Table of Contents
- Summary
- Situational analysis
- The traditional scale-up storage architecture
- Enter the SSD
- Scale-out storage
- Software-defined storage
- Hyper-converged pros and cons
- Software-defined networking advantages
- Combining SDS and SDN
- Key takeaways
- About Howard Marks
- About GigaOm
- Copyright
1. Summary
For decades, most IT organizations have relied on scale-up storage systems for the vast majority of their tier-one applications. Now, however, the modern data center’s increasingly demanding workloads combined with – paradoxically – the performance that flash memory provides as a storage medium, have stressed scale-out architecture. Fortunately, a new generation of storage systems has emerged to address the traditional storage system’s limitations.
Scale-up storage systems use a small number of storage controllers to manage multiple external shelves of drives. As the performance demands of today’s mixed workloads have increased, and flash-based solid-state drives (SSDs) have been increasingly able to meet those demands, the controller has become the bottleneck – the limiting factor in system performance.
In order to have products at various price points, scale-up storage vendors have as many as six models per product line, with each one of them, practically, only scalable over a limited range. Larger models have faster processors with more cores and at high-end proprietary ASICs or other hardware. As users reach the capacity of their systems’ controllers, rather than face a painful upgrade process, they order another system. This creates multiple islands of storage.
Instead of two controllers to manage ten dumb shelves of drives, the new generation of storage systems mounts those drives in dozens – or a hundred or more – industry-standard servers and through software running on those servers, creates an integrated storage system. Since the system’s compute resources increase with the storage media, scale-out systems increase their performance along with capacity. The most advanced of these systems also take advantage of the latest in local area networking (LAN) technology, software-defined networking (SDN) to optimize the connections from workload servers to storage.
This report will examine the traditional scale-up architecture and the challenges it faces in today’s datacenters. It will explore the evolution of the software-defined scale-out approach, discovering where this new model for storage is really a better idea.
Key findings from this report include:
- Scale-up storage controllers can’t deliver the high levels of storage performance that flash-based SSDs promised.
- Technology has surpassed the scale-up architecture, which can no longer handle the demanding workloads in today’s datacenter.
- The new scale-out, software-defined storage platforms have several significant advantages over traditional scale-up systems:
- They deliver scalability of capacity and performance, on demand
- They simplify management and provisioning
- They lower hardware costs through the use of industry standard x86 servers and SSDs
- They increase performance, data, and system resiliency
- Taking the software-defined, scale-out idea to the extreme produces storage solutions that deliver both performance and capacity while being significantly easier to manage than traditional storage systems.
Thumbnail image courtesy of: iStock/Thinkstock.