Comparison Edge-Core ECS4100-28T vs Juniper EX3300-24T

The bandwidth of a switch is the maximum amount of traffic that it can handle. Specified in gigabits per second.

This parameter directly depends on the number of network ports in the device (excluding Uplink). Actually, even if the bandwidth is not given in the specifications, it can still be calculated using the following formula: the number of ports multiplied by the bandwidth of an individual port and multiplied by two (since both incoming and outgoing traffic are taken into account). For example, a model with 8 Gigabit Ethernet connectors and 2 SFP ports will have a bandwidth of (8*1 + 2*1)*2 = 20 Gbps.

The choice for this indicator is quite obvious: you need to evaluate the expected traffic volumes in the serviced network segment and make sure that the switch's bandwidth will cover it with a margin of at least 10-15% (this will give an additional guarantee in case of emergency situations). At the same time, if you plan to often work at high, close to maximum, loads, it will not hurt to clarify such a characteristic as the internal bandwidth of the switch. It is usually given in a detailed technical description, and if this value is less than the total throughput, serious problems may arise under significant loads.

The maximum number of MAC addresses that can be stored in the Switch's memory at the same time. Specified in thousands, for example, 8K — 8K.

Recall that the MAC address is the unique address of each individual network device used in physical routing (at layer 2 of the OSI network model). Switches of all types work with such addresses. And it is worth choosing a switch according to the size of the table, taking into account the maximum number of devices that are supposed to be used with it (including based on the possible expansion of the network). If the table is not enough, the switch will overwrite new addresses over the old ones, which can noticeably slow down the work.

The number of optical network ports of the SFP standard provided in the design of the switch. We emphasize that we are talking about "ordinary" SFPs; SFP+ data is usually listed separately.

Specifically, in switches, the marking “SFP” usually means a connector for fiber with a connection speed of 1 Gbps. Technically, this is not much compared to RJ-45 speeds; however, this connection format has a number of advantages. One of the main ones is a greater effective range: the mentioned gigabit standard used in switches works with a cable length of up to 550 m, and by the standards of fiber, this is still very little. True, the cable itself is sensitive to kinks and requires quite delicate handling; on the other hand, it is completely immune to electromagnetic interference. On the other hand, in general, the SFP format is noticeably less popular in network equipment than RJ-45; therefore, there are few ports of this type even in advanced devices. So, solutions for 2 or 4 SFP connectors are most widely used, although there are more - 6, 8, or even 10 or more. It is also worth considering that the so-called combo connectors can be used in switches, combining SFP and RJ-45; the presence of such ports is specified in the notes, they are taken into account both in the calculation of RJ-45 and in the calculation of SFP.

To clarify, Uplink inputs also often use this type of connector; however, their...number is specified separately (see below).

The number of optical SFP+ ports provided in the design of the switch. Let's clarify right away that we are talking about ordinary network ports; Uplink inputs can also use this interface, however their number is specified separately even in this case (see below).

The general advantages of optical fiber over conventional Ethernet cable are longer communication range and insensitivity to electromagnetic interference. Specifically, SFP+ is a development of the original SFP standard; in switches, such connectors typically operate at a speed of 10 Gbps. As for the number of such ports, for all its advantages, fiber optics in network equipment is still used quite rarely. Therefore, the most common switches are 1 - 2, less often 4 SFP + connectors, although there are more. It is also worth considering that the so-called combo connectors can be used in switches, combining SFP + and RJ-45; the presence of such ports is specified in the notes, they are taken into account both in the calculation of RJ-45 and in the calculation of SFP+.

The type of connector(s) used in the switch as an Uplink interface.

See above for details on such an interface; here we note that the same network ports are usually used as Uplink, as for connecting individual devices to the switch. Here are the main options for such connectors:

- Fast Ethernet - LAN network connector (for "twisted pair") with support for speeds up to 100 Mbit. Such a speed is considered low by modern standards, while the Uplink port puts forward increased bandwidth requirements - after all, traffic from all devices served by the switch goes through it. Therefore, in this role, Fast Ethernet ports are used mainly in inexpensive and outdated models.

- Gigabit Ethernet - LAN connector with support for speeds up to 1 Gb / s. Such a speed is often enough even for a fairly extensive network, while the connectors themselves are relatively inexpensive.

- 2.5 Gigabit Ethernet - LAN connector with support for speeds up to 2.5 Gbps.

- 10Gigabit Ethernet - LAN connector with support for speeds up to 10 Gbps. Such features allow you to work comfortably even with very large volumes of traffic, but they significantly affect the price of the switch. Therefore, this option is rare, mainly in high-end models.

— SFP. Socket for fiber optic cable that supports speeds of about 1 Gb / s. At the same time, over Gigabit Ethernet, which has a similar bandwidth, this connector has one noticeable advantage - a lon...ger connection range (usually up to 550 m).

- SFP+. An evolution of the SFP standard described above. The switches usually provide a connection speed of 10 Gb / s; like the original standard, it noticeably outperforms an Ethernet connection in terms of effective range. On the other hand, the real need for such speeds does not arise very often, and SFP+ is quite expensive. Therefore, the presence of such Uplink connectors is typical mainly for high-end models with a large number of ports.

— SFP28. Another development of SFP with increased throughput up to 25 Gbps.

— QSFP / QSFP+. The fastest SFPs up to 40 Gbps.

We also note that the connectors described above (except perhaps Fast Ethernet) are rarely used as the only type of Uplink input. Combinations of electrical and fiber optic ports - SFP / Gigabit Ethernet and SFP + / 10Gigabit Ethernet - have become noticeably more common. This provides versatility in connection, allowing you to use the type of cable that is most convenient in a given situation; and if necessary, of course, you can use all Uplink inputs at once. However, it is worth considering that in some models, Ethernet and SFP interfaces can be combined in one physical connector. So before buying this nuance does not hurt to clarify separately.

There are also switches that use a combination of two types of SFP - SFP/SFP+; however, there are few such models and they mainly belong to the professional level.

— DHCP server. A feature that makes it easy to manage the IP addresses of devices connected to the switch. Without its own IP address, the correct operation of the network device is impossible; and DHCP support allows you to assign these addresses both manually and fully automatically. At the same time, the administrator can set additional parameters for the automatic mode (range of addresses, maximum time for using one address). And even in fully manual mode, work with addresses is performed only by means of the switch itself (whereas without DHCP, these parameters would also have to be specified in the settings of each device on the network).

— Stacking support. The ability to operate the device in stack mode. A stack consists of several switches that are perceived by the network as one “switch”, with one MAC address, one IP address, and with a total number of connectors equal to the total number of ports in all involved devices. This feature is useful if you want to build an extensive network that lacks the capabilities of a single switch, but do not want to complicate the topology.

— Link Aggregation. Switch support for link aggregation technology. This technology allows you to combine several parallel physical communication channels into one logical one, which increases the speed and reliability of the connection. Simply put, a switch with such a fun...ction can be connected to another device (for example, a router) not with one cable, but with two or even more at once. The increase in speed in this case occurs due to the summation of the throughput of all physical channels; however, the total speed may be less than the sum of the speeds — on the other hand, combining several relatively slow connectors is often cheaper than using equipment with a more advanced single interface. And the increase in reliability is carried out, firstly, by distributing the total load over individual physical channels, and secondly, by means of "hot" redundancy: the failure of one port or cable can reduce the speed, but does not lead to a complete disconnection, and when the channel is restored, the channel is switched on automatically.
Note that both the standard LACP protocol and non-standard proprietary technologies can be used for Link Aggregation (the latter is typical, for example, for Cisco switches). In addition, there are quite a few alternative names for this technology — port trunking, link bundling, etc.; sometimes the difference is only in the name, sometimes there are technical nuances. All these details should be clarified separately.

— VLAN. Support of the VLAN function by the switch — virtual local area networks. In this case, the meaning of this function is the ability to create separate logical (virtual) local networks within the physical "local area". Thus, it is possible, for example, to separate departments in a large organization, creating for each of them its own local network. The organization of VLAN allows you to reduce the load on network equipment, as well as increase the degree of data protection.

— Protection against loops. The switch has a loop protection function. The loop in this case can be described as a situation where the same signal is launched in the network in an endless loop. This may be due to incorrect cable connection, the use of redundant links and some other reasons, but anyway, such a phenomenon can “put down” the network, which means it is highly undesirable. Security prevents loops, usually by disabling looped ports.

— Limiting the speed of access. The ability to limit the data exchange rate for individual switch ports. Thus, it is possible to reduce the load on the network and prevent the "clogging" of the channel by individual terminals.

Note that the matter is not limited to this list: other features may be found in modern switches.

Recall that routing is the definition of the best path through which each data packet can be delivered to the recipient. For this, special tables are used, stored in the memory of the control network device with the routing function. According to the method of filling these tables, this procedure is divided into two main varieties — static and dynamic.

Static routing is a method in which all data routes (entries in the routing table) are manually written by the administrator; this applies both to the initial creation of the table and to making changes to it when changes are made to the network configuration. The main advantage of this method is the minimum load on the switch processor, which has a positive effect on the speed and reliability of the network. The main disadvantages of static routing are associated with the need for manual control. So, the larger the network, the more complex and time-consuming it is to manage it; Administrator's inattention can become an additional cause of failures; and diagnosing some problems is noticeably more difficult — for example, if there is a failure at the link layer, the static route remains visible as active, although no data is transmitted.

Static routing is carried out according to the standard scheme, but different protocols are used for dynamic routing. The idea of dynamic is that the route table is constantly edited programmatically, in automatic mode. To do this, network devices (more precisely, routing programs running on them) exchange service information with each other, on the basis of which optimal addresses are written to the table. One of the fundamental concepts of dynamic routing is a metric — a complex indicator that determines the conditional distance to a specific address (in other words, how close this or that route is to the optimal one). Different protocols use different ways to define and share metrics; here are some of the most common options:

— R.I.P. One of the most widely used dynamic routing protocols; was first applied back in 1969 on the ARPANET, which became the forerunner of the modern Internet. Refers to the so-called distance-vector algorithms: the metric in the RIP protocol is indicated by the distance vector between the router and the network node, and each such vector includes information about the direction of data transfer and the number of "hops" (sections between intermediate nodes) to the corresponding network device. When using RIP, metrics are sent over the network every 30 seconds; at the same time, having received from the "neighbor" data about the nodes known to it, the router makes a number of clarifications and add...itions to this data (in particular, information about itself and about directly connected network devices) and transmits further. After receiving up-to-date data throughout the network, the router selects for each individual node the shortest route from several received alternatives and writes it into the routing table.
The advantages of the RIP protocol include ease of implementation and undemanding. On the other hand, it is poorly suited for large networks: the maximum number of hops in RIP is limited to 15, and the complication of the topology leads to a significant increase in service traffic and the load on the computing part of the equipment — as a result, the actual network performance decreases. Thus, more advanced protocols such as (E)IGRP and OSPF (see below) have become more common for professional applications.

— IGRP. A proprietary routing protocol created by Cisco for autonomous systems (in other words, local networks with a single routing policy with the Internet). Also, like RIP (see above), it refers to distance vector protocols, however, it uses a much more complicated procedure for determining the metric: it takes into account not only the number of hops, but also delay, throughput, actual network congestion, etc. In addition, the protocol implements a number of specific mechanisms to improve communication reliability. Due to this, IGRP is well suited even for fairly complex networks with an extensive topology.

— EIGRP. An improved and modernized successor to the IGRP protocol described above, developed by the same Cisco. Created as an alternative to OSPF (see below), it combines the properties of distance vector protocols and standards with link state tracking. One of the main advantages over the original IGRP was the improvement in the algorithm for disseminating data about changes in the topology in the network, due to which the probability of looping (characteristic of all distance vector standards) was reduced to almost zero. And among the differences between this protocol and OSPF, higher performance and a more advanced algorithm for calculating the metrics are claimed with less configuration complexity and resource requirements.

OSPF. An open autonomous system routing protocol created by the IETF (Internet Design Council) and first implemented in 1988. Refers to protocols with link state tracking, uses the so-called Dijkstra algorithm (algorithm for finding the shortest paths) to build routes. The OSPF routing process is as follows. Initially, the router communicates with similar devices, establishing a "neighbor relationship"; neighbors are routers within the same autonomous zone. Then the neighbors exchange metrics among themselves, synchronizing the data, and after such synchronization, all routers receive a complete database of the state of all links in the network (LSDB). Already on the basis of this base, each of these devices builds its own route table using Dijkstra's algorithm. The main advantages of OSPF are high speed (speed of convergence), a high degree of optimization of the use of channels and the ability to work with network masks of variable length (which, in particular, is especially convenient with a limited resource of IP addresses). The disadvantages include the exactingness of the computing resources of routers, a significant increase in load with numerous such devices in the network, and the need to complicate the topology in large networks, dividing such networks into separate zones (area). In addition, OSPF does not have clear criteria for determining the metric: the “cost” of each hop can be calculated according to different parameters, depending on the switch manufacturer and the settings chosen by the administrator. This expands the possibilities for configuring routing and at the same time greatly complicates this procedure.

Modern switches may provide other routing protocols in addition to those described above.

— Built-in. The built-in power supply does not take up space on the outside, but can significantly increase the size and weight of the entire switch. Because of this, this option is quite rare — mainly among rack-mount models (see "Form factor"), where an external unit can create significant inconvenience, as well as among the most powerful desktop switches, for which restrictions on dimensions and weight is not critical.

— External. Theoretically , an external power supply requires additional space, and therefore is not as convenient as an internal one. In fact, most blocks of this type are quite compact in size and are equipped with “plugs” for sockets right on the case — in other words, the block is installed on a socket, and from there the wire stretches to the switch. And the absence of power circuits and transformers inside the case has a positive effect on compactness. Thanks to all this, this option is very popular among desktop models (see "Form factor"), primarily entry-level and mid-level.

— No BP. The absence of a power supply both in the design and in the delivery set is a rather rare case found in three types of switches. The first variety is models that use PoE power (see above) and do not require separate power sources. PoE power is relatively small, so relatively simple devices with a small number of ports fall into this category. The second variety is professional switc...hes, the power supplies for which are sold as separately installed internal modules; such equipment may even provide the possibility of using two PSUs simultaneously (main and backup) and hot-swapping them. The third type — switches with installation on a DIN rail (see "Form factor") and having terminals for connecting a specialized external power source.