Enterprise Storage: The SANs Cometh
While I don’t have an exact number, I am sure that, with all the stories, white papers, presentations and books I have written this year – not to mention the mountain of e-mail from disgruntled, er, fans of my work – I am storing pretty close to a yottabyte of data on my hard disk. The folks at YottaYotta tell me that a yottabyte equals one trillion terabytes. For lack of a meaningful equivalent, President Steve Mattioli says, "a yottabyte is more data than existed in the entire world two years ago, but less data than will exist two years from now."
Regardless, I thank Santa Claus that the year is at an end, and I can go through the annual purge of my Outlook Express inbox. I also thank Santa that I am not a company that needs to store all of my e-mail for seven years.
This is now an SEC requirement of all publicly traded firms: a safeguard to ensure that records of all transactions pursuant to a purchase of any product or service are maintained. The Justice Department likes the idea, too. They want to ensure that companies have a record of all e-mail, so that they can access any aggressive-sounding or blatantly illegal remarks made, in the event of a Judicial review.
Under the circumstances, publicly-traded companies will probably see yottabyte storage requirements a lot sooner than I will – which brings me to the point. How do you store a yottabyte?
Death of a Media
At a press event sponsored by Plasmon Inc. this past fall, I was invited to sit on a panel discussion about the future of secondary storage. Secondary storage is distinguished from primary storage, as nearly as I could tell, just because data isn’t stored on magnetic hard disks, but on tape or optical/magneto-optical media.
The press and analyst folks in attendance enojyed red carpet treatment by Plasmon. Despite this, someone had the nerve to ask the tactless question, "With the advent of SANs, wouldn’t you have to agree that tape and optical storage have no future?" I thought that our hosts would have felt, well, betrayed by the question, but they took it in stride. "Well, Mr. Toigo, what do you think? Are tape and optical dead?"
I took a moment to collect my thoughts, and then gave the response:
1. SANs do not yet exist. So, the question of SANs replacing tape and optical makes no sense.
2. Many of those storage networks (primordial SANs) that are being deployed in the field today are deployed expressly to share expensive tape automation systems, and not to replace them. By providing a back-end network for large data movements, storage networks, combined with server-less backup technologies, enable data to be backed up without interfering with production systems and LANs. That means no more struggling with backup "windows," and much more current backups.
3. As we approach yottabyte-sized storage backup requirements, or maybe, even just exabyte-sized, the time to restore data from tape following a smoke-and-rubble disaster may make tape backup less desirable. Disk-based data mirroring has always been the preferable, but much more expensive, option. As the time-based costs for data restoral increase, disk-based mirroring becomes more attractive as an option. This is true with or without SANs.
4. Having said the third, one must clarify that not all applications have the same time-to-data-restore sensitivities. Nor is it likely, for the time being at least, that "enterprise storage pools" – centralized SANs that store and share all data with all users and applications – will become a dominant architectural feature in companies, even once SANs are ready for prime time. What is more likely is that tape backup will likely be tailored, as it is today, to meet the needs of individual departments, business units or workgroups, who will not have exabyte or yottabyte storage requirements for quite some time.
5. Optical storage is another story. For the most part, magneto-optical disc (virtually indestructible glass platters about the size of a medium pizza), and to a lesser extent CD-R, CD-RW and DVD-? (pick a flavor for recording), serve niche markets today. They are preferred in situations where data must have a "high non-repudiation guarantee" – that is, no one can alter the data after it has been recorded on the media. The SEC and the Justice Department love that!
The other main use for optical is archival storage. Optical is generally considered more durable than tape for archival storage, but it is also subject to the same problem as archival tape: susceptibility to the obsolescence of the software used to record data to the media over time.
6. Thus, neither tape nor optical have anything to fear from the coming SANs. Even if they did, as things currently stand in the SAN world, tape and optical vendors can assure themselves of a pretty long runway for their products before SANs become a threat.