Virtual Observer

Searching for Intelligent Storage for Virtual Workloads

There are no one-size-fits-all storage solutions, so I've come up with some basic questions to ask the vendors who will be trying to get your attention.

• By Dave Bartoletti
• 10/05/2011

I've been paying close attention this year to the storage platforms my clients are relying on for their virtualization and private cloud deployments. The virtualization mega-trend has dramatically changed the way data and workloads are sized, controlled and protected. But traditional networked storage solutions are all too often rigid, complex and inefficient.

Many storage platforms simply weren't designed for the workload variability and rate of change that are the new "normal." Companies I speak with now demand storage that handles all types of server virtualization, rapidly expanding and massive data sets, and performance-hungry, business-critical database and productivity apps--simultaneously.

Traditional modular and monolithic storage solutions, unfortunately, have distinct price/performance characteristics that force customers into unattractive tradeoffs or into deploying multiple storage platforms to meet each different application need. Legacy storage is hobbled by limited and/or disruptive scalability, constrained performance and capacity optimizations and complex, expensive operations.

The result? Many IT departments are forced into a choice between two unappealing options: Either compromise their application and data flexibility, or reconfigure their storage systems every 2-3 years with a newer generation of the same inflexible storage platform.

Server virtualization has made some promises that traditional SAN, NAS and unified solutions haven't been able to keep. It has promised that we can have it both ways: consolidation and simplicity, flexibility plus efficiency, mobility without downtime. There's a lot of intelligence in your hypervisor to deliver these benefits. But how intelligent is your storage?

My review of the most innovative storage solutions on the market shows that some of the elements of intelligent storage are available in various forms from multiple vendors. However, the power is in the packaging and the devil is in the integrations.

To simplify your search for intelligent storage, I've identified the following "must-have" differentiators for storage solutions that have what it takes to keep up with your virtual workloads--without breaking the bank or your back.

Truly Scale-Out Architecture--Array architecture should be fully modular, and built to scale linearly in all significant dimensions: storage capacity, network bandwidth, and performance. The architecture must support abstraction (grouping, pooling, and sharing) of resources--a fully-virtualized storage foundation--that crosses the traditional disk, controller, and chassis boundaries of an array. No "forklift" upgrades!

Adaptable Capacity and Performance--Intelligent storage solutions must optimize all storage resources automatically. There's no room today for a bit of wasted capacity. There's no time for manual storage management, either. Data should be automatically and transparently placed across multiple (and diverse) physical storage resources, including arrays, RAID groups, disk types and controllers. Tiering functions must be automatic and multi-faceted, providing array-level intelligence for data movement in response to shifting demand.

Enterprise-Class Availability--Admins should not need an advanced degree in storage management to configure for maximum uptime or protect datasets at all times, nor should they have to scramble to recover from common, predictable faults. Faults should be automatically remediated, period. All storage should be RAID protected by default, and RAID configuration must be much simpler. And, a full suite of robust data protection features should be bundled in at no additional cost.

Radical Simplicity--Intelligent storage demands all-in-one hardware and software packaging, including integrated storage deployment, provisioning, management and troubleshooting features. Arrays should go from box to operation in hours, not days or weeks. Scale-out should be a no-brainer: New capacity and performance components must join existing resources transparently and start to load balance immediately and non-disruptively.

Intelligent Infrastructure Integrations--Your storage must evolve beyond being just an undifferentiated "bucket of blocks" or "cabinet of files." It must truly understand virtualization. This requires deep and comprehensive integrations with hypervisor management tools and with every storage optimization technology from VMware and Microsoft, at minimum.

Storage--whether block or file--is an integral, essential and a limiting factor in the virtualized datacenter. Intelligent storage must break its traditional limitations and deliver as much elasticity, flexibility and cost-efficient scalability as your hypervisor.

So, how intelligent is your storage?