By Rochna Dhand – Director, Product Management
IT infrastructure managers today are so busy that if you ask them for a list of the top ten challenges they’re facing, they’ll only have time to list five or six of them before rushing off to an urgent meeting.
The following half-dozen issues come up frequently in my conversations with IT leaders:
- Doing more with less – In most organizations, the line-of-business owners are constantly wanting to spin up new workloads, so IT is asked to increase productivity – to do more with fewer resources. Today, virtualization is almost always a key part of the solution.
- Monitoring and troubleshooting complexity – Virtualization has provided massive benefits, but it’s also made monitoring and troubleshooting more complex and time-consuming, due the additional layers of abstraction it introduces in the application and infrastructure stack as well as the IO Blender effect.
- Delayed projects – While virtualization has generally hastened the speed with which new applications can be deployed, there are still significant inefficiencies. These often stem from communication gaps across the different teams responsible for applications, storage and virtualization.
- Missed application SLAs – Application service level agreements (SLAs) are becoming tighter while the underlying infrastructure is becoming more complex. This makes it hard to understand the relationship between infrastructure configuration and application SLAs. Further, application needs from storage, as well the ability for the underlying storage to meet these needs, change often and independently. Often storage is managed on the lowest common denominator, instead of a la carte approach.
- Backup problems – This is a growing problem, as traditional backups choke performance, and recovery point objectives (RPOs) are unacceptably long. Often, backup windows are not long enough to cover backing up all the data that needs protection, leaving the IT administrator to make some hard choices.
- Security – Encryption is similarly inefficient and requires expensive technology.
Our solution to these problems is VM-aware Storage, which integrates three emergent technologies from Nimble Storage and VMware:
- VMware’s Storage Policy Based Management (SPBM)
- Nimble Storage’s Native Policy Based Management
- VMware’s Virtual Volumes (VVols).
Let’s explore how they fit together.
Storage Policy Based Management
Storage Policy Based Management (SPBM) is part of VMware’s policy-based framework, and is also a pillar of the Software Defined Storage (SDS) movement. When implemented correctly by the storage vendor, it enables the storage admin to package and advertise the underlying storage infrastructure’s capabilities in a vSphere admin-friendly way.
The vSphere admin in turn, combines these storage capabilities and creates VM policies. The VM policies are expressed using concepts that the application or VM owner understands – application SLAs that are an expression of application needs. The VM owner simply creates VMs, and assigns the VM policy that best meets the desired SLA. The framework ensures that the VM is created on the right storage resource with the most optimal settings.
This means that application needs are programmatically mapped to the most appropriate storage configurations. This mapping is enabled by the SPBM layer, which sits between the storage consumers (your apps) and the storage providers (your storage arrays). The vSphere layer provides the arbitration between the two. This abstraction of the application’s business needs from the underlying hardware capabilities is also the foundation of Software Defined Storage, which in turn is a significant element in the Software Defined Data Center (SDDC).
SPBM is very useful because:
- It ensures that each VM is provisioned with just the right storage settings, since nothing is lost in translation.
- Through compliance checks, SPBM ensures that the storage SLA is maintained throughout the VM’s life. For instance, if the capabilities of the underlying technology change such that it cannot honor the application SLAs anymore, the VM and vSphere admins are alerted, thus preventing a miss on their SLAs. The reverse is true as well. Because SPBM is effectively a broker that matches application needs to underlying storage hardware capabilities, any new capabilities introduced by the storage administrator or vendor can be seamlessly consumed by a VM.
- Managing VMs through a policy is much more efficient that managing them on a VM-by-VM basis. This result is faster app deployments, and less time required ongoing storage SLA maintenance.
Storage Policy Based Management and VVOLs.
Nimble Storage’s Native Policy Based Management
The next innovation that helps solve the problems outlined above is Nimble Storage’s Native Policy Based Management. Every Nimble Storage array comes with Application Performance policies, which are sets of performance and other settings pre-tuned for the most common applications seen in a typical IT org. The policies are easy to use, and are named after specific applications, such as Exchange 2013.
When provisioning a workload on a Nimble Storage array, the admin simply creates a volume and assigns the appropriate Application Policy. The Nimble Operating System ensures that the volume is provisioned with the block size, cache policy, and other parameters that are optimal for each application.
In part, the settings are based on each application vendor’s specific recommendations. But they’re also informed by the rich application performance insights that are uncovered by Nimble Storage’s unique predictive analytics engine, InfoSight.
VMware’s Virtual Volumes
The third pillar of Nimble’s VM-aware Storage is VMware’s Virtual Volumes, a disruptive new storage technology that comes with significant advantages for feature-rich storage vendors like Nimble Storage.
Essentially, the Nimble Storage implementation of VVOLS maps every virtual disk (vmdk) to its own native Nimble volume. Imagine a VM that has two virtual disks – one for the OS, and another for the application data. In the new word of VVOLS, this would translate into two separate native Nimble volumes. (Actually there would be a few more volumes but the concept is the same).
Operating at this level of granularity solves many challenges and opens up interesting opportunities. Some storage capabilities are best left to the storage array instead of being executed on a compute infrastructure that was designed for applications, not storage.
The list of Nimble Storage capabilities available at a VM granularity via VVOLs is long:
- Snapshots – Nimble Storage has one of the best snapshot technologies in the industry – application synchronized, and very CPU- and space-efficient. Further, we provide rich scheduling and retention controls. In fact, a very high percentage of our customers use our snapshots for primary backup. Snapshot based backups are superior to traditional back up as they have no application performance impact or need for separate storage devices, They also significantly reduce the back-up window, allowing you to back up more data than ever before. With VVOLs three things happen:
- The native vCenter Virtual Machine snapshots are offloaded to the Nimble Storage array. Because these are now hardware-based snapshots, a virtualization administrator can take more of them and keep them for much longer periods of time because there is no penalty when deleting them.
- These hardware-based snapshots are now available at a VM granularity. As a result backups can be tailored to each application, for instance, its RPO and retention needs.
- Further, the snapshots can be configured from the vCenter native pane-of-glass. There’s no need to use the array GUI, or even Nimble Storage’s vCenter plug-in.
- Cloning – Like snapshots, Nimble clones are very resource efficient. With VVOLs:
- Cloning operations are offloaded to the Nimble Storage array.
- Nimble Storage’s ability to instantly provide “zero copy” clones allows the virtualization administrator to quickly and efficiently provision virtual machines.
- And they can do so using the native vCenter interface.
- Replication – Nimble replication is performance- and bandwidth-efficient, and a very popular way to achieve DR (disaster recovery). As with snapshots, if desired replication can be configured at a VM granularity, straight from vCenter.
- Encryption – Nimble provides the ability to encrypt volumes based on a very efficient software-based encryption implementation. Many customers and service providers use Nimble-based encryption to meet their regulatory compliance needs. VVOLs brings additional benefits to a vSphere environment.
- VM encryption can be offloaded to Nimble’s built-in encryption-at-rest.
- VM Encryption now becomes cost effective – actually, it’s free, as Nimble always includes all features in the hardware SKU.
- VVOLS gives the VM and vSphere admin the ability to enable encryption and control it on a VM-by-VM basis, again, from the vCenter pane-of-glass.
- Application Policies – Application policies are evidence of Nimble’s obsession with simplifying storage management. With VVOLs VM and vSphere, admins can leverage this simplicity directly via the native vCenter pane-of-glass. This is in stark contrast to other vendors’ implementations where the complexity of the underlying technology is exposed to the non-storage admin. Concepts such as RAID, FlexVol, block size, and stripes are not something a vSphere admin should need to understand. With VVOLS, Nimble Application policies can be applied on a VM by VM basis. This means the underlying storage is optimized for each individual application. No more least-common-denominator provision.
- Deduplication – Nimble Storage recently added a very powerful set of dedupe and other data reduction technologies as part of the recent launch of the Nimble Predictive Flash Platform launch. Deduplication on Nimble is very effective with a very low performance impact. With VVOLs and SPBM, admins can enable (and disable) Nimble dedupe on a VM-by-VM basis from vCenter directly.
Manage policies, not VMs. Offload critical operations such as backup, replication, de-dupe, and security to the storage array.
So, how does Nimble Storage’s integration of these three technologies solve six of the core problems facing IT decision-makers today?
- Doing more with less – Storage management tasks that formerly required hours of time can now be performed in minutes by eliminating onerous cross-silo communications during new project provisioning, as well as ongoing troubleshooting and storage configuration changes. Further, management now occurs at a policy level instead of on a VM by VM basis. Offloading snapshots, backup, and security to the array leads to significant savings.
- Monitoring and troubleshooting complexity – InfoSight gives all our customers the ability to easily identify and predict problems, not just in the storage component but even elsewhere in the infrastructure stack. There are no more layers of abstraction or the IO Blender effect.
- Delayed projects – Not anymore. Policy based management means that even “one-off” projects are actually variations on a theme, which can be rapidly rolled out with extremely high reliability.
- Missed application SLAs – Policy-based storage management results in the articulation of storage capabilities in application SLA terms. This in turn leads to accurate first time placement of the VM on storage, as well as ongoing assessments to ensure nothing has changed.
- Backup problems – The combination of VVols plus Nimble snapshots provides an extremely powerful backup solution, even more so when administered at the policy level rather than VM by VM.
- Security – Nimble provides encryption at rest with virtually no performance penalty, so that’s one less thing to worry about.
In conclusion, using Nimble VM-aware storage enables an IT manager to stay focused on each application’s needs, and painlessly optimize the underlying Nimble Storage array for each individual application. Further, key data services can leverage a more efficient technology while significantly reducing total cost of ownership.
Managing by policy instead of a VM-by-VM basis leads to increased productivity, freeing up cycles to focus on more strategic activities. With VM-aware Storage you manage applications, not storage.