Getting to grips with server and storage virtualisation – Cloud Pro

The way that virtualisation has been talked about over the last decade you would be tempted to think every server and storage system has had the virtualisation treatment, but many organisation still haven't completely climbed on board the virtualisation ship headed for cloud nirvana. That said, there is still time and the benefits still stand.

According to Kong Yang, head geek at SolarWinds, virtualisation can bring many benefits to a business, from cost savings and flexibility to making IT workflows and processes more efficient and effective. "However, one of the key benefits is the ability to abstract infrastructure resources, which allows for the re-distribution of resources to applications at a radically faster rate," he says.

He adds the need to embrace virtualisation comes hand-in-hand with growing hybrid IT environments. The recent SolarWinds IT Trends Report 2017 reflects this, as it shows 52% of companies have server virtualisation included within their hybrid IT strategy.

Yang says hybrid IT is very much a reality for UK businesses, as many host some of their infrastructure in the cloud while also maintaining some on-premises. "In fact, our research shows that in the past 12 months, UK organisations have migrated applications (69%), storage (54%), and databases (37%) to the cloud -- this is more than any other area of IT," he says.

Consolidating servers and storage

So, what's the best way to go about consolidating servers and storage with virtualisation? It is critical to first understand the objectives of the virtualisation project and the outcomes that are desired. Often virtualisation forms part of a larger project, such as an application re-platform or datacentre consolidation or migration and it's important to understand how the larger picture affects the specific project.

"Once the objectives are clear, organisations should consider the individual workloads and operational areas to be implemented and how the implementation programme is to be conducted," says Pete Hulme, data centre technical lead at Dimension Data.

"For example, is it acceptable (or desirable) for test and development to share a platform with production? And is it important to be able to produce 'clones' of production for test purposes?"

He adds that companies should consider how business continuity and disaster recovery will be conducted and how the data management tools interact with the virtualisation platform. Adding, they should consider how security and segmentation is to be managed and who is to hold authority and control.

"Once these decisions are made it should then be possible to select a virtualisation platform for each component and to understand how these will interact and interoperate. It is essential to consider the management platform and interfaces that will be required to deploy and operate the platform and how the processes will interact with both new and existing processes," Hulme says.

Best practices

When virtualising infrastructure there are several best practices to be aware of. All x86 servers are now candidates for virtualisation. But according to Richard Stinton, enterprise solutions architect at iland, the main blockers to virtualisation are normally the risk associated with the migration process, especially downtime, and licensing policies of the software being run, the best example being Oracle.

"For this reason, the main servers still running on physical tend to be large transactional databases, such as SQL Server clusters, and Oracle," he says.

David Cottingham, director of XenServer product management and partner engineering at Citrix, says organisations should understand the characteristics of the workloads it is trying to virtualise, in terms of their CPU, memory, and I/O needs.

"A classic problem in virtualisation is for an overly-optimistic administrator to attempt to pack too many virtual machines onto a server and the end users experiencing poor performance when load is high," he says.

Technologies such as dynamic memory control and storage caching can help pack more VMs per server, workload balancing technologies can also support, "especially where VMs' performance is recorded over time, and then recommendations can be made on how to change the distribution of VMs across physical servers," Cottingham adds.

For storage, different issues come into play. Tom O'Neill, CTO international at all-flash storage provider Kaminario, believes the most important aspect of virtualising storage is to ensure the performance and capacity aspect are sized correctly.

"Storage is often the slowest part of any IT project and virtualised workloads often drive storage solutions harder," he says.

This is due to the I/O blender effect of multiple layers of virtualisation. Friendly sequential I/O can become increasingly random as extra layers of virtualisation are introduced. The impact of random vs sequential is much reduced with modern storage media like flash, but it's still considerable because of the caching and CPU deployed within the storage device (SSD).

Virtualising to move to the cloud

Virtualisation can also be considered a stepping stone to the cloud. However, if organisations opt for a private cloud, they should have a few resources in place - VM templates, resource pools, a user interface for self-service and a request process.

"VM templates let individuals requesting a new VM provision [it] themselves automatically, while resource pools designate a maximum quantity of resources that an end user can consume," says Arun Balachandran, product manager, at ManageEngine. "Organisations should also select the right tool as a user interface for self-service and have a request process in place."

He adds that in moving to private cloud, organisations must confirm that it remains cost-effective by means of constant performance monitoring and capacity planning.

"Without right-sizing VMs and failing to track and anticipate resource needs, businesses could face problems such as VM sprawl and resource overconsumption. Businesses should plan for the long term to ensure they have enough resources on hand to meet future business demands. To keep costs low, they can use low cost hypervisors or hypervisors from multiple vendors," he concludes.

Excerpt from:
Getting to grips with server and storage virtualisation - Cloud Pro

Related Posts

Comments are closed.