Microsoft is adding new resource-monitoring capabilities directly into Hyper-V 3.0, but the extent of these improvements still remains a mystery.
More on resource monitoring
Virtual machine host monitoring improves VM performance
Virtualization storage performance monitoring: Metrics and tools
Virtualization performance and server resource management guide
Now that server virtualization is the norm, IT shops rarely allocate an entire physical server to a single workload. As such, organizations must determine the amount of physical resources that individual virtual machines (VMs) consume to evenly distribute workloads across a data center and accurately calculate chargeback.
Up until now, add-on products could only handle those tasks. For example, System Center Virtual Machine Manager (SCVMM) and the Self-Service Portal can allocate a pool of resources to an individual user, so that person can freely create VMs through a simple Web interface. A chargeback mechanism tracks both the allocated resources as well as the resources that are actually consumed, so that the user is billed accordingly. There are also a number of other, third-party products that provide similar functionality, such as Nicus Software Inc. IT Chargeback & Showback and VMware Inc. vCenter Chargeback.
But these products are generally expensive and somewhat complicated to implement. In Hyper-V 3.0, however, Microsoft will build some of this resource-monitoring functionality into the Windows Server 8 core operating system.
Resource monitoring still a mystery
Currently, the exact nature Hyper-V 3.0’s resource monitoring remains unclear. Microsoft has released a fair amount of information about what to expect, but these features don’t exist in the Windows Server 8 Developer Preview. So we will have to wait until the first beta to see the resource-monitoring features.
Using resource pools in Hyper-V 3.0
Hyper-V 3.0 resource monitoring will center around a new multi-tenant model, according to the information that I received from Microsoft. It will allow for the creation of resource pools, which will serve as collections of physical resources that users can allocate to VMs. In a multi-tenant environment, each customer is assigned a dedicated resource pool, so that there is no resource overlap between customers. (There is no word yet on how this configuration will affect memory overcommit.) You could use this multi-tenant model presumably for internal customers as well situations in which virtual server resources are charged back to individual departments.
You could easily think of Hyper-V 3.0's new multi-tenancy and resource-pooling features as mechanisms for ensuring privacy and security, but it is also extremely useful for resource monitoring. Often times, tenants will have a large collection of virtual machines. You can group these virtual machines into a single resource pool, which you can monitor as a whole, rather than attempting to track the resource usage for each VM.
Hyper-V 3.0 resource-monitoring metrics
So what types of metrics will be available in Hyper-V 3.0? While it is too early to know the full extent of the new capabilities, here is what we know for sure.
Microsoft will expose Hyper-V 3.0’s resource-monitoring functionality through a series of PowerShell cmdlets. Additionally, Microsoft will publish a series of new application programming interfaces, so developers can utilize the new resource-monitoring capabilities. But it is unclear if Microsoft will expose these resource-monitoring features through the Windows Server 8 or Hyper-V Manager graphical user interfaces.
Additionally, the following resource-pool metrics will be available (Note that Microsoft has not specified what “over a period of time” will entail):
- The average CPU use for virtual machines over a period of time
- The average memory use in megabytes for virtual machines over a period of time
- The minimum memory used for virtual machines over a period of time
- The maximum memory used by virtual machines over a period of time
- The maximum disk allocation for virtual machines over a period of time
- The total volume of inbound network traffic for each adapter over a period of time
- The total volume of outbound network traffic for each network adapter over a period of time
Microsoft has already stated it will not provide metrics for the following resources:
- Storage accessed through a virtual Fibre Channel adapter
- SCSI pass through disks
- Network adapters configured with OffloadWeight
While there are still a lot of unanswered questions surrounding these capabilities, it’s nice to see Microsoft is laying the groundwork for effective resource monitoring in Hyper-V 3.0.