Software Defined Storage Challenges - EnterpriseStorageForum.com

Software Defined Storage Challenges

Vendors are self-defining SDS (software-defined storage) seven ways from Sunday. Let’s go with a definition that is broad enough to fit most implementations and narrow enough to be useful: Software-Defined Storage orchestrates multiple storage devices into a higher level software management layer. This control layer operates above the physical storage layer and pools storage resources for centralized management, provisioning, workload optimization, data protection, storage capacity pooling, and reporting.

There is no Super SDS that does it all but individual products will do one or more of these jobs exceptionally well. The benefits are quite substantial. For example, SDS implementations can optimize application workloads by needs and priority, centrally manage heterogeneous storage systems, or make it possible to scale-out cost-effectively without forklift replacement of existing systems.

Individual SDS implementations differ strongly. Understand different SDS classifications and the impact on your storage infrastructure. Match product strengths to your present and future needs before you buy.


We classify SDs products as virtual storage appliances, control layer intelligence, server-side controllers, and aggregated data management.  

· Virtual Storage Appliance (VSA). VSAs provide a wide range of virtualized storage services and are not limited to SDS. However, several vendors market VSAs as SDS if they abstract data management from the underlying storage systems, and enable VMs to share pooled storage and resources. The VSA enables underlying storage to be transparent to the VM, whether the storage is direct-attached, SAN, NAS or even the cloud. Because VSAs are a mature technology, those enabled for SDS represent the majority of SDS products in today’s market. A few examples include HP StoreVirtual VSA and VMware vSAN.

· Control Layer Orchestration. This implementation delivers SDS in a control layer atop physical storage systems. These products pool storage resources and abstract storage management tools from existing attached systems. The physical data paths are not changed, which allows IT to slot in control layer orchestration without significant upheaval to the storage infrastructure. Control layers enable SDS in heterogeneous storage environments, depending on what systems the vendor actually supports -- which may not be many at this early SDS stage. This product approach adds intelligence and centralized management, but takes heavy engineering to admit different models and vendor storage systems. EMC’s ViPR is a strong contender as are older but mature storage management software from IBM SVC and software-only DataCore.

· Server-Side Virtual Controllers. This product type enables SDS using virtualized storage controllers on physical or virtual systems. These systems generally depend on same-vendor storage systems to ease API and centralized management issues. IT uses a central management console to administrate the virtual controllers. The control layer may dispatch different policies to different controllers, allowing different workload treatments and resource policies for different workloads. These products are particularly useful in high I/O environments since they take advantage of server processing speeds instead of the slower physical storage controllers. A leading example of this architecture is Gridstore.

· Aggregated data management. This classification enables SDS across geo-dispersed storage systems of different types. Storage assets may be tape, disk, flash; cloud-based, remote or local. Tarmin GridBank uses a global namespace to pool distributed storage assets across multiple locations. Aggregating data processes enables Tarmin to add additional eDiscovery and analytics capabilities into its product.

Challenges of Software-Defined Storage

No one SDS product offers all benefits. Understand what your storage environment needs before you go shopping.

#1. Challenge: Large and/or complex workloads. SDS Solution: Optimize for different data needs. Some SDS products intelligently manage differing workload types by placing data on optimal storage types and tiers. For example, the product might store metadata and highly active data on flash and less frequently accessed data on disk. Physical and cloud hybrid storage systems have done this for a long time; the difference with SDS is that the product can use multiple storage systems for optimized placement.

#2. Challenge: Sprawling storage infrastructure. SDS Solution: Extend and simplify storage management. All SDS products offer some level of centralized control. This becomes particularly important with geographically dispersed storage infrastructure as well as heterogeneous storage in the data center. If your infrastructure is highly complex than look for SDS products that manage across several different storage systems and locations.

#3. Challenge: Poor scalability. SDS Solution: Cost-effective and efficient scalability. If you need scale-out storage without the scale-out price tag, look to SDS products that build on commodity hardware, or that leverage your existing storage systems for greater capacity and performance. Be aware that commodity storage may or may not provide the performance levels that you need. Performance does not automatically increase along with capacity; know your data workloads and performance needs before investing in commoditized storage with an SDS control layer.

#4. Challenge: High storage management costs. SDS Solution: Time and cost savings. Decoupling storage management from media saves on capital and operating expenses. It becomes easier to manage a sprawling storage infrastructure and optimizes workloads for faster processing. Centralized management enables IT to better create security and policies across the infrastructure instead of system-by-system using different interfaces.

#5. Challenge: Manage data silos. SDS Solution: Centralize management. Managing disparate storage systems in a single data center is challenging; managing storage systems across multiple locations is even harder. All SDS systems centralize management to a degree; if you need to centrally manage geo-dispersed systems, choose one that operates on distributed storage of different types. If you have heterogeneous storage in your data center, consider a control orchestration layer.

#6. Challenge: Provisioning multiple storage systems. SDS Solution: Policy-driven provisioning across systems. Even a single storage system can be tricky to provision, leading to under-provisioning with capacity over-runs, or over-provisioning for higher costs and management time. Centralized, automated provisioning across storage systems relieves under- and over-provisioning problems and saves IT management time.

Conclusion

There is no magic wand to make all storage systems play nicely with each other but SDS is a big step in the right direction. It is an up-and-coming technology that can provide significant cost and management time savings, and lower storage management risks.

The ultimate benefit of SDS is cost-effectively improving quality of service throughout the storage infrastructure. Even first-gen SDS is making distinct QoS improvements, and as development continues vendors will add more intelligence at the control layer. This will improve storage system functionality and efficiency at a far greater level than was previously possible.

Our take: don’t wait on a second-gen development. Take advantage now of an SDS pilot project to test the waters and take advantage of the real benefits SDS offers. As time goes on you can expand SDS into more and more corners of your infrastructure, growing your advantages and benefits as you go.

Christine Taylor is a well-known technology writer and industry-watcher.

Photo courtesy of Shutterstock.


Comment and Contribute

 


(Maximum characters: 1200). You have characters left.

 

 

Storage Daily
Don't miss an article. Subscribe to our newsletter below.

Thanks for your registration, follow us on our social networks to keep up-to-date