Really big data

By Anne Skamarock, Network World |  Storage

If you think you have storage issues, consider this: IBM is working with Seitel, a leading provider of seismic data for the oil and gas industry, to put together a storage-area network with an initial size of one petabyte. Wow, one petabyte of data!

To give you an idea just how big a petabyte is, it equals 1 billion (1,000,000,000) megabytes, or roughly 1 quadrillion (1,000,000,000,000,000) bytes. Another way to look at this is that it would take about 15,000 of the hard drives found in most home PCs these days to store that much data.

Roy Williams of California Institute of Technology maintains a Web page that tries to give examples of some of these large amounts of information:

http://webpages.shepherd.edu/TMCGIL01/one.htm

Taking one of his examples, the entire printed collection of the U.S. Library of Congress contains about 10 terabytes of information. A petabyte is 100 times larger than that!

While most companies are not currently trying to access petabytes of information, there are many industries that are getting very close. With the amounts of information moving to digital format, it won’t be long before all companies of any size will have to wrestle with these problems. So what does it mean to manage and access a petabyte of data?

First of all, let’s look at the physical space it would take to house a petabyte. If we use Seagate’s highest-capacity Fibre Channel Cheetah drives, soon to be available with formatted capacities of 180G bytes per drive, it would take approximately 5,600 of those drives to store a petabyte. If you stacked them all on top of each other, you would have a tower about 467 feet high. Or it you stacked them 10 high, you would need about 1,120 square feet of floor space. That is just to house the disk drives. Most storage systems come with power and cooling as well.

The next question is: How much time would it take to move a petabyte of data on a SAN? I have decided to take a leap of faith and say, for simplicity’s sake, we can get transfer rates on and off the disk and through a Fibre Channel SAN at 100 megabytes/second. I will note, however, that most applications running on Unix or Windows NT systems today cannot drive data that fast, so your results certainly will differ. To move a petabyte of data at a rate of 100M byte/sec will take nearly 116 days of constant, uninterrupted streaming.

Finally, I thought about how much it would cost to buy a petabyte of simple capacity (not fancy RAID or management software). If we say we can get capacity at \$0.04 a megabyte, it would cost \$40 million for one petabyte. Most "managed" storage costs between \$0.25 and \$0.35. At \$0.30 per megabyte, that petabyte of storage would cost \$300 million -- again, just for the storage -- in, perhaps, a RAID configuration.

StorageWhite Papers & Webcasts

White Paper

White Paper

White Paper

Smart Growth, Smart I.T. at Multi-Color Corp

See more White Papers | Webcasts