DE102013003544A1 - Shingled-magnetic recording (SMR) disk drive for SMR device, has electronic unit which stores set of exception records as received in write commands in memory E-region, and writes set of exception records to disk E-region in sequence order - Google Patents

Abstract

The disk has magnetic thin film which is provided with concentric tracks. I-region is contained with subset of tracks that are shingle-written with tracks partially overlapping previous tracks. Disk E-region is provided with subset of tracks to hold exception records. Memory E-region is provided with exception records stored in nonvolatile solid state memory. System electronic unit is provided to store set of exception records as received in write commands in memory E-region and to write set of exception records to disk E-region in sequence order. An independent claim is included for method of operating shingled-magnetic recording disk drive.

Description

Related applications

This application is related to provisional patent application Serial No. 61 / 489,174 filed on May 23, 2011, assigned to the assignee of the present application, and the advantages of this provisional application are claimed in 35 U.S.C. 119 (e).

A patent application serial number 13/135 953, filed on July 18, 2011 and assigned to the assignee of the present application, incorporated herein by reference, describes SMR drive embodiments having write-to-write cache areas disclosed in the present application are mentioned.

Field of the invention

The invention relates to the field of data storage device architecture, and more particularly to overlapped magnetic recording devices (SMR devices).

background

Conventional disk drives with magnetic media organize data in concentric tracks that are spaced apart. The concept of overlapping writing is a form of perpendicular magnetic recording and has been described as a way to increase the areal density of the magnetic recording. In an overlapping magnetic recording medium (SMR medium), an area (band) of adjacent tracks is described as being superimposed on one or more tracks previously described. The overlapping tracks, unlike conventionally separated tracks, which may be described in any order, must be described in a sequence. The tracks on a disk surface are organized into several overlapping areas (also referred to as I-areas). The direction of the overlapping writing for an I range may be from an inner diameter (ID) to an outer diameter (OD) or from OD to ID. The plate may also be overlapping in both directions on the same surface, with the two zones meeting approximately at the middle diameter point. The number of tracks that overlap one another in one area is a key parameter for the performance of the overlapping writing. Once described in an overlapping structure, an individual track can not be updated locally because it would overwrite and destroy the data in the overlapping tracks. Overlapping data tracks are therefore sometimes considered by the user to be end-to-end writable protocols. To improve the performance of SMR drives, part of the magnetic media is mapped to one or more so-called "exception areas" (E-areas) that are used as staging areas for data that is eventually written into an I-area , The E-area is sometimes referred to as an E-cache. Because it is expected that most of the data in an SMR drive will be sequentially stored in I ranges, the records that are not currently stored in the I ranges may be considered "exceptions" for sequential I range storage be considered. When writes are received in random order, they are generally stored in the receive order in the disk E space.

The U.S. Patent 7,965,465 by Sanvido et al. (June 21, 2011) describes techniques for using a cache memory to facilitate updating the records in overlapping blocks of an SMR disk that need to be written sequentially.

Address dereferencing in the internal architecture of the overlapping memory device is useful to shield the host from the complexities associated with the SMR. Host file systems conventionally use logical block addresses (LBA) in instructions for reading and writing data blocks, regardless of the actual locations (physical block address (PBA)) used internally by the memory device. Hard disk drives have had some degree of LBA PBA dereference for decades which, among other things, allows poor sectors on the disk to be reassigned to good sectors that have been reserved for that purpose. The address dereference is typically implemented in the control section of the drive architecture. The controller translates the LBA in host commands into an internal physical address or an intermediate address from which a physical address may ultimately be derived.

The conventional LBA PBA mapping for errors does not need to be changed frequently. On the other hand, in an SMR device, the physical block address (PBA) of a logical block address (LBA) may change frequently. The dereferencing system provides a dynamic translation layer between host LBA and the actual physical locations on the medium. In an SMR system, the LBA-PBA mapping may change each time a write occurs because the system dynamically determines the physical location on the medium where the host data for an LBA is written. The data for the same LBA will be written to another location the next time the host LBA is updated. In addition, the drive automatically moves data between write cache memories in RAM, write cache memories on the disk, e-slots on the disk, and I-slots on the disk. The LBA for the data remains the same wherever the data is stored on the drive. Background processes, such as defragmentation, are also performed by the device autonomously to move data sectors from one PBA to another while keeping the LBA constant.

Defragmentation is a generic term commonly used to describe a process of reorganizing records in a file or database system to eliminate or reduce fragmentation. In SMR I ranges, as records are updated or deleted, the number of small vacancies that are commonly referred to as invalid or "expired" data increases. The process of defragmentation physically moves the datasets to make them more coherent and to create larger and more useful free areas. DRAM is typically used in restaging because it allows efficient sorting of records into a suitable sequence. For SMR drives, efficient defragmentation is an important factor in the overall performance of the device.

Summary of the invention

Embodiments of the invention include SMR disk drives having a non-volatile semiconductor memory E-area in addition to the magnetic media E-areas on the disk. The combination of a semiconductor memory and plate E regions are referred to as hybrid E regions. According to embodiments of the invention, the semiconductor memory E-area may be used for operations referred to as destaging and / or restaging. The memory E area can be used to sequenced sets of exception records and thereby reduce the time and energy spent in performing disk E field searches. The ratio between the size of the semiconductor memory E-region and the total capacitance of the E-region on the plates can be optimized according to the invention for selected applications, using tradeoffs between performance and cost. For example, an embodiment having a storage area of the E-area that is 10% of the total capacity of the disk E-area, achieves a significant performance improvement over a disk-only E-space implementation, and also results in a lower cost than a Nur NAND memory E range would be required.

In one embodiment, a set of incoming write exceptions is first stored in the semiconductor memory E area and then rearranged before being destaged to a disk E area. The use of the semiconductor memory E-area allows rearrangement to improve sequencing and reduce the number of searches required when transferring the records to disk.

According to another embodiment, the nonvolatile semiconductor memory E-area is used for restaging data that has been previously written to the disk and that needs to be rearranged and then written back to disk, such as in a defragmentation process. The embodiments of the destaging and restaging can also be used together.

Brief description of the figures

1 FIG. 12 is an illustration of a hybrid E-domain SMR data storage device according to one embodiment of the invention. FIG.

2 Figure 4 is a graph showing continuous IOPs on the vertical axis versus read rate (seeks per second) on the horizontal axis for an SMR drive with a 4 drive disk-only E area in accordance with an hybrid E-range embodiment of the invention with NAND flash sizes of 1GB, 2GB, 4GB, and 8GB.

3 Figure 12 is a graph showing the size of the E-range required to match the PMR performance using a disk-only E range and a hybrid NAND E-range for destaging according to the invention.

4 Figure 12 is a bar graph showing the percentage of the size of the disk E area in user capacity for NAND E range and DRAM reorder buffer devices of different sizes and with different transmission lengths.

5 Figure 12 is a graph showing the sequential read power in MB / s for randomly written areas for a conventional PMR drive, a disk-only e-disk SMR drive, and a hybrid e-disk drive according to the invention.

6 Figure 12 is an illustration of the reordering of data blocks during the NAND flash E range to disk E range, according to an embodiment of the invention.

7 Figure 12 is an illustration of the use of the NAND flash E area in restaging blocks of data from a disk E area to facilitate reordering blocks of data during defragmentation of a disk E area in accordance with an embodiment of the invention.

8th Figure 12 is an illustration of the destage from the NAND flash E area of the rearranged blocks of data back to the disk E area during defragmentation of a disk E area according to one embodiment of the invention.

Detailed description of the invention

1 is an illustration of a data storage device (DSD) 10 using SMR with a system electronics unit 21 according to an embodiment of the invention. The system electronics unit 21 with the exception of what is described here, according to the prior art, and aspects of the system that perform functions of the prior art are not shown. The system electronics unit 21 may be a prior art system-on-a-chip unit, which is an integrated circuit including the host interface, the controller, servo functions, a microprocessor, firmware programs, etc. all in having a single chip.

The host user 11 For example, a computer may be of any type and communicate with the device by any means, such as through a network. The term "user" is used interchangeably with "host". Multiple hosts may also communicate with the device using known techniques. thin films 12 are magnetic thin film coatings typically deposited on both the top and bottom surfaces of a hard disk (not shown), and a device may also include multiple disks. The movies 12 are in 1 shown in a sectional view. In a plan view, the areas are a plurality of concentric circular bands. The magnetic thin films are formatted for use in an SMR architecture and according to this embodiment have a disk E area 16 , I areas 13 (also known as I-track areas), write cache areas 14 (also referred to as twice-writable cache areas) and protection areas or bands 15 on. Although in 1 only one is illustrated, a device may have multiple E-areas on each plate surface 16 and there are usually several plate surfaces. The nonvolatile semiconductor memory E range 19 is commonly referred to as NAND Flash E-Area or NAND-E Area. The NAND flash E range 16 is used in accordance with the invention to improve performance and save energy over a disk-only E-area implementation. In addition, this will be due to the size of the NAND E range 16 is designed to be considerably smaller than would be required for a NAND-only E range, saving costs.

A casual analogy to the problem solved by the invention can be made for mail delivery. Incoming letters are random and must be sorted into increasingly specific destinations such as country, region, state, postal code, postal delivery route, delivery order. It would be very inefficient if a postman happened to shuttle back and forth and deliver mail regardless of the address.

In this way, in this analogy, the write exceptions are received as letters in a substantially random order, however, writing them to a disk in the input order later, when the exceptions need to be rearranged, results in a high number of times and resource-intensive searches. The exceptions must be organized in various ways before they can be effectively written into an I-lane in an I-domain. High-level pre-sorting by zip code could be compared with organizing exceptions after the Goal I range. Other levels of sorting exceptions include an LBA sequence. The sorting of the exceptions according to the invention results in a continuous write performance that is highly independent of the search speed because the number of searches is significantly reduced.

Unlike what is described herein, general destaging and restoring algorithms can be used to manage data transfers between the NAND E-area and the disk E-area that are going back and forth. In the following description is the for the E-range 19 used non-volatile semiconductor memory commonly referred to as NAND flash, because this is the preferred write buffer based on current technology. However, any type of nonvolatile semiconductor buffer memory may be used with the invention, such as a phase change memory, a NOR flash, and / or an MRAM. The use of the term "NAND" is for convenience only and is not intended to limit the embodiments of the invention.

Destaging using the Hybrid E-Area

The use of the NAND Flash E range 19 as a write buffer allows a set of exceptions to be generally reordered into the sequence in which they reside in the disk e-space 16 be destaged. It is preferable to accumulate a large set of exception records for sorting before writing to the disk E area. Therefore, the destaging should generally be triggered when the flash e-area 19 is full to maximize the number of exceptions per I range when destaged. This reduces the number of searches (read-ups) during a defragmentation.

6 Figure 12 is an illustration of rearranging exception data blocks during NAND flash E range destesting 19 to the plate E area 16 according to an embodiment of the invention. Over time, the host may send write records in a sequential order, an inverse sequential order, a different non-random order, or in a substantially random order. The invention provides the highest gain over the prior art when the records are received in random order. In the example off 6 It is assumed that the LBA blocks on the left are received in a substantially random order from the host / user and are placed in the NAND flash E range. Although the exception records can be sorted into a sequence in the NAND flash E range, the physical order that the exception LBA blocks have within the flash E range is not critical because no mechanical, time-consuming searches are required to move from entry to entry. The important point is that the set of LBA blocks is destaged in the disk E area in an improved sequential order. This means that the plate E range 16 does not contain the raw random writes as in the prior art SMR. The initial reordering of a set of random host writes (exception records) increases the effective reuse rate of the individual exceptions. The number of searches is reduced, but this is to some extent independent of the rate of re-reading.

At this "skip sequentials", a burst read operation is performed after the restaging / defragmenting. The invention enables a significant reduction in the number of random searches during the disc E disk defragmentation process 16 , The reordering of the set of exception records before or while they are in the disk E space 16 is executed by the system electronics and can be achieved by executed by a microprocessor firmware programs.

Because more random read operations are required in a disk-only E-space design, it has the following disadvantages: a) poor short search mechanics, b) higher sort overhead, c) insufficient SRAM for large internal queues, and d) higher power consumption. In contrast, the invention results in fewer random searches when reruns are performed during a restore / defragmentation, and therefore, it improves on any of the problems with the panel-only e-scope design discussed above, including allowing for significant energy savings. Although a DRAM may be used instead of the NAND flash, there are advantages to using a NAND instead of the DRAM. NAND is about 10 times cheaper per MB than DRAM, and NAND, unlike DRAM, is non-volatile.

The hybrid E-zone design is also less expensive than a 100% replacement of the plate E-sections by a NAND flash E-section. The invention enables the size of the NAND E region to be significantly smaller than the overall size of the plate E region. For example, as shown in the data below, using a size of the NAND E range that is about 10% of the total disk E space size provides improved performance. Using a 10% size NAND E range also reduces the incremental cost to 10% of a NAND only E range design. Performance estimates (see Table 1) show that a 2 GB system for NAND destaging in a hybrid E-space design can achieve 498 IOPs, comparable to a 32 GB system of an NFC. NAND E range (421 IOPs). The estimates in Table 1 assume a user capacity of 2 TB and 4k RW IOPs. IOPs stands for I / O operations per second. NAND capacity (GB) % versus provision NAND D area IOPs NAND destage IOPs 2 0.11% 30 498 4 0.21% 59 568 8th 0.43% 116 626 32 1.72% 421 783 64 3.44% 747 954 Table 1.

The current perpendicular magnetic recording (PMR) drive technology performs random writes at random locations, requiring more searching and more latency before each write. SMR drives with hybrid E-arrays have faster random write IOPs because no random searches are required during user data writes. New user data is written either sequentially in an I-lane in an I-range or in a disk E-range in an SMR-drive.

2 Figure 12 is a graph showing the continuous IOPs on the vertical axis versus the read rate (seeks per second) on the horizontal axis for an SMR drive with a disk-only E range and for 4 drives with hybrid E-ranges with NAND flash Sizes of 1GB, 2GB, 4GB and 8GB. Each of the 4 examples of hybrid E-arrays results in higher IOPs over the rewrite range of 200 to 1000 searches per second than in the panel-only E range. Although a higher NAND flash size results in better performance, any doubling in flash size will result in a smaller increase than the previous one.

The sequencing of exceptions according to the invention removes almost completely the effect of the search speed from the continuous writing power. The NAND E-Area Destage ("NED") achieves equivalent performance to the NAND-only E-Area design and uses 80% -90% less NAND. Although the same performance advantage is possible when using a correspondingly large DRAM E-area buffer instead of the NAND, the NAND-E area has advantages over a DRAM.

The graphic in 3 compares the performance of long bursts when using the disk-only E-range with the use of a hybrid NAND-E range for a destaging according to the invention. The horizontal axis is a transmission length in 1 k units. The estimates assume a PMR performance level, a command queue depth (QD) of 32, and Write Cache Enabled (WCE) = 1, ie WCE On. The vertical axis is the percentage of the size of the disk E area versus the user capacity. If the size of the disk E area is allowed to be expanded, the rate of re-learning has a strong influence on the size of the E-range required to reach the same power level as the PMR. The extension of the E-range allows the re-reading to remain dominant. For a given exception read rate, the requirement of the size of the E-range doubles each time the transmission length is doubled. This keeps the average number of exceptions per track constant.

The top 3 lines in the graph are for read rates of 200, 500, and 1000 seeks per second for the disk-only E range. The lower line is provided for the hybrid E range according to the invention. With a 1k transmission length and a rewrite rate of 200, the The present invention assumes, with the above assumptions, a size of the disk E area that is about 10% of the size of a disk-only E range required to meet PMR performance. The advantage of the hybrid E-range decreases as the transmission length increases, but it is never worse than a disk-only E range. For drives that have or want low read rates, a NAND destage ("NED") can be used to reduce disk E-area requirements, making it possible to allocate a larger portion of the capacity of the drive to I-sections Assign user data storage. (Note: Lower read rates in one drive than another may result from design trade-offs, for example, to save space, to minimize search performance, or to reduce costs.) One example would be the selection of smaller, less powerful voice coils used for the Movement of the actuator are responsible.)

Restaging using the Hybrid E-Area

The NAND flash E range 19 may also be used for restaging (eg I-span or disk E-zone defragmentation) as well as initial reordering (sequencing) prior to destaging as described above. 7 is a representation of the use of the NAND flash E range 19 when restoring blocks of data from a disk E area 16 to facilitate the reordering of data blocks during defragmentation of a disk E area according to an embodiment of the invention. 8th Figure 12 is an illustration of the destage from the NAND flash E area of the rearranged blocks of data back to the disk E area during defragmentation of a disk E area according to one embodiment of the invention. The restaging can be achieved by the system electronics with firmware programs executed by a microprocessor. When the disk E area is defragmented, the use of the flash E area allows the records to be read from the disk E area in the physical order in which they occur to minimize seek time. Once the records are in the Flash E area, they can be easily sorted using a selected LBA sequence-like criterion.

• • Verringern der DRAM-Anforderungen (DRAM ist 10 Mal kostspieliger als NAND-Flash), beispielsweise von 16 MB auf 64 MB DRAM. Die Größe des DRAM-Puffers spielt eine große Rolle bei den Größenanforderungen des Platten-E-Bereichs.
• • Das Restaging ist eine zusätzliche Gelegenheit für das Umordnen von Ausnahmen vor dem I-Spur-Neuschreiben.

The benefits of restaging using the NAND E range include:

• • Reduce DRAM requirements (DRAM is 10 times more expensive than NAND flash), for example from 16 MB to 64 MB DRAM. The size of the DRAM buffer plays a large role in the size requirements of the disk E area.
• • Restaging is an additional opportunity for reordering exceptions before I-lane rewriting.

The bars in the graphic in 4 show the percentage of the size of the disk E area in user capacity for devices with different parameters. Each set of 3 bars corresponds to the selected NAND E range and DRAM read buffer sizes shown below the bars. For example, the first set of bars on the left is for NAND = 2 GB and DRAM = 32 MB. In each set of bars, the bars from left to right correspond to 128k, 64k and 32k transmission lengths. The results show that the DRAM buffer size plays a large role in the size requirements of the plate E region. Therefore, it may be advantageous to rewrite exceptions just before defragmentation into the NAND E range. For example, a smaller DRAM may be included in the design by the use of the NAND E-range instead of the DRAM for restaging.

In cases where the disk E area is much larger than the NAND available for destaging, it may be advantageous to store exceptions just prior to defragmentation back into the NAND with them. Fusion with the I-lane would still occur with extensive use of the DRAM. For example, if the E-Area is 10 times larger than the NAND Destage Buffer, an average of 10 searches are required to recover all disk-E exceptions for I-lane prior to defragmentation. Multiple I-traces of exceptions should be re-scanned for each search, increasing buffering requirements and reducing the re-read cost per I-lane.

The graphic in 5 Figure 14 shows the sequential read power in MB / s for randomly-described areas in a conventional PMR drive, a disk-only e-disk SMR drive, and a hybrid e-disk drive according to the invention. The horizontal axis shows the transmission length. The results show that NAND E-area destaging according to the invention reduces the number of random read operations required to host sequential reads as compared to a disk-only E-space SMR drive to comply with randomly written data.

Alternative embodiments and optional features

Several optional features or optimizations that may be incorporated into various embodiments of the invention will now be discussed. One option is the plate E range 16 in the vicinity of the outer diameter of the disk to use the higher number of bits per revolution, the hard disk drives have on the outer diameter. The tracks near the outside diameter are longer and last about twice the data per revolution than the inside diameter of typical plates.

Another option is to select the total disk E-Area storage as about 3% of the total disk space, for example, 30 GB for a 1 TB SMR drive. Because plate E area lanes come at the expense of I area lanes, an advantage of the invention is that the size of the disk E area may be lower than would otherwise be the case. The incremental benefits of higher E-range capacity decrease when larger NAND storage capacity is used (see 2 ).

As suggested above, choosing the total NAND E-Area Memory as approximately 10% of the total disk E-Area Memory or 0.3% of the total disk space is an attractive option to greatly enhance performance reasonable costs. If the total disk E area memory is selected to be about 3% of the total disk space and the NAND E area is 10% of it, then the NAND E area accounts for 0.3% of the total disk space.

Another option to consider is that the DRAM can be used to buffer incoming writes before writing to either the NAND and / or the disk. DRAM and SRAM memories are used by the disk controller to help reorder and move user data.

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• US 7965465 [0005]

Claims (14)

Disk drive with overlapping magnetic recording, comprising: a magnetic thin-film-coating plate on which data is stored in a plurality of circular concentric tracks, an I-domain having a first subset of tracks written in sequence overlapping, where tracks partially overlay previous tracks, a disk e-space having a second subset of tracks containing exception records; a memory E area containing exception records stored in a nonvolatile semiconductor memory, and System electronics which first store a set of exception records as received in write commands in a first order in the memory E area and then place the set of exception records in a sequential order other than the first order in the disk array. E-area writes. The disk drive of claim 1, further comprising a plurality of disk E areas having a total memory size of the disk E area, and wherein the size of the memory E area is significantly smaller than the total memory size of the disk E area. The disk drive of claim 2, wherein the size of the memory E area is about 10% of the total memory size of the disk E area. The disk drive of claim 2 wherein the sequential order results in less search time being required to write the set of exception records than would be required to write the set of exception records in the first order as received. The disk drive of claim 1, wherein the sequential order results in less search time being required to write the set of exception records than would be required to write the set of exception records in the first order as received. The disk drive of claim 1, wherein the system electronics reads the disk E area by restaging a second set of exception records from the disk E area to the memory E area in a first order as read from the disk E area. defragments and then writes the set of exception records in a sorted sequential order to a selected criterion of the second set of exception records in the disk E area. A method of operating a magnetic overlap magnetic disk drive comprising: Storing a set of exception records in a memory E-area located in a non-volatile semiconductor memory, the set of exception records being received in write commands in a first order, and Writing the set of exception records into a disk E area in a sequential order different from the first order. The method of claim 6, wherein the sequential order results in less search time being required to write the set of exception records than would be required to write the set of exception records in the first order as received. A method of operating a disk drive according to claim 8, wherein the disk drive further comprises a plurality of disk E areas having a total storage size of the disk E area and wherein the size of the storage E area is significantly smaller than the total storage size of the disk E range. A method of operating a disk drive according to claim 9, wherein the size of the memory E area is about 10% of the total memory size of the disk E area. The method of claim 10, wherein the sequential order results in less search time being required to write the set of exception records than would be required to write the set of exception records in the first order as received. The method of claim 7, further comprising: Reading a second set of exception records from the disk E area into the memory E area, the second set of exception records being read in a physical order from the disk E area, and writing the second set of exception records into the memory E area Disk E area in a sequential order different from the physical order, the sorted sequential order being a sorted order for a selected feature of the second set of exception records. The method of claim 12, wherein the sorted sequential order results in less search time being required to write the set of exception records into an I-range than would be required to write the set of exception records in physical order. The method of claim 7, further comprising sorting the set of exception records in a sequential order in the memory E area before writing the set of exception records in sequential order to the disk E area.

Images