The Alchemy of Services-Oriented Storage Area Networks
August 30, 2010
By Dave Trowbridge
Alchemists long sought a hypothetical substance called alkahest, possessing the power to dissolve anything. The IT world has found it. It's called virtualization, and it's driving the cloud computing phenomenon. Just as with alkahest, for which no container would suffice, virtualization is well on its way to dissolving the data center itself.
Along the way, it's dissolving every data center technology as well, and storage is no exception. Within individual storage arrays, virtualization can create multiple volumes or LUNs (Logical Unit Numbers), much as server virtualization creates logical processors. A Virtual Storage Area Network (VSAN) divides a physical SAN consisting of many networked arrays into multiple virtual SANs, much the way Virtual Local Area Networks (VLANs) create multiple virtual LANs.
Networked storage virtualization is becoming especially important to enterprises with between 50 terabytes and one petabyte of data, according to Rick Villars, vice president for the Storage Systems & Executive Strategies practice at International Data Corporation (IDC).
"In the last year or so, these enterprises have recognized that improving efficiency in their networked storage is absolutely pivotal to be able to continue their server virtualization efforts," he says. "Without virtualization solutions for networked storage, they start over-provisioning, run into backup and compliance problemsthey simply can't scale effectively."
Scaling with Storage Virtualization
Christian Teeft knows these problems first hand. He is Vice President of Product Development at NTT America Enterprise Hosting. His team of almost 50 engineers manages the computing needs of hundreds of enterprise customers in the company's data centers in Northern Virginia, Northern California and New York, so the dependence of server virtualization on storage virtualization has long been clear to him.
"I now have greater freedom to purchase storage based on the precise features and capabilities that matter to me. And I can swap out technology when it makes the most economic sense. I get more value from my storage investments."
"Agility is an overused term, but in our business it's essential," he says. "Customer expectations never go down: They want IT services their way and they want them faster. We have to be able to bring up storage as fast as we can a server."
There's more data to store, too, due to increasing business reliance on digitized records, a flood of unstructured data video and images in particular and longer-term storage requirements imposed by regulatory and risk management concerns.
"You also have a multiplier effect," says Rajeev Bhardwaj, director of storage product management at Cisco. "When you create a chunk of data, you need a dynamic copy to work on, a local backup copy, a copy in a remote data center for disaster recovery and then you may have to hold on to it for years. So storage needs grow even faster and it gets harder to manage."
Cloud Computing with a Services-Oriented SAN Fabric
NTT America had already built a high speed SAN fabric using Cisco MDS-series switches; in fact, a popular social media site has been running on it since February, 2008. However, Teeft wanted more.
"Our primary challenge was data migration, which happens a lot when you're hosting so many different companies," he says. "We have to balance keeping a storage asset as long as possible to extract maximum value, against customer demand for the latest and greatest. If migration is difficult, you end up with islands of storage that tend to last longer than they should."
The next step, then, was to further abstract the SAN by moving essential services into the fabric. "The point of Cisco's services-oriented SAN (SOS) fabric," says Bhardwaj, "is that it no longer matters who made the storage device, where it is, what protocol it uses, how fast it is or even what type it is disk, tape, virtual tape. You get migration, I/O acceleration, and encryption services as part of the fabric. This is essential for the transition to cloud computing, whether private or public."
Simplifying Data Migration
The migration service now available on MDS-series switches is the Cisco Data Mobility Manager (DMM), a SAN application with simple wizards to automate data migration. For Teeft, DMM delivers both agility and financial benefits.
"Certainly, from the customer side, there's a big difference," he says. "The downtime involved in data migration is no more than the time it takes to reboot the box, 30 seconds to 2 minutes. This is amazing compared to what it was before. We had similar capabilities from each storage vendor, but they didn't play well with each other, and sometimes data migration could take hours."
It is the ability of a services-oriented SAN fabric to simplify the management of heterogeneous storage assets that delivers a key financial benefit for NTT America. "I now have greater freedom to purchase storage based on the precise features and capabilities that matter to me," says Teeft. "And I can swap out technology when it makes the most economic sense. I get more value from my storage investments."
Making the Transition to a Services-Oriented SAN Fabric
NTT America is currently "soft launching" the new capabilities offered by the services-oriented SAN, making them available to customers and studying how best to develop them into full-fledged product offerings. "Putting more intelligence in the SAN fabric is definitely in line with our long-term product plans," says Teeft. "It's aligned with the demand for cloud-like services that we see from more and more customers."
This is not a casual step to take, even for a service provider with wider experience than any enterprise. Rick Villars at IDC cautions enterprise IT managers to proceed with care. For one thing, the ability to easily provision storage for dozens or hundreds of virtual machines means less predictability in application demands.
"There's no doubt that services in the fabric will play an important role in making the virtualization of storage assets practical for many companies," he says. "The real challenge is not implementing the technology. Careful attention to detail is critical to avoiding the potential disruption of a very finely-tuned SAN fabric with new applications and services that have very different behavior from traditional SAN traffic."
Evolution of Enterprise IT
Another challenge involves the human dimension, as "technical silos" collide. This was the experience of Derek Masseth, senior director of infrastructure services at the University of Arizona, a leading U.S. public research university with 38,000 students and over 14,000 staff. There, the convergence of data and storage triggered a collision between Fibre Channel (FC) and Ethernet engineers.
"For the Ethernet guys, a port in the data center is much like any of the other 50,000 data ports they're managing. To them reliability means retransmission of dropped packets. But mention that to a Fibre Channel guy and he'll get red in the face. Storage can't ever, ever drop a packet as far as they're concerned," he says.
This experience has convinced him that moving to a services-oriented approach, whether in storage or elsewhere, requires less focus on technological prowess and more on people with a willingness to learn whatever technology is needed to support a given service.
"Above all, the enterprise IT operational model will have to evolve further," says Bhardwaj. "For instance, migration is no longer a function of the server or the storage device, so your network team ends up responsible for it. The same is true of the other services in SOS, and the rate of change in operational models can only accelerate as more and more services move into the network."
Dave Trowbridge is a freelance writer based in Boulder Creek, CA