IRF - factory scaling technology
The massive use of server virtualization applications, the introduction of server blades and the use of multi-core server processors changed the requirements for data center networks and led to the emergence of new products, technologies and standards related to the construction of data centers.
For example, using VMware server virtualization technology, you can move virtual machines between different physical servers in real time using vMotion technology. To do this, the servers must be located in the same Layer 2 subnet, so for effective use of vMotion, you need to build a large network Layer 2 domain in a large data center.
Widely used in campus networks, the Spanning Tree Protocol (STP), like the modern versions of its Rapid Spanning Tree Protocol (RSTP) and Multiple Spanning Tree Protocol (MSTP) extensions, is poorly suited for this task because of:
The disadvantages of STP are to eliminate the new technologies IEEE 802.1aq SPB (Shortest Path Bridging) and IETF TRILL (TRansparent Interconnection of Lots of Links), however, until they receive the status of official standards.
')
HP Networking is already offering the possibility of building large Layer 2 subnets using IRF (Intelligent Resilient Framework) technology, which is based on 3Com's developments. IRF technology is used in both modular switches and switches with fixed configuration of A-series network ports, and (relatively recently) E-series stackable switches. Although this technology is based on a proprietary protocol, it is able to interoperate with other systems via standardized interfaces.
Using IRF, you can build a network fabric of nine switches and 1024 Gigabit Ethernet / 10 Gigabit Ethernet ports, combining access, aggregation and core levels, and these switches are managed by the same IP address (even if physically mounted in different racks and even in different data centers), which can scale up to 1024 ports. To centrally manage the entire factory as a large virtual switch, HP offers the Intelligent Management Center (IMC) console.
Console "Intelligent Management Center"
One of the switches of the IRF factory is the main one (primary) and is responsible for the control plane functions and updating the routing tables on the remaining switches of the factory, which provide Layer 2/3 functions for network-connected servers. If the main switch fails, the IRF automatically transfers primary functions to one of the other switches in the factory.
All switches in the IRF factory use the same configuration file and one software image that are copied from the main switch, so in fact, the network administrator needs to maintain only one switch. If another switch is connected to the IRF factory, then this configuration file and the software image will be automatically uploaded to it. IRF also supports the use of Link Aggregation Control Protocol (LACP), so that traffic between switches can be transmitted in parallel across several links, which reduces delays in packet transmission and provides protection against single link failure.
Topology convergence in case of failure of one switch or link in the IRF factory takes no more than 50 milliseconds, whereas using STP this procedure can take several seconds (this is important if critical applications run on virtual machines with maximum availability requirements). In addition, the In-Service Software Upgrade (ISSU) feature allows you to temporarily disable one switch to upgrade its software or maintenance without interrupting the normal operation of the entire factory.
When using 10 Gigabit Ethernet, this technology allows you to connect two data centers in one network factory, separated by up to 70 kilometers, and introduce virtual machines moving between data centers using vMotion technology to protect individual server applications from disasters.
So, the factory of HP Networking A and E series switches with IRF technology is an alternative to using STP when building a network infrastructure of a virtualized data center, providing more efficient scaling of a Layer 2 subnet and high fault tolerance, as well as simplifying its configuration and maintenance.
For example, using VMware server virtualization technology, you can move virtual machines between different physical servers in real time using vMotion technology. To do this, the servers must be located in the same Layer 2 subnet, so for effective use of vMotion, you need to build a large network Layer 2 domain in a large data center.
Widely used in campus networks, the Spanning Tree Protocol (STP), like the modern versions of its Rapid Spanning Tree Protocol (RSTP) and Multiple Spanning Tree Protocol (MSTP) extensions, is poorly suited for this task because of:
- relatively large delays in convergence of the network topology in case of failures of individual links or switches
- inefficient use of bandwidth (since there can be several physical paths between two switches, but due to the fact that some of the paths are blocked by STP, packets are transmitted over only one active connection)
- complexity of configuring switches and eliminating failures due to incorrect configuration parameters of one of the switches.
The disadvantages of STP are to eliminate the new technologies IEEE 802.1aq SPB (Shortest Path Bridging) and IETF TRILL (TRansparent Interconnection of Lots of Links), however, until they receive the status of official standards.
')
HP Networking is already offering the possibility of building large Layer 2 subnets using IRF (Intelligent Resilient Framework) technology, which is based on 3Com's developments. IRF technology is used in both modular switches and switches with fixed configuration of A-series network ports, and (relatively recently) E-series stackable switches. Although this technology is based on a proprietary protocol, it is able to interoperate with other systems via standardized interfaces.
Using IRF, you can build a network fabric of nine switches and 1024 Gigabit Ethernet / 10 Gigabit Ethernet ports, combining access, aggregation and core levels, and these switches are managed by the same IP address (even if physically mounted in different racks and even in different data centers), which can scale up to 1024 ports. To centrally manage the entire factory as a large virtual switch, HP offers the Intelligent Management Center (IMC) console.
Console "Intelligent Management Center"
One of the switches of the IRF factory is the main one (primary) and is responsible for the control plane functions and updating the routing tables on the remaining switches of the factory, which provide Layer 2/3 functions for network-connected servers. If the main switch fails, the IRF automatically transfers primary functions to one of the other switches in the factory.
All switches in the IRF factory use the same configuration file and one software image that are copied from the main switch, so in fact, the network administrator needs to maintain only one switch. If another switch is connected to the IRF factory, then this configuration file and the software image will be automatically uploaded to it. IRF also supports the use of Link Aggregation Control Protocol (LACP), so that traffic between switches can be transmitted in parallel across several links, which reduces delays in packet transmission and provides protection against single link failure.
Topology convergence in case of failure of one switch or link in the IRF factory takes no more than 50 milliseconds, whereas using STP this procedure can take several seconds (this is important if critical applications run on virtual machines with maximum availability requirements). In addition, the In-Service Software Upgrade (ISSU) feature allows you to temporarily disable one switch to upgrade its software or maintenance without interrupting the normal operation of the entire factory.
When using 10 Gigabit Ethernet, this technology allows you to connect two data centers in one network factory, separated by up to 70 kilometers, and introduce virtual machines moving between data centers using vMotion technology to protect individual server applications from disasters.
So, the factory of HP Networking A and E series switches with IRF technology is an alternative to using STP when building a network infrastructure of a virtualized data center, providing more efficient scaling of a Layer 2 subnet and high fault tolerance, as well as simplifying its configuration and maintenance.
Source: https://habr.com/ru/post/115501/
All Articles