Find out what skills storage administrators will need as SSDs, object storage and other trends gain traction.
The world of data storage is evolving faster than dinosaurs after the asteroid struck. Much of the old storage “theology” is on the block as we move to a world of solid-state, software-defined, open source, cloudy appliances and leave RAID arrays behind. That inevitably means that the skills needed to be a successful storage administrator also are changing.
Let’s first look at some timelines. Solid state is already mainstream and 2017 will see a massive jump in usage as 3D NAND hits its stride. With the industry promising 100 TB 2.5 inch SSDs in 2017, even bulk storage is going to change from hard-disk drives. Software-defined storage (SDS) is really just getting started, but if its networking equivalent (SDN) is a guide, we can expect to see it gain traction quickly.
Open source code, such as Ceph and OpenStack, is already a recognized business alternative. Cloud storage today is mainstream as a storage vehicle for cold data, but still emerging for mission-critical information. This year, we can expect OpenStack hybrid clouds to transition to production operations with the arrival of new management tools and approaches to storage.
Coupled with these storage changes are several transitions under way in servers and networking. The most important is the migration of virtual instances to the container model. Not only do containers impact server efficiency, the ability to manage them and integrate data and network storage resources across the hybrid environment is going to be an in-demand skill in the next-generation data center.
One poorly understood but important issue is how to tune performance in the new environment. We are still getting the wheels to turn in so much of this new stuff, but at some point the realization will hit that a well-tuned data management approach will prevent many of the ills that could arise in performance and security.
In this environment, demand for many traditional storage skills will decline. With cloud backup and archiving rapidly becoming standard, anything to do with traditional backup and tape libraries has to top the list of skills on the way out. Tape has been declared dead regularly for decades, but now the low prices and built-in disaster recovery benefits of the cloud make any tape-based approach impractical.
RAID-based skills are in the same boat. Array sales are dropping off as small Ethernet appliances make for more flexible solutions. In fact, the block-IO model, which struggles to scale, is in decline, replaced by REST and object storage. Skills ranging from building Fibre-Channel SANs to managing LUNs and partitions will be less needed as the decline of the traditional SAN occurs, though IT is conservative and the SAN will fade away, not instantly disappear.
NAS access is in many ways object storage with a different protocol to ask for the objects. While the file model will tend to stick around, just as block-IO will take time to go away, increasingly it will be offered on an object platform, which means that a NAS admin will need to become skilled with object storage approaches.
Continue on to find out what data storage skills will be in demand in the years ahead.
(Image: Mark Agnor/Shutterstock)
Based on unit volumes shipped, half the storage admins in the world still shy away from SSDs. However, if you understand the collateral changes in usage, away from dual-port drives and RAID-based data integrity, most of the FUD around SSDs evaporates. That means that knowing how to source, integrate, and use SSDs will replace caring for legacy RAID as a primary skill on the storage team.
2017 will bring 100 TB SSDs, while hard drives struggle to pass 15 TB. Add in the 1000x random performance and the 100x streaming of SSDs, which are already reaching 10 GB per second, and any residual pricing deltas will be offset by the far fewer servers needed. Understanding the dynamics of all of this is a crucial storage skill.
Ceph is now delivered with filer and block-IO protocols in addition to RESTful interfaces to object storage. This type of “universal” storage is likely to become the norm as other object storage packages emulate Ceph. Most of the object solutions in the market are similar to Ceph, often with specialized features such as compression still lacking in the Ceph approach.
Knowing the alternatives and how to take best advantage of object storage for both structured and unstructured data is a very useful skill to have. Being the go-to resource on all things object storage would be an enviable position indeed. The hot players are Ceph (supported by Red Hat), Scality, Caringo, and OpenStack Swift. It also would be useful to understand IBM Cleversafe.
Cloud-based backup, archiving, and disaster recovery
Most IT shops back up at least part of their data to the cloud. This is still an evolving area with the emergence of hybrid clouds and containers, while vendors are adding new features and faster ways to archive and recover data.
DR is a very important but underrated issue in the cloud model. Few data centers take a comprehensive view of recovery, such as where the recovery servers are located (hint: in the cloud!) and how to rebuild data access to petabytes of cold storage and snapshot backups.
With the cloud and the ability to fire up a working data center emulation, downtime likely will to drop from tens of hours to minutes. There’s a career to be made from knowing how to do this!
Docker is one of the hottest areas in IT today. The additional agility of containers compared to traditional hypervisors, coupled with a much more efficient use of memory and compute resources, offers cost and space savings. Understanding the storage needs for Docker, and networked data integrity, is a skills opportunity that promises a good payback.
More advanced container storage skills include managing interactions with hypervisors ( especially VMware vSphere), understanding the performance needs of both instance types and different classes of data, and managing data across a hybrid cloud.
Hybrid cloud data management
Anyone who has studied hybrid clouds asks the question, “Where should I put my data?” For example, placement of data can turn cloud bursting from a near-instant solution for workload issues into hours-long data moves that satisfy no one. Understanding data encryption methods also is critical to cloud success.
Automation promises to address hybrid cloud issues, so following the trends in hybrid cloud management software is a necessary first step in becoming a cloud data management expert.
(Image: I'm Friday/Shutterstock)
Policy-based management systems
All management on a well-designed cloud will be policy-based, allowing departmental cloud users to set up their own virtual spaces as needed. While this type of orchestration has been typically thought of in the server context, the arrival of SDN and SDS will make this a part of the storage admin team's workload and, if well executed, should result in a good deal of manpower savings and better user satisfaction as setups become much faster.
As storage appliances have shrunk, they’ve become almost identical to typical servers in a cluster. Taking this a step further, hyperconverged infrastructure conceptualizes servers and storage as essentially identical, with the drives in each box shared as a pool across the cluster and the storage software hosted in virtual instances. This is a recent approach that's still evolving. It does fit the concept of SDS remarkably well, resulting in a high-growth rate. With a good possibility that this is the way of the future in storage, knowing this area and understanding advanced issues such as clustering with RDMA are important new storage skills.
NVMe and RDMA
NVMe is a post-SCSI method for queueing and executing I/O that is much faster and flexible with lower overhead than Fibre Channel or SAS. We are moving out from typical PCIe-based solutions to include miniature cards (still with terabytes of flash) and solutions based on a fabric interconnect, such as Ethernet, InfiniBand or FC.
This is the future of primary storage, with advanced packaging concepts such as NVDIMM heading for the fabric interconnect approach. All of these approaches use Remote Direct Memory Access (RDMA), which allows very direct memory to memory operations. Performance increases dramatically, but so does complexity related to configuration optimization and performance tuning.
Understanding NVMe/RDMA is a must-have skill in the storage team if big-data analytics are on the agenda.
(Image: InfiniBand Trade Association/RoCE Initiative)
Open source storage
Storage is poised for a Linux-like revolution, with open source code delivering the core software stacks that will turn low-cost COTS platforms into powerful storage appliances. Ceph is today’s leading contender in the open source storage space, together with scale-out file systems from Red Hat and the Hadoop File System, among others.
Add to these offerings a variety of packages unbundled from proprietary hardware platforms that are lining up to meet the SDS approach and the ability to build storage solutions al-la-carte looms on the near horizon. As with Linux, having the right skills and knowledge here will make a storage administrator valuable in the enterprise.