Sun's 7000 series: The future of storage?

In the immortal words of Kent Brockman, I, for one, welcome our new storage overlords

Today Sun unveiled the Sun 7000-series Unified Storage System. Three models were announced: the low-end 7110, the midrange 7210, and the high-end 7410. All are built on OpenSolaris and ZFS, and the 7210 and 7410 leverage SSD technology to speed up I/O. Luckily, I happen to have a 7210 in my lab currently and will be publishing a full review of the unit in the very near future. For now, however, I can talk a bit about what these units actually represent.

The 7210 is built on the Sun x4500 "Thumper" platform, which incorporates 48 SATA drives in a 4U chassis alongside a server mainboard running OpenSolaris. The 7210 isn't technically a server though, it's an appliance -- no shell or root access to the system is possible for the moment, and thus you won't be running any custom applications on the device. What the 7210 and its brethren provide is an elegant storage framework that includes the best-of-breed ZFS file system coupled with read- and write-biased SSDs, support for CIFS, iSCSI, NDMP, FTP, (and others) driven by a truly breathtaking interface.

The SSDs are used explicitly for caching and logging, and only the 7410 offers both -- the 7210 has read-biased SSDs, and the 7110 doesn't have SSD support. In discussions with Sun engineers, they claimed that the addition of the read-biased SSD caching in conjunction with ZFS' predictive caching algorithms means that 7,200-rpm SATA drives perform just as well, if not better than 10K SAS drives. In fact, they're conducting trials to determine if they can use 4,200-rpm SATA drives in these devices without sacrificing I/O performance. If that's possible, then the price point, power consumption, and heat generation drops across the board. Either way, I have to say that I've been a fan of ZFS for some time now, and it's only getting better and better.

I'll get into the specifics of the 7210 in my review, but as far as the interface goes, it's really impressive to be able to quickly pull up real-time user- and file-level throughput statistics complete with live graphs.

Watching the evolution of storage over the past few decades, it's becoming obvious that IT has overcome tremendous challenges in terms of storage technology, all the while building new problems that will need to be dealt with over the next decade. It's a tremendous advance to go from circa-1998 DEC Alpha-driven Network Appliance filers that took three racks of 9GB SCSI drives to reach a terabyte all the way up to a single 1.5TB SATA drive, but it's also a tremendous problem to back up, catalog, and secure the simply enormous amounts of storage present on just about every network. Tape has always seemed archaic, and even with the newer high-capacity Ultrium drives, there's still plenty of problems with backing up to tape, not the least of which is cost. It's generally cheaper to buy two identical storage solutions and replicate them than it is to continually buy and store tapes over time. That said, it's definitely not easier to locate those storage solutions in different physical locations without expensive pipes connecting them.

The other problem is that when you give essentially unlimited storage to a set of users, they will always use it. The saying "A project will tend to expand to consume all available resources" is getting out of hand now, and I'm guilty of that as well, looking at my 500GB home directory.

It used to be that resources were so tight that Unix commands were truncated to save resources (cp, rm, ls, and so forth), and code had to be as concise and elegant as possible for the same reason. That's not nearly as much the case anymore, as evidenced by the 16GB base installation of Microsoft Vista and Office Premium. But even that's not as serious an issue as dealing with per-user storage bloat. When every user had a limit of, say, 25MB on a central file server, locating small but important files was easier. Now, wading through a few hundred gigs of storage to find a very important 100K Word document is a chore, and nonindexed file searches take far longer than ever.

At the same time, the new capabilities offered with products like the Sun 7000-series storage appliances are paving the way for new takes on old ideas, such as video-on-demand. Let's just say that a 48TB 7210 can store quite a few HD movies, and there's no such thing as an overabundance of scratch space when working with audio and video projects.

So, in the immortal words of Kent Brockman, I, for one, welcome our new storage overlords.