When unexpectedly heavy use of the virtual desktop infrastructure outgrew existing local storage capabilities at Oregon State University’s College of Business, Engineer–Systems Administrator Alan Sprague had an idea: What about software-defined storage (SDS)?
The College of Business suffered performance bottlenecks for high input/output operations per second workloads, Sprague says. There was no money in the budget to invest in adding a physical storage area network, but a virtual SAN cluster that would absorb those existing drives and balance workloads for optimal performance could be accommodated.
The solution would also cost about a third to a quarter less than a hardware SAN, he says.
“We had servers and drives in them,” Sprague says. “All I needed was to add a couple of solid-state drives and some 10 Gigabit network interfaces — instant SAN.”
“When we talk SDS inside VSAN, and inside VMware’s vCenter Server, [which provides a centralized platform for managing VMware vSphere environments], it takes two clicks to make a VSAN cluster and have it automatically assign drives to it,” Sprague says.
Some other SDS solutions enable access to storage management capabilities from within vCenter, but as a separate interface, which he doesn’t prefer. With VSAN in place, Sprague’s worries about whether workloads are spun across the right drives and correctly balanced are gone.
“It’s all just dealt with automatically, and I always get the best possible performance. It’s an incredibly fast performance.”
Software-Defined Solutions Gaining Traction
Software-defined storage is gaining followers across every industry, including higher education. IDC has reported that sales of traditional stand-alone systems could decline 13 percent through 2018 while sales of new system technologies — defined as all-flash, hyperconverged and software-defined — could grow by 22 percent.
A component of the software-defined data center, SDS essentially abstracts storage management from physical hardware. It presents educational institutions and other organizations with the chance to simplify disparate storage infrastructures — leveraging shared resource pooling and automated management, often in conjunction with commodity hardware — while gaining efficiency, agility and cost savings in the process.
“In my mind, the difference between storage virtualization and SDS boils down to which layer of abstraction you want to refer to,” says Paul Ryan, systems programmer at the College of Education, University of Hawai’i at Mānoa, which is using an open-source SDS system as part of its private cloud implementation.
“Storage virtualization is the first layer of abstraction above physical disks, where you can group these disks into virtual drives,” Ryan says. “SDS, on the other hand, is one layer of abstraction above this, where you speak of clustered file systems that can be based on the capabilities of the underlying disks, so you have more control over quality of service, throughput and latency.”
Educational and similarly sized enterprise environments tended to adopt the open-source technologies related to SDS first. Today, SDS manifests itself in everything from pure software-defined solutions for heterogeneous storage systems (such as VSAN), to hyperconvergence appliances using industry-standard components whose value lies primarily in their SDS capability. Front-end SDS layers have close ties to established storage players’ high-end hardware solutions. The concept has also extended to running some vendors’ storage services on public clouds in order to provide consistent experiences across public, private and hybrid cloud settings.
IT leaders feel “there needs to be an alternative in the way that storage is managed and deployed,” Robinson says. “Organizations are managing more data every year, and the cost of managing their environments is really eating at their budgets and resources.”
Bring On the Data
Oregon’s Sprague says one of the most appealing features of VSAN is that he can use a variety of storage hardware devices from many manufacturers — his team just drops them in and they work. At the Biodesign Institute at Arizona State University, the university’s first interdisciplinary research institute, Scott LeComte, associate director of IT, enjoys similar capability with its DataCore SANsymphony-V SDS solution.
“DataCore has given us flexibility so that we don’t have to buy expensive disks,” LeComte says. “We can buy cheaper disks and DataCore layers the size, performance or capacity based on whatever we tell it.”
Given the Biodesign Institute’s specialization in bioinformatics and other biomedical research — and the many new scientific machines pulling related data into its storage infrastructure — such flexibility is hugely important. The institute comprises about a dozen research centers, and storage has been segmented so that each center gets its own volume for easier management.
“If a center generates a lot of data and needs more space, it’s easy for us to go in and make it 5 terabytes instead of 3, for example,” LeComte says.
The solution also supports thin provisioning for virtual disks, eliminating the need to dedicate full capacity to each center upfront, while still ensuring that each center will get the capacity it needs when it actually needs it.
A lot of obstacles can be overcome in the SDS approach, which is well suited to scaling out. Adding capacity merely entails buying an extra node, adding it to a cluster and rebalancing data across that cluster. Multicampus colleges that opt for hyperconverged appliances with built-in compute, network and storage resources, as well as SDS technology, can simply repeat those deployments across all their locations — and they won’t require storage experts onsite in each one.
“This completely changes the way you can approach storage,” Oregon State University’s Sprague says. “The days of a monolithic SAN with all its associated management, and care and feeding and dedicated storage administrators because of all of the specialized tools that were needed — that’s gone.”
At the University of Hawai’i at Mānoa, Ryan appreciates the fact that its SDS deployment makes it possible to configure for optimal data redundancy. Its cluster lives across three individual servers, each with 12 spinning disks and three solid-state disks attached, for what amounts to triple replication. A failing drive can be replaced and the cluster will rebuild itself.
“The layer of abstraction that SDS provides is important in that we don’t have to manually recover drives,” Ryan says. “The cluster does that.”
According to MeriTalk’s 2014 “Cloud Campus: The Software-Defined College” report, just one in five colleges and universities has so far deployed broader software-defined technology resources; however, the report also states that more than twice as many respondents saw it as an effective solution for their IT challenges. A truly centralized, software-defined campus, according to the survey, could lead to benefits that include increased operational efficiency, improved continuity of operations, greater security and decreased operating and capital expenditures.
As a piece of a larger software-defined technology vision, Sprague says software-defined storage has “the potential to provide a very nimble storage environment at significantly lower cost.”