IBM Corp. became the latest major storage vendor to add a clustered NAS product to its enterprise data storage portfolio, as scale-out quickly becomes a checklist feature for unstructured data storage systems.
The new product, Scale Out Network Attached Storage (SONAS), is based on IBM's General Parallel File System (GPFS) and a cousin of the Scale-Out File Services (SOFS) IBM has offered through IBM Global Technology Services. IBM also launched standalone bundles based on SOFS for information archiving and cloud computing in October, but SONAS project executive Todd Neville said SONAS is a different animal.
While GPFS is the foundation of both types of scale-out systems, "those [Smart Business bundles] offered a variation in back-end storage between the XIV and DS series, and between blade and rackmount servers," Neville said.
SONAS has a fixed underlying hardware infrastructure that consists of between two and 30 IBM System x3650 server nodes, between two and 30 Gigabit Ethernet or 10 Gigabit Ethernet (10 GbE) ports, and a 20 Gbps InfiniBand network connecting the compute nodes to what IBM calls Storage Pods (SP). Storage Pods are subunits consisting of redundant x3650 servers attached to a 4u storage device that holds 60 SAS or SATA disk drives. The storage side of SONAS can scale from one to 30 Storage Pods.
GPFS allows up to 256 snapshots per file system, and up to 256 file systems are allowed per SONAS system. Neville said SONAS is also IBM's first product with a clustered Samba implementation on the front end, which allows the underlying object-based storage system to be accessed over the CIFS and NFS standard file network protocols. SONAS can also be accessed with FTP or HTTP protocols. IBM "made a conscious effort to design this system without anything proprietary or unique," Neville said. Though pricing for SONAS isn't being specifically disclosed by IBM, Neville indicated IBM is using SONAS to target the largest of large shops in the enterprise and cloud data storage markets, at petabyte scale.
SONAS will ship on its general availability date in March with a Tivoli Storage Manager (TSM) client already loaded on each server node. Because GPFS already separates metadata from back-end data, the metadata database can be queried through SONAS to return a list of files sorted by certain parameters, like "files that are 90 days old." That list can then be passed to TSM for backup.
"It's important in the multibillion file space," Neville said. "Finding what needs backing up is often harder than doing backup itself." Neville said IBM plans to add automated tiered storage software features to the system when that product becomes generally available later this year.
The metadata operations that are possible with SONAS are what most intrigued Jeff Boles, senior analyst and director, validation services at Hopkinton, Mass.-based Taneja Group. "Being able to index everything at blazing speed makes backup possible for huge things like this that currently, backup is impossible for," he said.
Details of SONAS were made public on IBM's website earlier this week, touching off some buzz in the industry about whether IBM's use of its own scale-out IP was an indication one way or another about the fate of IBM network-attached storage (NAS) OEM partner NetApp's long-anticipated Ontap 8 scale-out integration. In August 2009, NetApp said it was beginning to roll out the release, but that integration between Ontap 7G's data protection features other than mirroring and Ontap GX scale-out capabilities would come in subsequent releases. This final convergence hasn't yet been publicly announced.
"This is another wake-up call for NetApp. How many more do they need?" wrote StorageMojo analyst Robin Harris to SearchStorage.com in an email. "They may have a technology response with Data OnTap 8 -- but they don't have an economic one that meets the commodity scale-out storage price point."
However, Taneja Group's Boles said it's too early to speculate on the fate of the IBM-NetApp relationship based on SONAS. "I don't think this is necessarily a statement on Ontap 8, because Ontap 8 isn't officially here already," he said. "GPFS could be a different use case from Ontap 8. I don't know that IBM has optimized SONAS for highly random I/O, NFS deployments behind Oracle, or the other things NetApp generally goes after, and probably will with Ontap 8."
Given its scale and object-based back end, could SONAS be more of a competitor for EMC Corp.'s Atmos cloud storage scale-out system? "I think it's somewhere in between," Boles said. "It has massive capacity and extensibility like Atmos, but at the same time, it also has a lot of NAS functionality Atmos doesn't."