Today we would like to talk about two quite similar switch virtualization technologies, which allow to combine several switches into one logical one. We will talk about Cisco Virtual Switching System (VSS) and HPE Intelligent Resilient Framework (IRF) technologies. Within the framework of the article we will consider in more detail how the VSS technology works, after which we will talk about IRF technology.
Both technologies (VSS and IRF) allow us to combine switches using conventional Ethernet ports. In general, these technologies can be referred to as stacking technologies. But both vendors try to call them virtualization technologies. Cisco avoids the word stack in relation to VSS at all.
Cisco VSS
VSS technology allows you to combine two physical switches into one logical one. But unlike more classic stacking technologies (StackWise, FlexStack), Ethernet ports are used to connect switches to each other instead of specialized cables. Thus, switches can be located at a relatively large distance from each other.
Once merged, the switches work as a single logical one (Figure 1). Both switches are active and provide packet transmission. Both switches are managed by one of the devices. In other words, the data plane is active on both devices. But the control plane is only on one device. Let's remind that control plane (I suggest to use a foreign name in the future) is responsible for the logic of the switch operation: processing of all network protocols (L2/L3), formation of the routing table, filling of CEF, ACL, QoS tables, etc.
VSS technology is supported on the following Cisco switches:
Cisco 4500E and 4500X
Cisco 6500E and 6800
But you can't take any two of these devices and combine them using VSS technology. Firstly, not all supervisors, not all line cards and not all service modules support VSS technology. For example, for the 6500E VSS technology is supported on Sup720-10GE and Sup2T supervisors. Secondly, VSS technology works only between the same platforms, for example, between two 4500X or two 6500E+Sup2T. The packing of the devices (line cards and service modules) may differ. The chassis size may also differ for 4500E and 6500E switches. There are a lot of nuances, so it is highly desirable to see the current requirements for hardware, software versions and licenses on the vendor's website.
Once we have combined the two switches, the overall system performance doubles. This is due to the fact that both switches are responsible for packet processing. Thus, we get it:
- up to 1.6 Tbps for 4500X series switches,
- up to 1.8 Tbps for 4500E series switches with Sup 8-E supervisor
- and up to 4 Tbps for 6500E/6800 series switches with Sup2T supervisor.ttps://habrastorage.org/files/aab/5a5/036/aab5a503676c4854923e7ece1e5d2043.png
VSS architecture is shown in Figure 2. One of the switches is chosen as the main one, the second one as the backup one. On the main control plane switch it becomes active (Active), and on the backup switch it goes to the Hot Standby state.
The active control plane controls both switches. Also in the process of operation there is a constant synchronization of the state between the active control plane on the main switch and the control plane on the redundant switch to ensure fault tolerance. Management and synchronization are performed via a special channel - Virtual Switch Link (VSL).
The VSL channel is a direct connection between two switches (no intermediate devices are allowed). To provide a VSL link, the switches are connected to each other via regular Ethernet ports. But when I talk about regular ports, I'm a little bit sly. As you can guess, these ports also have certain requirements and these requirements vary depending on the switch platform. In addition to all the packets that are transmitted through the VSL channel, a special header is added - the Virtual Switch Header (its length is 32 bytes):
The VSL can consist of several physical channels (which is recommended). This is necessary for the fault tolerance of our system and for obtaining the necessary bandwidth. Aggregation is performed using PAgP or LACP protocols. I.e. we can have 8 active channels united into one logical VSL. For example, if we use 10 Gbit/s ports, we will get up to 80 Gbit/s when aggregating 8 such channels.