Speakers: Kenon Owens, Senior Product Marketing Manager, Microsoft and Fahad Ahmed, Infrastructure Architect, Microsoft.
This is a VMM 2012 session on building the private cloud fabrics. Or you could read Microsoft Private Cloud Computing to learn all this.
You create pools of physical resources, aka, clouds, give users access to them, define resources that they can use, and give them a quota. The physical resources in question are compute, storage, and network.
You can attach Configuration Manager to do additional management such as patching, DCM, auditing, compliance, security, etc.
- WinPE downloads and prepares a partition
- Downloads a VHD from VMM for boot to VHD
- Does Plug and Play for the system
- Boot the machine into OOBE
- Join domain and enable Hyper-V
- Reboot – and it’s now in a VMM host group
Uses SMI-S. Storage vendors still slow to implement.
- End to end mapping = create associations between storage and VM. ID storage consumed by VM, host and cluster
- Capacity management: add storage to a host or cluster through masking operations. Add capacity dring a new cluster creation
- Rapid provisioning: create new VMs leveraging SAN LUN cloning.
Can tier storage via classification pools using labels of your choice.
In Fabric: storage is a fabric. They’ve deployed 3 NetApp arrays via SMI-S providers. They have created 3 tiers of storage pools based on quality of disk, picked from the various arrays. In the VMM console, they create LUNs that will be used as a cluster witness disk and a CSV in a later cluster build.
Logical Abstraction for the network fabric
- Logical networks: Classify networks for VMs to access, map to network topology, allocate to hosts and clouds.
- Address pools: allocate static IP to VMs from a preconfigured pool, create and IP pool as a manage range of IPs, create a MAC address pool
- Load Balancers: apply settings for LB capability in service deployment, control LB through vendor provider
You can allocate logical networks to physical NICs, e.g. create Prod and DMZ networks, and allocate those logical networks to hosts in different clouds as appropriate.
- IP pools: assigned to VMs, hosts, and virtual IPs (LBs), specified use in VM template creation, checked out at VM creation, returned on VM deletion
- MAC pools: same as with IP pools, but for MAC assignment
- Virtual IP pools: assigned to service tiers in a service template that use a LB. Assigned to clouds. Checked in/out on creation/deletion. Reserved within IP pools
Supported LB: MSFT NLB, Citrix NetScaler, F5 Big IP, Brocade ServerIron ADX. Each requires a provider. Specify type of LB, e.g. round robin, etc.
Demo of Cluster Creation
2 Nodes, A and B. They are discovered in VMM, using a RunAs account. Creates a new cluster in Fabric. Adds the two hosts from the host group – must be in a single host group like in 2008/R2. Can optionally do the cluster validation tests (recommended). Assign a cluster IP. Now allocate the previous storage to the cluster. Checkbox to enable CSV in the disk selection. And that creates the cluster – some simplification of the networking story here.
Can manage Hyper-V, vSphere 4.1 (via vCenter only) and XenServer 6.0.
Demo of Cloud Creation
Create a cloud in VMs And Services. Select a host group or VMware resource pools. Select logical networks. Select LB VIP profiles. Any additional storage to allocate? You can set quota on CPU, RAM, storage, custom quota points, and VM number. You can control which type of hypervisor can be used.
Creates a new role in Settings/Security. Select from admin, read only, or self-service for a cloud. Select the clouds to assign to the role. You can override the previous quotas, e.g. for the role or for the entire role, as a subset of the cloud’s quotas. Add resources to the role, e.g. templates they can see. Then you specify actions they can do.
In App Controller, we see the delegated rights this role has, e.g. what they can deploy, how much, and what actions they can do. The Self-Service Portal is there only for backwards compatibility, it’s been deprecated.
Typically people make a few clouds, e.g. Prod and Test, and then use roles to divide up the shared pool of resources.
They aim to support vSphere 5.0 with System Center 2012 VMM SP1.