Blog

Bringing Web-scale Infrastructure to Hyper-V Environments

By Manish Pandey

The Nutanix Virtual Computing Platform is a converged infrastructure solution that consolidates the compute (server) tier and the storage tier into a single, integrated appliance. The platform delivers web-scale performance, scalability, and data management to VMware ESXi and KVM based virtualized systems. Nutanix, with the recent NOS 3.5.2 release, now brings these benefits to virtualized infrastructure running Windows Server 2012 R2 with Hyper-V on Windows Server 2012 R2 with Hyper-V SMB 3.0 storage. With this addition, Nutanix provides its customers the broadest choice of hypervisors among all converged platforms. With this choice, customers can run different applications and workloads on different hypervisors and make economically rational choices for their hypervisor, OS and application software licensing strategy.

SMB or Server Message Block was originally designed for providing access to file shares and other resources over the network. In the early days of the PC, IBM developed NetBIOS, a simple networking system for small LANs. Microsoft added the ability to allow disk I/O over NetBIOS interface, which made disk space sharable over the LAN. This file-sharing protocol became known as SMB or CIFS. SMB 3.0, released in 2012, has come a long ways from this humble origin, and was designed to support file server storage for Hyper-V and applications such as SQL server. It is aimed at running high-density virtualized workloads, and provides highly available storage through transparent client failover, and scales performance by including additional fault-tolerant network connections.

SMB 3.0 is designed to support workloads with a variety of traditional storage technologies ranging from JBODs to expensive SAN or FC solutions. However, the fundamental bottleneck of accessing traditional storage over network makes datastore scaling a hard problem. There are multiple hurdles for traditional solutions — overhead of networking protocols, shared common bandwidth among the potentially hundreds of VMs accessing the datastore, high cost, incremental upgradability, to name a few. If you do not have unlimited IT budget, and unlimited time to deploy and manage storage systems, these hurdles presents an insurmountable barrier for traditional storage. Traditional storage vendors are touting expensive interconnect fabric and flash storage trays. However, these serve as little more than ineffective bands aids, given the fundamental trends of data access and storage needs for virtualization, big-data and cloud.

So, why does Nutanix shine here? The Nutanix Virtual Computing Platform eliminates the server-storage interconnect bottleneck by providing a converged solution that creates a single unified storage pool from server-side flash and locally attached spindles. This fundamental architectural advantage holds across all hypervisors, and with NOS 3.5.2, we support Hyper-V hypervisor on Nutanix clusters. Each node in the cluster runs Hyper-V, and a Nutanix controller VM handles all I/O operations for the local hypervisor. This provides a highly available, resilient and performant software defined storage, that is now available as an SMB 3.0 share (in addition to previously supported NFS and iSCSI interfaces). Each of the nodes below is a server running Hyper-V, which accesses the SMB 3.0 storage to provide the virtualized storage to the user virtual machines (UVMs) (Figure 1).

Two-way mirroring

Figure 1: Nutanix Hyper-V Cluster with SMB 3.0

Enterprises can run large-scale server virtualization deployments, virtual desktop initiatives, test and development applications, and big data applications such as Hadoop and Splunk on Hyper-V, while eliminating network-based storage. The scalability and performance of the world’s largest, most efficient web-scale data centers is available to customers running Nutanix clusters with Hyper-V.

This solution with SMB 3.0 available on the Virtual Compute Platform, provides customers a number of capabilities such as transparent failover of client access, offloaded data transfer (ODX) and TRIM.

Transparent failover of client access allows UVMs on a Hyper-V host to stay connected to storage and working, even though the server they were talking to fails. The failover capability, allows the Hyper-V server to recover and continue talking to different SMB server in a cluster if needed (Figure 2).

Figure 2: Transparent Failover

ODX support speeds up copy and move operations by offloading the entire operation to the storage system which orchestrates the reads and writes without reading data onto the server, buffering it and writing it back. Besides speed up, it cuts down on server CPU cycles, and results in a large server-client network traffic reduction. This feature is available to Nutanix Hyper-V clusters.

TRIM support is another feature available to Nutanix Hyper-V clusters. This feature allows the operating system to communicate the inactive block IDs to the storage system. The storage system can wipe these unused blocks internally. Trim support enables efficient garbage collection in SSDs and minimizes write amplification and flash cell wear.

In the coming weeks and months we will have additional blogs on Nutanix Hyper-V systems. Watch out here.