This content is part of the Essential Guide: VMware VSAN features and realities

EMC explains differences between ScaleIO, VSAN and ViPR software

What's the difference between EMC's ScaleIO, VMware VSAN software and ViPR? That's a question EMC execs get a lot these days.

LAS VEGAS -- Integrating ScaleIO block storage capabilities into its ViPR software platform at EMC World this week...

brought on questions about EMC's software-defined storage strategy.

Why do customers need ViPR and ScaleIO if they share characteristics? And doesn't ScaleIO compete with new virtual SAN (VSAN) software recently launched by EMC-owned VMware?

EMC acquired ScaleIO last summer, in between disclosing plans to add ViPR software-defined storage at EMC World 2013 and VMware's launch of the VSAN beta at VMworld. ScaleIO and VMware both can turn commodity hardware into shared storage.

EMC labels both VMware Virtual SAN and ScaleIO as software-defined- storage. They are both block-based, distributed transactional software stacks with similar architectural models, according to EMC executives. And both have a connection with ViPR.

EMC executives admit the similarities but say ScaleIO scales better and supports more types of servers than VMware-specific VSAN.

"VSAN and ScaleIO are both software-defined storage that provide hyper-converged storage," said Erez Webman, ScaleIO founder and currently CTO and chief architect for EMC's ScaleIO/server storage software group. "When customers look for a solution that is VMware only, they can choose between VSAN and ScaleIO. I can see such environments, but most environments where heterogeneous (external) storage is a concern, I would see ScaleIO."

The products are even more connected now that ScaleIO is integrated into the ViPR controller that manages block, object and Hadoop Distributed File System (HDFS). The ViPR controller can also plug into VMware vSphere VSAN.

What separates the ScaleIO and VSAN is scalability, said David Goulden, CEO of EMC's Information Infrastructure.

"VSAN and ScaleIO are both block platforms," Goulden said Monday during a media question and answer session. "But they're for different workloads. VSAN is hyper-converged storage. Think of VSAN as an extension of (VMware) vSphere, it only runs when physically connected to vSphere. ScaleIO is aimed to scale. You can add thousands of nodes. VSAN is limited in size, it scales to 32 nodes.

"Just like you have multiple block physical arrays, you have multiple types of software-defined storage."

And EMC does not worry about having multiple types of overlapping storage. There is overlap between many of its platforms, such as its VNX, VMAX, Isilon and Atmos arrays and even its data protection applications. EMC's acquisition of DSSD this week will bring more flash to the vendor that already has hybrid and all-flash arrays plus server-side flash products.

"Different workloads have different requirements," Goulden said. "There's no way one size fits all."

Razor Technology, an enterprise systems integrator, has six ScaleIO installations among its customers, according to Razor distinguished engineer Oleg Lukyanov. One Razor customer is running more than 30 servers.

"Whenever you grow the system, you get immediate performance," Lukyanov said. "If you run out of performance and you need more IOPS, you just add more drives. I can take drives from virtual and physical systems and create a pool. It rebalances all the LUNs automatically in the background. This is very unusual for any system. The system is so simple."

Lukyanov said to get that kind of storage scalability, the choices are buy expensive traditional storage systems or build an iSCSI system. In either case, he said, management is not so simple.

"You have to manage all the storage that is allocated, track what is shared from where and which volume you pulled out of it and which servers are using the volumes," he said of traditional SANs. "The management overhead is high and it just kills any Capex in savings."

Adding more capacity to a ScaleIO configuration is not as complex, he said.

"You basically pick your server. You go back to ScaleIO and allocate a pool volume out of it. Done," Lukyanov said. "Then you can ask ScaleIO where the volume is and allocate to it. You don't need spreadsheets."

Dig Deeper on Converged Infrastructure Vendors