More powerful VMs and better networking capabilities
1. VMs with 32 vCPUs—Previous versions of Hyper-V were limited to four virtual CPUs (vCPUs) per VM, which was adequate for smaller servers but obviously wasn't able to support workloads with very high scalability requirements. Hyper-V 3.0 VMs will support up to 32 vCPUs, which is adequate for almost all workloads.
2. VMs with 512GB of RAM—Another huge advance in VM scalability is support for up to 512GB of RAM per VM. This is a big increase over previous Hyper-V VMs, which were limited to 32GB of RAM. VMs that support 32 vCPUs and 512GB of RAM provide scalability comparable to vSphere 5.1.
3. 16TB VHDX format—Another scalability enhancement that Microsoft has added to Hyper-V 3.0 is support for a new Virtual Hard Disk (VHD) format called VHDX. The VHDX format provides support for VHDs with up to 16TB of storage. Previous versions of the VHD format were limited to 2TB.
4. Multiple concurrent Live Migrations and Storage Live Migrations—Live Migration was added to Hyper-V 2.0, which was introduced with Windows Server 2008 R2. It let you move VMs between hosts in the same cluster with no downtime. However, it was limited to performing a single Live Migration per occurrence, and migrating VM storage still required downtime. Hyper-V 3.0 addresses both of these problems by allowing for unlimited multiple concurrent Live Migrations as well as introducing Storage Live Migrations.
5. Hyper-V Replica—A completely new feature in Hyper-V 3.0, Hyper-V Replica is a disaster-recovery solution for VMs that doesn't require an expensive SAN. Hyper-V Replica provides asynchronous replication of VMs from a primary site to a backup site without the need for specialized storage or networking hardware. In the event of a failure at the primary site, the administrator can fail over to the replica VMs at the disaster-recovery site.
6. Extensible virtual switch—The virtual networking capabilities provided by Hyper-V 3.0 have been significantly enhanced. Microsoft's new virtual switch can provide minimum and maximum bandwidth guarantees. The virtual switch is also extensible through an API that allows capture, filtering, and forwarding extensions to be added to the switch. Microsoft will subject all Hyper-V virtual switch extensions to testing through a new logo program.
7. 63 node clusters—Although this isn't strictly a virtualization feature, Hyper-V 3.0 can take advantage of Windows Server 8's vastly expanded clustering capabilities. Earlier versions of Windows Server Failover Clustering were limited to 16 nodes. Windows Server 8 supports clusters with up to 63 nodes and up to 4,000 VMs per cluster. Cluster nodes can be physical systems or they can be VMs. Clusters of VM guests can span multiple virtualization hosts.
8. Support for native NIC teaming—Another important underlying enhancement to Windows Server 8 that Hyper-V 3.0 can leverage is Windows Server 8's native support for NIC teaming. NIC teaming was possible in earlier versions of Windows Server. However, you needed specialized NICs from either Intel or Broadcom. Windows Server 8's NIC teaming works over all types of heterogeneous NICs and can be used by Hyper-V 3.0 virtual networks.
9. Affinity and anti-affinity rules—With Hyper-V 3.0, administrators can use affinity and anti-affinity rules to control when multiple VMs should fail over together or to prevent specific VMs from running simultaneously on the same virtualization host. Affinity and anti-affinity are configured by setting the cluster service properties.
10. Hyper-V client—One surprising move was support for Hyper-V on the desktop. Hyper-V 3.0 will be built in to the client version of Windows 8. Unlike the older Microsoft Virtual PC, Hyper-V on the client will run directly on the system hardware. And unlike Hyper-V 2.0, it will also support power-management functions such as suspend and hibernate.