Server virtualization increases efficiency, but it also places new demands on server input/output (I/O). Increased network and storage traffic can cause bottlenecks that impact the user experience. You can minimize bottlenecks by adding more network and storage connections, but at the cost of greater management complexity, higher capital expense cost, and larger servers.

Xsigo Delivers Virtual I/O and a Unified Fabric
Virtual I/O eliminates bottlenecks and reduces cost and complexity by consolidating the physical resources to a single high speed link. Within the server, the storage and network connections (HBAs and NICs) are replaced by virtual resources that appear exactly as their physical counterparts.

From the point of view of the hypervisor, virtual I/O behaves like traditional I/O. It appears within the hypervisor as conventional NICs and HBAs, and can be associated with virtual switches just as traditional I/O would. But with virtual I/O, connectivity can be deployed on the fly, migrated from server to server, and controlled with fine-grained bandwidth management tools.

Solution benefits include:

1. Predictable Application Performance
   It can be challenging to guarantee application performance and isolation when relying on shared connectivity. Virtual I/O delivers predictable application performance in three ways:
   • Dynamically allocated bandwidth: 20Gb bandwidth to each server is dynamically shared among virtual machines.
     
   • Resource isolation: Connectivity can be assigned to specific VMs and terminated on specific external ports on the Xsigo I/O Director. It is the equivalent of a dedicated link without the need for dedicated cabling.
     
   • Quality-of-service (QoS): Ensures that critical applications receive the bandwidth required. Control both network and storage bandwidth to specific virtual NICs and HBAs.
2. Lower Server Cost and Power Consumption
   Connectivity requirements for virtualization often drives deployment of larger, more costly servers. With virtual I/O, you can deploy all needed connectivity within a 1U server. Just two adapter cards lets you deploy up to 64 redundant vNICs and 32 vHBAs to any server.
   
   Fewer I/O cards also means less power consumption, saving as much as 40 watts per server.
   
3. I/O Resource Mobility
   Virtual I/O eases the challenge of building fault tolerant datacenter environments by allowing connectivity to be rapidly moved from one server to another. A server's I/O personality is defined by its connectivity, MAC addresses and WWNs. With Xsigo, you can move all of those attributes from one server to another in a single operation, effectively moving the server personality. This mobility helps enable:
   • Server replacement: When a server fails, its full connectivity profile can be instantly moved to another server.
     
   • Isolation: To ensure physical isolation, a server may be configured only with the connectivity currently needed, and no more. When requirements change, the configuration can be changed without entering the data center. The IT manager is never forced to implement "big flat networks" or open storage.
     
   • Disaster recovery: The exact I/O resources at one site can quickly be created at another, thus accelerating switchover.
4. Development to Staging to Production, Without Disruption
   Move a server or specific VM from one physical network to another without re-cabling. Simply direct the server's traffic to another external port on the Xsigo I/O Director. Get network isolation without relying on VLANs.
   
5. Faster VMotion
   Accelerate VMotion with a "dedicated" high-speed link. Xsigo virtual I/O incorporates a low-latency switched fabric that enables exceptionally fast server-to-server communications. Within that fabric it is simple to configure an isolated VMotion network that is isolated from production traffic.
   
6. Less Latency
   Server-to-server communications benefit from the high-speed and low-latency switching of Xsigo's underlying fabric. By sending information from one vNIC to another, servers move traffic at nearly 10Gb per second with a latency of less than 30 microseconds using standard TCP/IP protocols.
   
7. Scalable I/O for Blades
   Escape the I/O constraints of blade systems. With virtual I/O, storage resources can be deployed as needed, up to 32 virtual NICs and 32 virtual HBAs per blade. Each hypervisor may be configured with up to 32 vNICs and 12 vHBAs, so you can better support VMware best practices of dedicated connectivity for critical VMs and management functions.
   
8. Centralized I/O Management
   Manage I/O from within the VMware user interface. The Xsigo graphical user interface is a browser based interface that provides centralized management of all I/O connections, system-wide, from a single screen. In addition to running as a stand alone system, the Xsigo Management System has been integrated with the VMware Infrastructure Client. This provides complete management of virtual machines and virtual I/O from a single application.
   

The Xsigo I/O Director

The Xsigo I/O Director is an enterprise-class, hardware-based solution that provides LAN and SAN connectivity for up to hundreds of servers. Under software control, any server may be configured with access to any LAN or SAN. Because the vNICs and vHBAs can be configured on-the-fly without a server reboot, the server's network connectivity is truly flexible and reconfigurable.