Virtualisation has had a profound effect on IT in recent years. As it has become more and more pervasive, applications have evolved, virtualised and become the centre of many business practices. But despite the significant advances promised by virtualised applications, many IT teams are discovering there is a disconnect between those apps and the storage that supports them.
Put simply, there is a glaring divide between many businesses’ physical storage infrastructure (based on LUNs and volumes) and their virtualised apps. In the past, physical workloads were mapped to their own storage LUNs, but that environment is rapidly fading now that 75% of workloads are virtualised, including mission critical applications that run on virtual machines.
When issues arise, IT often looks to the application or the network to find the root cause of the problem, but visibility into the supporting storage infrastructure is not normally available. This is a pity because without it, IT teams are often left shooting in the dark.
Stuck in the I/O blender
While virtualisation has revolutionised much of IT infrastructure, storage has lagged behind, taking far too long to adapt to a world increasingly based on virtual machines (VMs) and containers. The reason for this disconnect is mainly because mapping virtual workloads to physical storage is complicated and time consuming. Unfortunately, the sluggishness and lack of awareness from businesses in adapting storage to fit the new world of virtual workloads has led to the creation of the "I/O blender effect".
This is a phenomenon where virtual environments generate far more random I/O patterns that can seriously choke hard disk-centric storage and increase latency. The I/O blender effect degrades storage performance and many IT teams are still battling to counter it, years after implementing their virtualised environments. To make things even more difficult, the effects are likely to be compounded as businesses virtualise more of their infrastructure and I/O becomes even less predictable.
The two most common methods used to reduce the I/O blender effect are over-provisioning and the addition of solid-state drives for caching. However, both of these approaches bring their own problems, as well as added expense.
How to get out
One way to escape the I/O blender is to map storage resources directly to each individual VM and the application running on it. This would enable IT teams to effectively monitor storage performance and distribute resources based on the importance of the application and the number of users depending on it.
This can be accomplished by adopting storage built specifically for virtualised environments that provides VM level visibility and VM level manageability. This would enable IT teams to create thresholds for VMs to track specific applications.
Storage built for virtualisation, VM-aware storage (VAS), strips out the complexity of LUNs and volumes attached to physical workloads and integrates with applications in a virtual infrastructure. VAS enables businesses to profile applications and base their planning for capacity and performance needs on exact VM behaviours.
Storage for the virtualisation age
Traditionally, organisations have attached a high priority to ensuring the efficiency of the application and the underlying network infrastructure it runs on. But in a virtualised environment, that’s only half the battle. Without the right storage resources, all the fine-tuning in the world will only be able to improve the performance a limited degree. Trying to run virtualised applications and workloads with storage developed for physical applications and workloads is often a recipe for declining performance.
By implementing storage fit for the virtualisation age, organisations can make virtualisation predictable and easily scalable, improve performance significantly, slash costs and provide higher IT productivity. In contrast to the storage constructs (LUNs and volumes) attached to physical workloads and applications, VAS can provide the cost effective capacity, simplified management and scalability suitable for virtualised environments and applications.
Now that virtualisation has become commonplace at the server and application level, it makes perfect sense to provide a storage platform that can deliver the full benefits of a virtualised environment. Trying to get the best out of virtualised applications without the right storage platform to deliver the performance it requires is like trying to run a brand new version of an operating system on an ageing computer. It will probably work, but it’s not going to give you the performance you’d expect.
