Software Defined Storage (SDS) is the abstraction of data services from the storage hardware and it promises a lot: flexibility, cost savings and the elimination of vendor lock-in. Despite the hype, organizations have been slow to adopt the technology and sales of turnkey vendor solutions continue to be strong. The hesitation to move to SDS can be a timing issue as organizations wait for their next storage refresh. Comfort with the new technology is also a factor. Deciding if and when to move to an SDS infrastructure is a critical decision for IT to make this year.
The Value of SDS?
Before determining if and when SDS should be deployed it is important that the IT planner has an understanding of what SDS will bring to an organization that adopts it. Today, SDS brings flexibility and simplicity to the data center. While every data center wants to be flexible, not every data center needs to be just yet. Flexibility is required by data centers that acquire new storage systems, frequently through growth or already having a variety of storage systems on the data center floor. SDS flexibility means that the data center can manage these various storage systems and add new ones from almost any vendor while maintaining a centralized point of data services execution.
This flexibility should also drive down the cost of and simplify the process of acquiring new storage hardware since the IT planner only needs to evaluate the hardware capabilities of the storage system. They can focus on hardware features like performance, reliability and serviceability instead of factoring in a software value add. But the organization should be careful not to go with the cheapest solution available since even in a software defined world, hardware quality and capabilities still matter. As Storage Switzerland discusses in its paper “The Hardware Requirements for Software Defined Storage” storage hardware should be evaluated differently if an SDS solution is in place. This paper is available exclusively to registrants of the on demand webinar “Leaping to SDS? Make Sure You Can Answer These Questions”.
The other key factor is management simplicity. This is not to say that SDS is simpler to operate than vendor provided software but rather that it allows for a single application to execute data services across multiple storage hardware platforms. This capability should lead to reduced training and support costs.
Types of SDS
In addition to understanding the value of SDS it is important to understand the types of SDS that are on the market. With multiple vendors claiming to have an “SDS” solution the IT professional can be overwhelmed by the options.
The first type is hardware defined SDS. A legacy storage hardware vendor that has moved to a software focus typically delivers this type of SDS solution. Their own hardware has often become off the shelf storage servers. The two are combined and sold as one package. Essentially the storage vendors unique IP (Intellectual Property) becomes the software not the hardware. But this type of SDS does not tend to inherit the flexibility and value benefits described above. While the vendor can shift hardware platforms when they deem it is necessary, the customer is locked into just the hardware that vendor has chosen.
The second type of SDS is a SDS solution that runs on a physical appliance that logically sits between the storage system and the attaching hosts. This form of SDS best mimics the legacy dual controller architecture but abstracts it on to a separate appliance. This form of SDS delivers all the benefits above. There is however the potential for the dedicated appliance to become the bottleneck as all I/O has to flow through it but few environments can actually tax this type of deployment. The I/O function is still distributed between the controllers on the storage system (which now have less work to do) and the CPUs on the SDS appliance (which only have to focus on advanced data services). For situations where bottlenecked SDS appliances are a concern, some vendors have added a scale-out architecture to their SDS solution, scaling I/O across SDS server nodes.
The third type of SDS runs in a virtual context. Virtualized SDS comes in two forms. The first is the virtualization of the above dedicated appliance. The appliance essentially runs as a virtual machine across multiple hosts. Other than sharing compute with applications this form of SDS still has access to dedicated storage arrays.
The other form is virtualized SDS that runs across the physical hosts and then aggregates internal storage. This form of SDS is also known as a hyper-converged architecture. In most cases this last form can not aggregate external storage systems into the architecture. The risk with this form is that too much is being asked of the compute tier which is running applications, running storage software and routing storage I/O between internal hard disks. As the environment scales this form of SDS, it becomes difficult to assure predictable performance.
Should SDS Be Implemented?
The big question facing IT is should they implement SDS today? The answer depends on a number of factors, but the two at the top of the list should be how many storage platforms does the data center have and how well does the SDS solution being considered support the various environments that the organization is running? The more storage platforms, the more complex it becomes to individually manage those platforms. Administrators have to be trained on each storage system’s features and each system has to be independently monitored and managed. If the solution supports existing shared storage architectures, SDS levels the playing field by consolidating data services and making them commonly available across the various storage systems.
As the data center continues to grow and deploys new environments such as big data analytics, server and desktop virtualization and the build out of dedicated database clusters, it will result in an increasing number of new storage systems. Many of these new environments will also contain multiple projects as Hyper-V is implemented alongside VMware and new databases like NoSQL are implemented alongside traditional Oracle and MS-SQL architectures. The data center may feel forced to purchase separate storage systems for each new environment and possibly also for each of the sub-projects within each environment.
As the number of storage systems increases the unified management of SDS becomes very compelling. This feature alone should reduce costs and increase IT efficiency. Some environments are better off starting with SDS with virtualization being a good example of such a case. The virtual environment, because of the variety of workloads it will support, can often justify multiple storage systems. The data center may want a flash storage system for performance critical VMs or for desktop images, then a hybrid system for most day to day workloads and a high capacity hard disk system for workloads that are not performance sensitive or are simply inactive.
SDS solutions are a great way to integrate functionality and even the hypervisor vendors themselves are getting into the act. But the IT professional may be better served by an SDS solution that can support both the virtualized and non-virtualized use case.
When Should SDS Be Implemented?
If the decision to deploy a SDS solution has been made, then the final consideration is when to implement it. Most IT professionals tend to focus on the next storage refresh. This could be a good time, especially if that refresh is additive to existing storage resources. But a refresh is still an excellent time to re-consider consolidation. There are systems that do an effective job of integrating flash and HDD while still providing a robust feature set. This combination may allow for a single system to support the mixed environments described above. If the data center can be consolidated to a single array then the need for SDS is eliminated and managing one of something is always easier than managing multiples, even if those multiples have a common interface.
An important consideration is to make sure that the consolidated storage system has the ability to integrate with a future SDS strategy. Even if consolidation is successful, eventually additional storage systems almost always work their way into the data center. Being able to pull off the data services from the consolidated arrays so it better integrates with the SDS strategy is a key differentiator to look for.
A more appropriate time to implement SDS is when a storage addition is needed. The data center does not have the time or budget to refresh the entire architecture and they have a specific point problem to address that can’t be resolved with the current storage assets. Instead of bringing in a net new storage system and learning another new interface and set of data services, it may make more sense to bring in a more hardware focused solution, combine it with SDS and then start integrating the current storage resources into that solution. This allows for a gradual transition to SDS instead of having to “jump in with both feet”. Over time the cost to manage the environment should come down, and additional storage purchases become more seamless.
A final area to consider SDS is when the data protection architecture is being addressed. As Storage Switzerland discussed in its article “Designing Primary Storage to Ease the Backup Burden” replicating the snapshots of a primary storage system to a secondary storage system provides excellent recovery point and recovery time objectives (RPO/RTO). SDS is a key enabler of this strategy as it allows the purchase of a second more cost effective system in which those snapshots and replication can all be triggered from a single management solution.
Depending on the environment SDS is an ideal solution for many data centers. It is best when the solution can be implemented in a crawl-walk-run fashion. The ability to leverage SDS to support the addition of more hardware focused storage systems that address a pressing performance or capacity problem or to leverage SDS as the start of an advanced data protection strategy are two excellent ways to begin.
Software Defined Storage can bring value to almost every data center that has multiple storage systems to manage and that are rapidly expanding. Once those value points are understood and verified, the next step is to decide which of the available SDS solutions should be implemented. Finally, proper timing of the move to SDS is critical and taking a gradual approach allows the organization to gain comfort in the solution’s capabilities. But SDS is not a license to buy junk hardware. Quality, reliability and performance of the hardware still matters. Additionally, the storage hardware has to be able to adapt transitioning from a full featured array to a hardware focused array that can leverage SDS.
Sponsored by X-IO
X-IO Technologies is a leader in performance-driven storage with the industry leading price, performance, and capacity ratios of comparable solutions on the market. They provide flexible data services that can be turned off when used with a SDS solution. Their ISE product family continues to provide a 5-year service guarantee as well as the capacity and performance scaling to meet the needs of any data center exploring or standardizing on SDS.