Reservations on Layer2 without STP
The STP protocol was coined by the standards of information technology in the Stone Age. At that time, the task of redundancy and security was already in place, with a sufficiently long convergence time, the STP protected the network from broadcast storms on the one hand (when looping) and provided some level of redundancy on the other. With the growth of the Dig Data effect and the maximum utilization of both computational and transmission capacities, STP began to grow into various “gadgets” and rose to MSTP, which was designed to maximally load all the existing links between equipment by building a tree for groups of vlans. But what to do if there is a lot of traffic going in vlan and bandwidth of one physical line is not enough?
This article will look at an example of the practical implementation of the Multi-Chassis Link Aggregation Group (MC LAG) view protocol on HP ProCurve hardware.
Distributed Trunking (DT) is an HP communication channel aggregation technology that allows you to use 2 pieces of equipment at one end of the aggregation, supported by 5400, 3500, 8200 series switches.
Next, we analyze the scheme and application of DT technology in it, as well as options for connecting various equipment to this technology. The example did not call for solving specific problems, but I think the ideas used in this scheme will help assess the advantages and disadvantages of the approach to building L2 networks.
General wiring diagram connection. Parts of which will be reviewed
Wiring switches access level switches to the kernel switches
')
The use of the STP family protocol in this scheme is not optimal. With a high load, in any of the vlans, the maximum speed will be limited by the speed of the channel to the kernel switch, in the specific case of 1 Gbit / s. If the amount of traffic is not constant, it is difficult to define groups for MSTP, having received at least some balance in the lines. We introduce DT technology for kernel switches. Switches ESC1 and ESC2 are combined into one logical switch, on access switches they aggregate lines to the core switch. As a result, the logic circuit L2 will have the following form.
Thus, the scheme does not suffer from loopback and provides 2 Gbit / s bandwidth to the core.
A little more about the configuration and functions of the ports.
Inter-Switch Connection (ISC) is responsible for transferring data between switches of a group, there may be either a separate port or a group of aggregated lines.
Peer-keepalive is responsible for monitoring the partner’s life, only service information and Hello packets are transmitted, traffic does not go.
Distributed Trunk Interface (DTI) is part of a common logical line of a virtual switch.
Due to the limited set of equipment on which this technology can be implemented, I will not describe its configuration, and there are enough resources with sufficiently detailed manuals, tyts .
For devices on the network that are connected to the DTI ports, a pair of switches is visible as one; this aspect allows connecting any equipment capable of aggregation.
I will cite a few examples, the easiest way with switches is that they almost all know how to aggregate up-links, and just combine a couple of ports directed to the core and that's it. With routers, everything is a little more complicated, not all of them know how to LAG (Link Aggregatin Group). In particular, for Cisco equipment, the 3925 router is able to use EtherCnannel, and 1841 is not able (without an additional switch module). At 3925 we combine two ports over EtherCnannel and that's it. But from 1841 it is necessary to be tricky, in practice a bunch of BeckUp interface + ip unnambered technologies was used to solve the problem. I suspect that it was possible to use it in the BackUp Interface + bridg group bundle, but was not tested in practice. Perhaps the community will offer other options.
The servers are still quite simple, for servers on Windows and server network cards most often there are drivers or utilities that allow you to combine network interfaces, for VMware there is NIC Teaming, etc.
In conclusion, I would like to say that this approach to building L2 networks has already been developed in the technology of large players; there is VSS and VPC from Cisco, DT and IRF from HP. Everything is moving towards the maximum utilization of all the capacity of the equipment park with a sufficient level of redundancy, and the technologies that do not meet these requirements go away.
This article will look at an example of the practical implementation of the Multi-Chassis Link Aggregation Group (MC LAG) view protocol on HP ProCurve hardware.
Distributed Trunking (DT) is an HP communication channel aggregation technology that allows you to use 2 pieces of equipment at one end of the aggregation, supported by 5400, 3500, 8200 series switches.
Next, we analyze the scheme and application of DT technology in it, as well as options for connecting various equipment to this technology. The example did not call for solving specific problems, but I think the ideas used in this scheme will help assess the advantages and disadvantages of the approach to building L2 networks.
General wiring diagram connection. Parts of which will be reviewed
Wiring switches access level switches to the kernel switches
')
The use of the STP family protocol in this scheme is not optimal. With a high load, in any of the vlans, the maximum speed will be limited by the speed of the channel to the kernel switch, in the specific case of 1 Gbit / s. If the amount of traffic is not constant, it is difficult to define groups for MSTP, having received at least some balance in the lines. We introduce DT technology for kernel switches. Switches ESC1 and ESC2 are combined into one logical switch, on access switches they aggregate lines to the core switch. As a result, the logic circuit L2 will have the following form.
Thus, the scheme does not suffer from loopback and provides 2 Gbit / s bandwidth to the core.
A little more about the configuration and functions of the ports.
Inter-Switch Connection (ISC) is responsible for transferring data between switches of a group, there may be either a separate port or a group of aggregated lines.
Peer-keepalive is responsible for monitoring the partner’s life, only service information and Hello packets are transmitted, traffic does not go.
Distributed Trunk Interface (DTI) is part of a common logical line of a virtual switch.
Due to the limited set of equipment on which this technology can be implemented, I will not describe its configuration, and there are enough resources with sufficiently detailed manuals, tyts .
For devices on the network that are connected to the DTI ports, a pair of switches is visible as one; this aspect allows connecting any equipment capable of aggregation.
I will cite a few examples, the easiest way with switches is that they almost all know how to aggregate up-links, and just combine a couple of ports directed to the core and that's it. With routers, everything is a little more complicated, not all of them know how to LAG (Link Aggregatin Group). In particular, for Cisco equipment, the 3925 router is able to use EtherCnannel, and 1841 is not able (without an additional switch module). At 3925 we combine two ports over EtherCnannel and that's it. But from 1841 it is necessary to be tricky, in practice a bunch of BeckUp interface + ip unnambered technologies was used to solve the problem. I suspect that it was possible to use it in the BackUp Interface + bridg group bundle, but was not tested in practice. Perhaps the community will offer other options.
The servers are still quite simple, for servers on Windows and server network cards most often there are drivers or utilities that allow you to combine network interfaces, for VMware there is NIC Teaming, etc.
In conclusion, I would like to say that this approach to building L2 networks has already been developed in the technology of large players; there is VSS and VPC from Cisco, DT and IRF from HP. Everything is moving towards the maximum utilization of all the capacity of the equipment park with a sufficient level of redundancy, and the technologies that do not meet these requirements go away.
Source: https://habr.com/ru/post/183208/
All Articles