Data recovery software - Urgent Recovery Software v3.2

WELCOME TO URGENTRECOVERY NEWS

Hybrid Hard Disks and Beyond December, 2005 By Victor Loh

Last Tuesday, IDEMA hosted a symposium in entitled "HDD Dynamics: Interfaces, Electronics, Architecture, and Reliability". IDEMA, which stands for International Disk Drive Equipment and Materials Association, is a trade association representing 340 companies from the $30 billion disk drive industry.

At the event we attended, representatives from Adaptec, Agere Systems, Broadcom, Hitachi GST, Intel, M-Systems, Microsoft, Network Appliance, Samsung, Seagate, and Western Digital were given the opportunity to speak on issues such as emerging interfaces, new electronic designs and integration, sector architecture, and reliability testing.

One of the topics discussed at the symposium dealt with the incorporation of solid-state storage with traditional magnetic disks. Hybrid hard disk drives, as they're called, combine a magnetic hard disk drive and flash memory into a single device.

The use of solid state media in consumer and enterprise storage is already widespread: It's in our cameras, MP3 players, USB drives, and mobile phones—and its market share will continue to expand.

At the same time, the price of flash memory has been decreasing by about 30% to 50% per year. The areal density of magnetic drives continues to grow as the price per gigabyte falls. With the hybrid HDD, manufacturers seek to combine the robustness, low-power consumption, and speed of flash media with the extraordinary storage density and low cost of magnetic drives.

Microsoft's hybrid hard disk design features a non-volatile (NV) cache added to the hard drive to store frequently used sectors for fast boot and resume times. M-Systems's Director of Technical Marketing, Esther Spanjer, notes that 90% of users, on average, use only 10% of operating system sectors.By duplicating the most commonly used data onto the NV cache, you can minimize dependency on the hard drive for frequently accessed sectors and speed up performance. A cache size of at least 128MB is recommended, but drive manufacturers could conceivably

incorporate caches as large as 4GB or more to store critical system data like registry or favorites info. Microsoft's hybrid drive design is currently supported only in Vista, but wider adoption is expected as standardization of the platform expands.

Intel's "Robson" cache takes a different approach, placing the NV cache directly onto the motherboard. With the entire operating system loaded onto flash memory, booting can occur directly from the cache. Once again, a minimum 128MB cache is recommended, but 2GB to 4GB sizes could provide more flexibibility.

What do we get from all this hybrid integration? Lower power consumption, for starters. Disk reads and writes from flash allow the hard disk to stay in a power-saving state with the spindle stopped.

Samsung's Dan Barnetson estimates a 9% lower consumption overall. On a laptop, the power savings could add up to an hour of additional battery life.

Faster boot and resume is another advantage. The first 3 to 5 seconds after initially powering up a PC are spent spinning up hard drive platters and synchronizing heads before the boot process can begin. Booting straight from the flash cache eliminates this latency without sacrificing disk performance: Flash can nearly saturate the SATA protocol's 1.6 GB/sec throughput. Working models using Intel's Robson cache have been able to shorten boot time down to less than half a minute. It's the same story for resuming from either shutdown or suspend modes. Critical data is written to the flash cache before your laptop goes to sleep, hibernates, or powers off. Upon resume, it's available instantaneously.

Finally, hybrid storage solutions are more reliable. Mobile applications favor robustness, and flash, as a solid-state medium, boasts superior shock resistance. Also, lower operating duty cycles for your HDDs means less spinning, less heat, longer drive life, and less frequent disk failure.

Three strategies exist for implementing non-volatile memory into hybrid HDDs. The first involves an embedded flash disk. It's an easy, small-footprint solution that incorporates a controller and flash memory onto a single chip. It requires minimal motherboard real estate and provides a NOR interface that works with the existing HDD controller. From a design standpoint, it is much simpler than other options below. It doesn't require completely redesigning existing hard drive architecture. Management software for the flash disk can be supplied by the flash manufacturers.

Also a one-chip solution, using raw NAND flash (the kind found in CompactFlash and USB drives) with existing HDD controllers is cheaper than an embedded flash disk, but it requires additional research and development costs. With the broad array of NAND memory types, cross compatibility is not guaranteed, as many flash chips are not compatible among vendors. The software to manage the NAND memory will also need to be developed.

The third solution uses raw NAND with a dedicated NAND controller. It solves the issue of multi-source flash chips. This two-chip solution requires more PCB space than the embedded flash disks or the NAND disks with HDD controllers. It also has similar drawbacks to the NAND-and-HDD-controller design, including a lack of existing software support and additional development requirements.

HDD Sector Architecture
Increased hard disk areal density is one of the primary driving forces for the tremendous growth in HDD storage capacity over the past few decades, but we've reached a point where further bit per inch (bpi) and track per inch (tpi) growth is making reliable data retrieval difficult. Areal densities in excess of 150–200 Gbits/in² have made traditional 512-byte sectors inadequate for error-correcting code (ECC).

An IDEMA-sponsored committee found that increasing the standard sector size to 4096 bytes would provide a sufficient platform for future high-areal density drives. This larger 4K standard would allow ECC to maintain the same bit error rate (BER) without sacrificing signal to noise ratio (SNR) . Longer sectors decrease the per-sector overhead among larger swaths of data, and bit space is used more efficiently. With larger 4K sectors, drives will benefit with not only enhanced reliability and tolerance, but greater capacity as well.

The migration from 512B to 4K sectory sizes will require systematic changes that will occur over time. Production of 4K sector drives is expected to start in 2007, but a transition period will follow for a number of years after as the market moves to standardize on the 4K native block size drives. Operating systems like Windows XP, for example, will still need to support 512B emulation for the continued function of legacy drives. Vista, scheduled for release in 2006, will take another 2-3 years before becoming a mainstream OS.

According to Curtis Stevens of Western Digital, support for both 512B and 4K native sector drives in the meantime may incur some performance penalties. Mapping 512B logical sectors to larger physical sectors, for example, may require a read-modify-write (RMW) process if a host writes data to a logical block address (LBA) that is out of alignment with the physical sector boundaries. While there may be no measurable penalties for read operations or write operations that begin and end on physical sector boundaries, writes that start or end in the middle of a physical sector may suffer a slowdown. Writes within a track, for instance, may require an additional revolution. Two revolutions may be required for writes that span an entire track.

PARTNER PROGRAM

FAQ

GLOSARY

ERRORS