|
 |
| ||||||||||||||||||||||||||||
 |
sections
home features from the editors opinion interview toolkit issue index news forums about us about queue advisory board media kit about ACM privacy policy writers guidelines feedback 
|
The Most Pressing Issue and what else Might ChangeThe last vestige of that old geometric model for disk drives is the fixed 512-byte sector. This still maps directly to the physical unit written to the media. It is probably what drive manufacturers would most like to get rid of next. The use of error correction codes is an important drive component for protecting data and ensuring that it can be recovered from a disk and returned accurately. ECC information is appended to each 512-byte sector of data. This additional information means that a little more disk space is required to store the information. As data bits have become smaller, flaws have become effectively larger—that is, each flaw can corrupt more bits. The ECCs that enable the drive to recover from these flaws have had to become longer, from perhaps 16 bytes of ECC per sector in 1995 to more than double that today. To continue increasing areal density, it will get longer yet. To minimize the proportion of disk space dedicated to ECC overhead, drive vendors universally desire a longer sector size—4,096 bytes is the proposed length. Whereas the work required to support 4K sectors in a drive is straightforward, the difficulty of getting the rest of the computer system to employ it is daunting. It is not clear when this might happen, but companies are already investigating the OS and driver problems. Some research argues for exciting future changes in the programming model, including hiding the physical sector size from the host altogether. Exposing the geometry of a drive to the host operating system in a useful way is impossible, because it is subject to change and variation. When a specific method for laying out files is being determined, however, knowledge of the exact drive geometry, including information such as flaw location, would benefit performance. One of the motives behind object-based storage device (OSD) research by the Storage Networking Industry Association (SNIA) and the T10 Technical Committee of the International Committee of Information Technology Standards is the recognition that the drive could be assigned the responsibility for organizing the information storage space [refer to "Working Project Draft T10/1355-D Revision 06 Information Technology—SCSI Object-Based Storage Device Commands (OSD)," Ralph Weber, editor, Aug. 2, 2002]. It could then employ an exact geometric model, because it understands the unique specifics of its three-dimensional space and could in theory allocate data intelligently based on that knowledge. OSD would also solve the long-sector problem; it completely abstracts the underlying drive format by transferring data in byte lengths. Security is a growing concern among users. OSD research advocates an appealing security model. Recent research at MIT found that a significant amount of sensitive data could still be recovered from drives that had been "erased" and discarded ["Selling a computer? Be sure to erase hard drive files," by Justin Pope, Minneapolis Tribune, Jan. 20, 2003]. OSD includes a security model that would marry the security policy to the data at the drive level. Special information such as a unique password would always be required to decode the information stored on the media. Because the data cannot be separated from the security control, this sort of vulnerability could be avoided. Concurrency control is another area of research that could have implications with respect to the programming model. Two projects suggest that the drive, because it is at the point of convergence for accesses from multiple systems, could help make clusters more scalable [see "Scalable Concurrency Control and Recovery for Shared Storage Arrays," by Khalil Amiri, Garth Gibson, and Richard Golding, CMU-CS-99-111, February 1999; and "The Global File System," by Steven R. Soltis, Thomas M. Ruwart, and Matthew T. O'Keefe, Proceedings of the Fifth NASA Goddard Space Flight Center Conference on Mass Storage Systems and Technologies, Sept. 17-19, 1996]. Other examples include applications for disk locks such as fencing and revocation of access rights. Seagate and other drive manufacturers have participated in several aspects of this ongoing research, and most feel that many of these ideas are worthwhile candidates for future disk-drive architectures.
| |||||||||||||||||||
Previous Page (6/7)