By Mark Young, Director of Systems Engineering EMEA, Tintri
Quality of Service (QoS) is nothing new in the data centre. More than a decade ago, when Information Lifecycle Management was hyped in the industry, QoS was already important, albeit for physical servers at the time. Today, after the virtual revolution, administrators are struggling to juggle QoS for their critical VMs. But traditional storage, with its ancient storage concepts like LUNs and volumes, limits the storage admin’s ability to manage VMs at a VM level.
What IT admins need is a tool that allows them to not only manage those critical VMs easily, but also eliminates the guesswork of QoS settings. With traditional tools it is only possible to set QoS at the LUN-level, forcing admins to plaster performance across dozens or even hundreds of very different VMs. New smart storage solutions make it possible to define QoS for every individual VM and to make sure that each one gets the exact input/output per second (IOPS) it needs.
Why is traditional storage limited in managing QoS?
The trouble with conventional QoS is that it works at the volume level. So when you set minimum or maximum IOPS you can only do so for an entire LUN or volume and all the dozens of very different VMs inside get the very same QoS. And if you get those settings wrong, you only hear about it later when users start complaining.
What admins need is VM-centric storage that delivers QoS at the VM-level to guarantee storage performance for VMs. VM-centric storage offers full visibility of every single VM, not just LUNs and volumes. Furthermore, you can see the latency broken down for every part of the system: Latency from the hypervisor host, the network, the storage contention from system overload, the flash component, the disk component from flash-miss and the throttle associated with QoS settings.
This level of visibility of a VM-centric storage dashboard gives very clear insight into performance. With the ability to manage QoS on a VM level, there are no noisy neighbours and each VM gets its own swim-lane with its fair share of performance resources.
How can admins manage QoS at the VM-level?
A classic: Users start to complain that Exchange is running slow and a support ticket is raised. Finding the problem’s source with a system managed on a VM-level is child’s play. As all VMs are listed on the dashboard, it is easy to find the rogue VM sucking up bandwidth. But what can be done about that? With a VM-level storage system you set a QoS policy for that rogue VM to bring it under control, with real-time graphs that show the IOPS the rogue VM is consuming. QoS for any VM can be configured by just right-clicking on the VM and simply setting maximum IOPS it can consume. Any issue with QoS can be fixed in minutes and the result is immediately visible in the dashboard and to the user that complained about the slow system in the first place.
Another example: Many businesses have very important servers that are mission critical and should be guaranteed high performance on that VM at all times. A click on the VM might show that it has some contention latency, meaning it is fighting with other VMs for resources. Just like setting a maximum for rogue VMs, VM-aware storage can easily be used to guarantee a minimum level of IOPS. The same principle applies: Just a right-click on the VM lets you configure QoS and you can easily set a minimum.
QoS set on LUNs and volumes does not help if what you care about is VMs and applications. VM-level storage is simple, powerful and visual. It puts the admin back in control and saves time and money. It is the only storage solution that can guarantee storage performance at the VM-level. That is smart storage.