Inside System Storage -- by Tony Pearson

With all the announcements we had in June, it is easy for some of the more subtle enhancements to get overlooked. While I was at Orlando for the IBM Edge conference, I was able to blog about some of the key featured announcements. Then, later, when I got back from Orlando to Tucson, I was able to then blog about [More IBM Storage Announcements]. For IBM's Scale-Out Network Attach Storage (SONAS), I had simply:

In my defense, IBM numbers its software releasees with version.release.modification, so 1.3.2 is Version 1, Release 3, Modification 2. Generally, modification announcements don't get much attention. The big announcement for v1.3.0 of SONAS happened last October, see my blog post [October 2011 Announcements - Part I] or the nice summary post [IBM Scale-out Network Attached Storage 1.3.0] from fellow blogger Roger Luethy.

Here is a diagram showing the three configurations of SONAS.

I have covered the SONAS Appliance model in depth in previous blogs, with options for fast and slow disk speeds, choice of RAID protection levels, a collection of enterprise-class software features provided at no additional charge, and interfaces to support a variety of third party backup and anti-virus checking software.

The basics haven't changed. The SONAS appliance consists of 2 to 32 interface nodes, 2 to 60 storage nodes, and up to 7,200 disk drives. The maximum configuration takes up 17 frames and holds 21.6PB of raw disk capacity, which is about 17PB usable space when RAID6 is configured. An interface nodes has one or two hex-core processors with up to 144GB of RAM to offer up to 3.5GB/sec performance each. This makes IBM SONAS the fastest performing and most scalable disk system in IBM's System Storage product line.

I thought I would go a bit deeper on the gateway models. These models support up to ten storage nodes, organized in pairs. The key difference is that instead of internal disk controllers, the storage nodes connect to external disk systems. There is enough space in the base SONAS rack to hold up to six interface nodes, or you can add a second rack if you need more interface nodes for increased performance.

SONAS with XIV gateway

XIV offers a clever approach to storage that allows for incredibly fast access to data on relatively slow 7200 RPM drives. By scattering data across all drives and taking advantage of parallel processing, rebuild times for a failed 3TB drive are less than 75 minutes. Compare that to typical rebuild times for 3TB drives that could take as much as 9-10 hours under active I/O loads!

In the configuration, each pair of storage nodes can connect to external SAN Fabric switches that then connect to one or two XIV storage systems. How simple is that? These can be the original XIV systems that support 1TB and 2TB drives, or the new XIV Gen3 systems that support 400GB Solid-state drives (SSD) and 3TB spinning disk drives. In both cases, you can acquire additional storage capacity as little as 12 drives at a time (one XIV module holds 12 drives).

The maximum configuration of ten XIV boxes could hold 1,800 drives. At 3TB drive per drive, that would be 2.4PB usable capacity.

The SONAS with XIV gateway does not require the XIV devices to be dedicated for SONAS purposes. Rather, you can assign some XIV storage space for the SONAS, and the rest is available for other servers. In this manner, SONAS just looks like another set of Linux-based servers to the XIV storage system. This in effect gives you "Unified Storage", with a full complement of NAS protocols from the SONAS side (NFS, CIFS, FTP, HTTPS, SCP) as well as block-based protocols directly from the XIV (FCP, iSCSI).

SONAS with Storwize V7000 gateway

The other gateway offering is the SONAS with Storwize V7000. Like the SONAS with XIV gateway model, you connect a pair of SONAS storage nodes to 1 or 2 Storwize V7000 disk systems. However, you do not need a SAN Fabric switch in between. You can instead connect the SONAS storage nodes directly to the Storwize V7000 control enclosures.

To acquire additional storage capacity, you can purchase a single drive at a time. That's right. Not 12 drives, or 60 drives, at a time, but one at a time. The Storwize V7000 supports a wide range of SSD, SAS and NL-SAS drives at different sizes, speeds and capacities. The drives can be configured into various RAID protection levels: RAID 0, 1, 3, 5, 6 and 10.

Each Storwize V7000 control enclosure can have up to nine expansion drawers. If you choose the 2.5-inch 24-bay models, you can have up to 480 drives per storage node pair, for a total of 2,400 drives. If you choose the 3.5-inch 12-bay models, you can have up to 240 drives per node pair, 1,200 drives total. At 3TB per drive, this could be 3.6PB of raw capacity. The usable PB would depend on which RAID level you selected. Of course, you don't have to limit yourself all to one size or the other. Feel free to mix 2.5-inch and 3.5-inch drawers to provide different storage pool capabilities.

All three SONAS configurations support Active Cloud Engine. This is a collection of features that differentiate SONAS from the other scale-out NAS wannabees in the marketplace:

• Policy-driven Data Placement -- Different files can be directed to different storage pools. You no longer have to associate certain file systems to certain storage technologies.
  
  
• High-speed Scan Engine -- SONAS can scan 10 million files per minute, per node. These scans can be used to drive data migration, backups, expirations, or replications, for example. It is over 100 times faster than traditional walk-the-directory-tree approaches employed by other NAS solutions.
  
  
• Policy-driven Migration -- You can migrate files from one storage pool to another, based on age, days since last reference, size, and other criteria. The files can be moved from disk to disk, or move out of SONAS and stored on external media, such as tape or a virtual tape library. A lot of data stored on NAS systems is dormant, with little or no likelihood of being looked at again. Why waste money keeping that kind of data on expensive disk? With SONAS, you can move those files to tape can save lots of money. The files are stubbed in the SONAS file system, so that an access request to a file will automatically trigger a recall to fetch the data from tape back to the SONAS system.
  
  
• Policy-driven Expiration -- SONAS can help you keep your system clean, by helping you decide what files should be deleted. This is especially useful for things like logs and traces that tend to just hang around until some deletes them manually.
  
  
• WAN Caching -- This allows one SONAS to act as a "Cloud Storage Gateway" for another SONAS at a remote location connected by Wide Area Network (WAN). Let's say your main data center has a large SONAS repository of files, and a small branch office has a smaller SONAS. This allows all locations to have a "Global" view of the all the interconnected SONAS systems, with a high-speed user experience for local LAN-based access to the most recent and frequently used files.

If you want to learn more, see the [IBM SONAS landing page]. Next week, I will be across the Pacific Ocean in [Taipei], to teach IBM Top Gun class to sales reps and IBM Business Partners. "Selling SONAS" will be one of the topics I will be covering!

technorati tags: IBM, SONAS, gateway, disk, tape, scale-out, NAS, RAID, rebuild