Virtual Switches Demand Rethinking Connectivity with Servers
Technology Brief: Intel® Ethernet Server Adapters, Virtualization
With the creation of multiple virtual machines on each physical host server, network adapters now connect servers to the network with equivalency to a traditional top-of-rack switch.
The adoption of virtualization has led to a paradigm shift in the data center. At the core of these changes is the fact that a single host server running multiple virtual machines (VMs) can now represent the equivalent of multiple physical servers. Traffic among the VMs and with the outside world is now managed and coordinated by a virtual switch.
The virtual switch exists in software built into the hypervisor, meaning that it resides within the server itself, rather than as a separate physical device. Therefore, the network connectivity that was situated between the servers and a top-of-rack switch now occurs within the host server. By extension, the host server’s network adapters now handle the connections that formerly existed between the top-of-rack switches and core switches.
This shift can be further described as follows:
Yesterday’s data center: Physical edge switches. In conventional, non-virtualized data centers, servers were connected to top-of-rack switches by multiple Gigabit Ethernet (GbE) connections, and those switches connected to core switches with large trunked uplinks.
Read the full Rethinking Connectivity Technology Brief.
2688KB
Leider ist diese PDF-Datei nur im Download-Format verfügbar.
Virtual Switches Demand Rethinking Connectivity with Servers
Technology Brief: Intel® Ethernet Server Adapters, Virtualization
With the creation of multiple virtual machines on each physical host server, network adapters now connect servers to the network with equivalency to a traditional top-of-rack switch.
The adoption of virtualization has led to a paradigm shift in the data center. At the core of these changes is the fact that a single host server running multiple virtual machines (VMs) can now represent the equivalent of multiple physical servers. Traffic among the VMs and with the outside world is now managed and coordinated by a virtual switch.
The virtual switch exists in software built into the hypervisor, meaning that it resides within the server itself, rather than as a separate physical device. Therefore, the network connectivity that was situated between the servers and a top-of-rack switch now occurs within the host server. By extension, the host server’s network adapters now handle the connections that formerly existed between the top-of-rack switches and core switches.
This shift can be further described as follows:
Yesterday’s data center: Physical edge switches. In conventional, non-virtualized data centers, servers were connected to top-of-rack switches by multiple Gigabit Ethernet (GbE) connections, and those switches connected to core switches with large trunked uplinks.
Read the full Rethinking Connectivity Technology Brief.
