by 32, i.e. we calculate the specific performance of the port, we get 2,812,500 packets per second. Everything is very good, all 32 ports operate at full wire speed. Now we install a second port module (i.e. we add another 32 ports), we get a total of 64 ports. Again, we divide the performance of the control module (90 million packets per second) by the number of ports (64), we get 1,406,250 packets per second.

Ideally, some alarm should go off, a big yellow light should blink, a siren should go off. We have already dropped below, just a little bit, but below the wire speed. But this is not happening, no network equipment manufacturer is interested in this and will not do it. The user, who does not suspect any trick, installs another (third) module, another 32 ports and the result is obvious: each port has 468,750 packets per second left. Alas! The locomotive has started to move much slower, the train is dragging along at a speed almost three times lower than the wire speed, each port has three times less of the control module's performance than it needs.

The picture is bleak… How to deal with such a problem? The solution is obvious: each car must be equipped with a locomotive. This locomotive must have the power necessary to efficiently service only its own car, i.e. it must not be so productive as to pull the entire train. But a train consisting of such pairs can be as long as desired, because each additional car is accompanied by an additional power for its maintenance.

Such a design of the network device assumes absolute equality of all chassis modules, each of them must be intelligent enough to efficiently service its ports and not seek advice from a «higher authority».

The new chassis from Enterasys Networks are made in exactly this way. The new series of devices, namely the three-slot N3 chassis and the seven-slot N7 chassis, assume the installation of modules that are a «car with a locomotive». In other words, each module is equipped with the necessary set of specialized microcircuits that provide routing at wire speed on all ports of the module. The chassis does not assume the use of any control modules, all modules are equal, each of them performs switching and routing without the help of any additional devices.

A reasonable user should be scared: a completely new solution built on unproven principles! And he would be right, but the point is that by the standards of Enterasys Networks, these devices represent the development of principles and technologies that have been developed since 1991! The first device built in accordance with the logic of «each carriage its own locomotive» was called Smart Switch 9000 and has proven itself well. Its descendants, the modular switches Matrix E5, Matrix E6, Matrix E7 are offered on the market today and provide reliable, productive operation of the networks of many large corporate customers.

The N3 and N7 chassis buses are matrices that connect modules together in accordance with the «each with each» principle, and are scalable to terabit performance.

The range of modules for these chassis covers all variations of gigabit and Fast Ethernet, as well as modules of the completely new 10 gig Ethernet standard (802.3ae).

Today, such a design is unique. No other company producing network devices is able to offer anything similar.

The range of application of these chassis is huge and includes all those industries and businesses that carefully and seriously treat the issues of performance and scalability of their network infrastructures.

Добавить комментарий

тринадцать + девятнадцать =

Мы используем cookie-файлы для наилучшего представления нашего сайта. Продолжая использовать этот сайт, вы соглашаетесь с использованием cookie-файлов.
Принять