There’s a great whitepaper out from Storage Strategies NOW (December 2012) that discusses the storage systems requirements for mid-sized and enterprise datacenters. As we’re working with customers on storage strategies, we’re seeing the same concerns arise. The demands should be for storage systems that deliver the best productivity and performance, scales economically and has self-healing capabilities.
They outline eight features that are essential in today’s datacenter. They are:
- Capacity utilization must be efficient. The key to maximizing system capacity while minimizing cost is having an efficient algorithm for in-line compression, with sufficient compute resources to avoid any performance impact.
- Performance should be designed into the architecture. Like storage capacity, performance needs to be optimized for maximum efficiency. In virtualized workloads, even sequentially operating applications become randomized when other workloads are mixed together on a single physical machine. The system should have the right hybrid mix of solid state hardware and rotating hard disks with appropriate algorithms. This maximizes performance and intelligently places data for retrieval at rates appropriate to application requirements without breaking the bank.
- Scaling should be non-disruptive and independent. The system should be able to increase capacity by the addition of disk shelves.
- Data Protection should be built in. Storage systems should have features to help protect the data against application or user errors, or even disasters. Functionality such as backup and recovery should be integrated and not the function of a different piece of hardware.
- Data resilience and high availability should be integrated. Every storage system should have multiple levels to ensure the integrity and resilience of data against possible corruption. A RAID implementation that can efficiently withstand multiple drive failures without losing data is essential.
- A vendor should deliver application optimized storage. An advance knowledge of the customer’s application environment is essential.
- Applications workloads should be virtualized without specialized knowledge beyond the hypervisor.
- Storage monitoring and corrective actions should occur within the system. The system should monitor its conditions and report problems proactively.
Tim Krueger, PEI