Shared-nothing live migration: What it means to virtualisation admins

Shared-nothing live migration: What it means to virtualisation admins

Daniel Eason

Shared-nothing live migration is a storage capability that will be included in two upcoming virtualisation platforms -- Microsoft Windows Server 2012 Hyper-V and VMware vSphere 5.1. While it may not dramatically alter a virtual infrastructure, IT pros must not dismiss it because it brings the much-needed virtual machine mobility and storage mobility.

Shared-nothing live migration allows admins to migrate a virtual machine (VM) from one physical server with direct-attached storage to another physical server with direct-attached storage.

It requires no clustering or shared storage between both servers because just a small connection such as a Gigabit Ethernet (GbE) cable will suffice resulting in zero down time.

When Microsoft announced that it will release Windows Server 2012 Hyper-V in October 2012 with the shared-nothing live migration functionality, experts gave a thumbs up to the new feature. Virtual machine mobility and storage mobility are very interesting features that provide for a previously unknown flexibility, they said.

Hyper-V 3.0 shared-nothing live migration
For virtualisation admins wondering what’s special about a feature that allows live migration of VMs between hosts, the Hyper-V 3.0 shared-nothing live migration differs significantly from other forms of migration techniques. Hitherto, live migration capability was limited to VM migration between only virtual hosts with VMs on a volume that is hosted on common shared storage subsystems, and this meant that live migration in most scenarios was limited to just the hosts in the same local data centre.

Even that limited level of migration capability provides an unfounded level of flexibility to data centre administrators, but in order to use it, they must purchase the expensive clustered infrastructure and networking tools.

Shared-nothing live migration will offer live migration of VMs that are hosted on inexpensive local disk SMB v3.0 volumes, ie, “\\servername\volumename”. Virtualisation admins can migrate files including VM virtual disk file, memory state and any configuration that is present within the VM folder. Such live migration is a seamless activity and does not affect any services running within the VM environment.

It still has the same level of discipline for compatibility rules across the host hardware as standard Microsoft Live Migration -- meaning IT pros must still need to have compatibility between Hyper-V 2012 hosts for key components such as processor family, network speeds. In addition, it must be in the same AD domain and have 1 GB networking capability.

Storage-nothing live migration will be a third type of migration capability within Windows Server 2012 Hyper-V and can be viewed as a “bronze” level of capability compared to others. Other techniques include Live Migration and Storage Live Migration that are comparable to VMware Storage vMotion.

Don’t be fooled into thinking that shared-nothing live migration will be a waste of time and an unnecessary feature that you won’t exploit.

Use cases for shared-nothing live migration
Shared-nothing live migration will most certainly hit a sweet spot for cash-strapped SMB shops that have always been limited to running VMs upon direct-attached storage or local disk storage. Small and medium-sized enterprises will most likely have the prerequisites to satisfy the feature’s criteria, including the 1 GB networking connectivity that has become available at a much lower price with the availability of 10 GB networking in the industry.

For IT pros planning a deployment, the feature gives them more flexibility and capabilities that previously didn’t exist.

As the pressure on IT department to deliver more for less intensifies, shared-nothing live migration  will allow SMB IT admins to think like a large enterprise when it comes to migration and availability of VMs.

While shared-nothing live migration has particular benefits for SMBs, IT pros at large enterprises should not dismiss it. Regardless of the size of the organisation, the functionality is useful in the event of planned outages. Using shared-nothing means IT does not have to use reactive features such as the standard live migration functions and can avoid potential impacts that the migration may have on the virtual infrastructure and IT services.

The feature will also help the IT staff obtain an instant migration capability between clustered nodes and avoid network performance and timing issues involved in migrating VMs reactively.

Technical considerations
While shared-nothing live migration functionality is recommended for planned outage scenarios, it is important to emphasise that it should not be used for high availability or immediate clustering needs because of potentially large VM disk sizing in these scenarios.

The new VHDX file format in Hyper-V 3.0 means the largest possible supported virtual hard disk size will be a total of 64 TB, and migrating a file of such size isn’t practical.

As virtualisation grows, the size of virtual hard disks will also increase and IT should consider using the shared-nothing live migration feature to migrate smaller VMs, not VMs with large connected secondary disk volumes such as Exchange or SQL databases.

Even for migration of smaller sized VMs, IT could experience some impact on network performance, so admins should consider using a dedicated network connection or a crossover cable link as the migration network between Hyper-V hosts. 

Admins should also avoid using the same primary VM network for migrations because the sheer volume of traffic could create bottleneck havoc. Another way to avoid bottlenecks is to bundle network interface card interfaces on virtual hosts so that there is more network throughput between hosts.

Lastly, in terms of the overall management of Hyper-V virtual machines that need to be migrated by shared-nothing live migration, IT should perform a review of the size of the VMs targeted for migration and calculate the potential time this will take against the outage period they will have for the underlying server hardware.

IT admins should consider breaking the time down to calculate estimated migration timeframes across the network so that they have a better understanding of the overall time frames to complete migrations. If admins don’t calculate the time frames, they could face a long night ahead of them or an aborted outage period.

What the feature means to virtualisation admins
Shared-nothing live migration won't dramatically change the way you do things today, but don’t dismiss it as just another standard feature. It can add value to a virtualisation admin’s existing tool sets.

It is possible that software vendors within the Microsoft ecosystem will develop products that use the feature as an enabler for their own backup or disaster recovery tools.

IT admins should get used to the concept of shared-nothing capability.  VMware plans to introduce this functionality in vSphere 5.1, too, so it is certainly not a flash in the pan and will become a de facto migration choice across multi-vendor virtualised environments.

Daniel Eason is an infrastructure architect at a multinational company and is based in the UK.

 

COMMENTS powered by Disqus  //  Commenting policy