SAN FRANCISCO -- VMware unveiled vSphere 5 today, introducing changes to nearly every aspect of its virtualization and cloud computing platform.
VSphere 5 features enhancements to the hypervisor itself,
“We need to make infrastructure become something that people can just depend upon and allow them to focus on the things that really make a difference for their businesses,” CEO Paul Maritz said. “Infrastructure is a means towards an end, and it needs to become fundamentally more automated.”
New vSphere 5 storage features
With a new feature called Profile-Driven Storage, vSphere 5 aims to put more storage-provisioning power into the virtualization administrator’s hands. This feature allows users to define tiers of storage that can span across different arrays, then associate these pools with service levels for both I/O performance and capacity availability.
“Before, the admin kind of had to guess what’s the best quality storage … and then they had to put their virtual machines on the right one,” said David Davis, director of infrastructure for Trainsignal Inc., an IT training company. “Now it’s more automated.”
With Profile-Driven Storage, administrators can assign profiles and service levels to a virtual machine (VM) that take storage -- not just CPU, RAM and the network -- into account, Davis explained.
VMware also overhauled Storage I/O Control in vSphere 5, adding new visibility into storage arrays with vStorage APIs for Storage Awareness (VASA). Storage I/O Control allows VMware administrators to set quality-of-service (QoS) priorities for VMs in the event of network or storage I/O contention.
In vSphere 4.1, Storage I/O Control wasn’t aware that the physical disks underlying two wide-striped LUNs were the same, resulting in latencies that it interpreted as contention -- a “major flaw,” according to Bob Plankers, a virtualization architect at a large Midwestern university.
“It [didn’t] know it [was] competing with itself,” Plankers said.
In vSphere 5, the new VASA application programming interfaces (APIs) allow vSphere to understand how the physical disk layout corresponds to logical data stores. The VASA APIs also enable another new feature: Storage Distributed Resource Scheduler (DRS), which allows for automated Storage vMotion of Virtual Machine Disk (VMDK) files around storage pools according to policies, guaranteeing storage bandwidth for the VM.
Storage DRS may overlap with automated tiered storage functionality from some storage subsystem vendors, but Plankers said the virtualization layer of the infrastructure offers a more intelligent way to optimize storage. Storage arrays have a block-centric view of things, limiting their effectiveness, he said.
”The higher you get in the stack, the more intelligent the decision can be,” he added.
In case it wasn’t clear that VMware intends to barge into the storage infrastructure market, the company also launched its own storage appliance, which allows the pooling of ESXi hosts’ internal disks as shared storage. This appliance competes directly with other offerings on the market, such as Hewlett-Packard’s P4000 Virtual Storage Appliance.
Shared storage is the underpinning for a number of advanced features within vSphere, so creating this low-end storage appliance is “essential to go into [a] smaller set of customers that haven’t used the high-end capabilities to date,” Maritz said.
VSphere Replication in Site Recovery Manager 5.0
With Tuesday’s announcement of Site Recovery Manager (SRM) 5.0, VMware also took disaster recovery into its own hands. SRM’s new vSphere Replication feature provides replication between sites at the VMDK level, which allows storage at each site to be heterogeneous -- and potentially less expensive than using matching arrays at each end of the wire.
SRM also introduces another long-awaited capability, automated failback. Previously, SRM could orchestrate the failover of a primary site to a secondary site in the event of a disaster, but when it came time to restore the primary site, users had to coordinate failbackon their own.
VMware HA becomes Fault Domain Manager
VMware also completely rewrote its High Availability (HA) feature, revamping legacy code that limited its scalability and how machines failed over, said Duncan Epping, principal architect for the company.
The previous version of HA was limited to a maximum of five primary nodes, responsible for powering on VMs on another host in the event of a failure, and it did not automatically elect a new primary node. So if something were to happen -- if, for example, the networking team rebooted a physical network device without telling the virtualization team -- VMs wouldn’t always restart.
“That’s one of the big sticking points for implementing HA for me,” Plankers said. “Someone screws up the network connectivity, and everything sort of kills itself.”
The new version, called Fault Domain Manager, operates using a “master-slave” concept. If the master host fails, the slaves detect that a failure has occurred and automatically elect a new master, which restarts the workload again. The new version can also use datastores as a secondary communication mechanism.
What else is new in vSphere 5?
As previously reported, VMs will be able to expand under vSphere 5, which will accommodate VMs with up to 1 TB of RAM and 32 virtual CPUs. In addition, a new auto-deploy feature -- which uses the PXE boot process to configure a VM image and deploy it to a host over a network -- will make life easier for some users.
“I've built PXE installation servers to deploy ESX and ESXi, and this requires many steps, scripting and troubleshooting,” said Shannon Snowden, consulting partner with New Age Technologies in Louisville, Ky., in an email. “I like the idea of having a plug-in and a more streamlined way to rapidly deploy ESXi.“
Other changes and related product updates in vSphere 5 include:
- Network I/O Control now allows for QoS at the VM level, rather than the host level.
- VMware introduced vShield 5, which can apply security and compliance policies to VMs according to the specific data they contain. If sensitive data such as credit card or Social Security numbers are discovered on a VM that’s not properly protected, that VM can be moved into a trust zone using vShield App.
- VMware announced that many of these new products, including vSphere 5, SRM 5.0, vCenter Operations Manager and the updated vCloud Director 1.5, will comprise its new Cloud Infrastructure Suite. VCloud Director 1.5 includes the ability to provision VMs from inside the self-service portal using Linked Clones, plus new APIs that allow for integration with third-party change management databases or IT service management software. In addition, vCloud Director 1.5 will support Microsoft SQL Server as well as Oracle for its back-end database.
Beth Pariseau is a senior news writer for SearchServerVirtualization.com. Write to her at email@example.com.