Data storage is a crucial part of a solution provider’s offering, whether they offer cloud services or are planning to. Regardless of whether you’re a large-scale cloud infrastructure as a service (IaaS) provider or a small regional managed service provider (MSP), a strong storage platform can push you ahead of the competition and help you quickly expand your business. On the other hand, a weak storage platform could become the Achilles’ heel that sets a provider back, whether it be from a missed business opportunity or service-level agreement (SLA) penalties.
Businesses that have sizable storage budgets and employees with specialized skills are well-suited to traditional storage systems. In contrast, service provider operations designed for scale with margin-optimizing efficiency require considerably more storage with considerably less complexity. Many traditional vendors, in trying to meet these requirements, have compelled providers to make sacrifices in the key areas of performance, cost and availability.
With this in mind, service providers looking for a better approach have increasingly invested in flash storage technology. Flash technology has promised -- and largely delivered -- better IOPS and increased throughput for critical, I/O-intensive applications, though often at a much higher cost per GB than traditional hard disk drives (HDDs). But performance is just one attribute cloud service providers are looking for in a storage solution. Also high on their list are requirements such as:
- High availability and reliability. Cloud service providers can’t afford storage infrastructure downtime, lest they risk breaking SLA-driven availability commitments and irreparably damaging their reputations.
- Cost-effective scalability. Flash storage must scale easily and nondisruptively to accommodate rapidly growing capacity requirements, and at a cost that enables service providers to maintain affordable and competitive prices.
- Built-in operational efficiencies. In a similar vein, cloud service providers are looking for storage systems that are frugal in terms of space, power and cooling requirements, since every dollar of cost means one less dollar flowing to the bottom line.
Unfortunately, most flash storage systems today don’t meet these requirements. We believe this gap between solution requirements and actual capabilities, along with some confusion about how and where flash technology can best be deployed, has inhibited the adoption of flash storage among cloud service providers. Let’s take a closer look at these requirements and, in the process, briefly examine how a couple of innovative flash storage vendors are addressing them. Though flash storage technology can be deployed in various ways to meet service provider challenges, we’ll focus on all-flash array solutions.
It’s important to note that many enterprises have similar needs. So while emerging solid-state products satisfy service provider requirements, they also meet those of enterprises looking for high-performance, reliable and scalable storage.
Requirement: Availability and reliability
For service providers looking for a storage platform, high availability is clearly one of the top requirements. Most cloud IaaS providers’ SLAs are based on minimum availability guarantees, meaning storage must operate without disruption in the event of outages or planned maintenance activities.
Unfortunately, many early flash storage offerings have been built around single controller systems, and have lacked the availability features needed to meet these requirements. Nimbus Data Systems, a provider of all-flash arrays, is addressing availability with a dual-active, redundant controller architecture designed to eliminate single points of failure. A companion set of capabilities, such as hot-swappable flash modules and nondisruptive capacity expansion, are in place to prevent outages due to planned maintenance activities. Just as important from a reliability and durability perspective, Nimbus uses enterprise-grade enterprise multi-level cell (eMLC) NAND silicon for an extended life 10 times better than MLC flash chips.
SolidFire, an all-solid-state drive (SSD) vendor that recently emerged from stealth mode, also has a system architected for high availability. The company’s clustered, scale-out storage system employs storage virtualization to spread volumes across the entire drive pool and its Helix technology to replicate data. If a node or drive suffers a hardware failure or needs to be taken offline, the affected data is instantly re-replicated across the cluster, allowing processing to continue without disruption.
While flash storage performance generally lived up to service providers’ expectations, scalability of flash systems didn’t.
With an eye on scalability requirements, Nimbus has designed its flash storage technology with service providers in mind. At 10 TB per rack unit, Nimbus offers one of the highest density primary storage systems on the market today. Service providers can seamlessly add Nimbus enclosures to an existing array, so capacity can be added without forklift upgrades.
SolidFire clusters are also built for scalability, aggregating multiple storage nodes over an iSCSI-based, 10 Gbps Ethernet grid. The cluster scales linearly with each additional node, aggregating performance and capacity resources across all nodes. The company’s primary storage solution scales to 100 nodes, representing more than 1 PB of capacity.
Requirement: Operational efficiencies
In the highly competitive cloud service provider space, efficiency advantages can make or break profitability. Service providers are constantly looking for ways to pack more storage wallop into a smaller footprint, and to wring every last watt of power and BTU of cooling efficiency out of their arrays. Unfortunately, most flash storage offerings to date have fallen short of meeting those efficiency goals.
Nimbus’ approach delivers high-density flash storage with three times the density of 15K HDDs. A single Nimbus E-Class system can crank out the same level of performance using just 7% of the footprint of HDD storage. E-Class systems consume just 5W of power per terabyte, an 80% savings over conventional 15K disks.
For its part, SolidFire employs real-time inline deduplication and compression across all data to streamline capacity needs. All new capacity is thin provisioned. Leveraging the strengths of the scale-out clustered architecture, these capabilities enable SolidFire to achieve 85%-plus capacity utilization without performance degradation.
What this means for cloud service providers (and enterprises)
Flash memory has the potential to take cloud storage availability, scalability and efficiency to a whole new level. Innovative players such as Nimbus Data Systems and SolidFire are bringing flash storage products to market that help realize the potential of solid-state storage.
About the author:
Jeff Byrne is a senior analyst and consultant at Taneja Group.
This story was previously published in Storage magazine.