Operation of perpendicular magnetic recording. In contrast to longitudinal recording, the magnetic field has to enter the recording track twice. 

Context in source publication

Context 1

... shingling is a new approach to improving data density on disks, it builds on existing technology, and disks that use it must, at least initially, fit into roles filled by traditional disks. In this section, we describe approaches that disk designers are using to improve density, and the roles that disks using such technologies must fill. The dramatic improvements in disk data density will eventually be limited by the superparamagnetic effect, which creates a trade-off between the media signal-to-noise ratio, the writeability of the media by a narrow track head, and the thermal stability of the media; Sann et al. call this the media trilemma [28]. Various approaches to this problem have been proposed; of these, shingled writing offers perhaps the most elegant and currently realizable solution, but there are other technologies that may also increase disk density. One possible approach to address the superparamagnetic limit is to radically change the makeup of the magnetic layer, as is done in Bit Patterned Media Recording (BPMR) [24]. BPMR stores individual bits in lithographically defined “mag- netic islands.” This approach faces several design problems, however. The most obvious is the challenge of manufacturing the surfaces themselves to have such islands. An additional non-trivial challenge, however, is the requirement that writes must be synchronized with the locations of the islands. A second approach to increasing density involves temporar- ily changing the receptivity of a standard magnetic layer by “softening” the magnetic material, making it easier to magnetize. This can be done with microwaves ( Microwave Assisted Magnetic Recording —MAMR, [34]) or by heat- ing the writing area with a laser ( Heat Assisted Magnetic Recording —HAMR [3], [16], [27], or Thermally Assisted Magnetic Recording [18]). Making the magnetic media easier to magnetize allows the use of a smaller magnetic field, and can also allow smaller areas to be magnetized, since lasers and microwaves can be focused on small areas to restrict the “softening” to a smaller region than that covered by the magnetic field. Unfortunately, both of these approaches offer significant challenges in construction and mechanical design, and differ significantly from existing magnetic disk designs. Because of the difficulty in manufacturing such devices, disk designers have been pursuing other approaches to increase data density. In contrast to these techniques which require radical changes to the structure of the underlying media, shingled writing builds directly upon existing magnetic recording technologies. A fundamental problem in magnetic recording is the control of the magnetic field whose flux emanates from the write head and must return to it without erasing previously written data. While perpendicular recording allows much more stable mag- netization of the magnetic grains, it complicates engineering the magnetic field for writing because the flux has to enter through the recording media in order to do its desired work, but also has to return back through it to the head. In order to protect already stored data, the return flux needs to be sufficiently diffused, limiting the power that the magnetic field can have, as shown in Figure 1. Shingled writing addresses this problem by allowing data in subsequent, but not prior, tracks to be destroyed during writes. Shingled writing uses a write head that generates an asymmetric, wider, and much stronger field that fringes in one lateral direction, but is shielded in the other direction. Figure 2 shows a larger head writing to track n , as used by Greaves et al. in their simulations [7]. Because of the larger pole, the strength of the write field can be increased, allowing a further reduction of the grain size because the technique can use a more stable medium. The sharp corner- edge field brings a narrower erase band towards the previous track, enabling an increase in the track density. Shingled writing overlaps tracks written sequentially, leaving effectively narrower tracks where the once-wider leading track has been partially overwritten. Reading from the narrower remaining tracks is straightforward using currently-available read heads. Taken together, the smaller grain size and increased track density result in an areal density increase by a factor of at least 2.5 [30] and possibly higher (3–5) according to our industry sources. Greaves et al. modeled shingled writing and found a maximum density of 3 Tb/ in 2 [7], a nominal increase of 3 × over the superparamagnetic limit of 1 Tb/ in 2 . Without III. shingled I NTEGRATION writing, OF avoiding SWD S interference INTO S YSTEMS with, or erasure SWDs of, adjacent are an tracks ideal limits replacement the maximum for tape storage for backup density and of a archival device. data, Two-Dimensional thanks to the largely-sequential Magnetic Recording nature (TDMR) of writes [4], to [12], these [13], devices. [28] can Moreover, be used the in addition ability of to SWDs shingled to be writing read randomly to place tracks makes even them closer more together attractive by than changing tapes for inter-track archival and interference backup. from However, an obstacle to be to most an instrument. useful and TDMR to provide reads a from mass several market, adjacent SWDs tracks must and remain uses inter-track capable of interference serving as primary to decode random-access the signal from storage the target devices. track. Kasiraj It uses et more al. propose sophis- organizing ticated signal the processing disk into bands [13], , where [33] and each write band encodings. stores a single In a file traditional such as disk a large architecture multimedia with file [11]. a single Bands read are head, separated reading by a a gap single of k sector tracks, with so that TDMR a write involves to the last reading track the in a sectors band does on not adjacent destroy tracks, the data requiring in the additional first track disk of the rotations. subsequent To band. avoid While this problem, we agree TDMR that some disks type could of banding use multiple is necessary read heads to store on the the bulk same of slider, the data thus in restoring a SWD, traditional there are many read ways service of times. using banding TDMR presupposes to manage data shingled in a writing, SWD and but other shingled design writing issues can to be used without TDMR. In our view, the viability of Shingled Write Recording (SWR) depends mainly on integrating SWDs into current storage and computer systems, whereas much research and development is still needed to assess the viability of TDMR. Shingled writing can be used alone or in conjunction with other new magnetic recording technologies. Shiroishi et al. recently proposed a possible transition path to incorporate these future technologies [29]. While perpendicular magnetic recording (PMR) reaches densities of up to 1 Tb/ in 2 , the next generation of technologies might use BPMR in conjunction with HAMR or MAMR and Shingled Write Recording, with a transitional use of Discrete Track Recording (DTR) as a predecessor to BPMR to reach 5 Tb/ in 2 . With TDMR in the mix, they see the possibility of densities of 10 Tb/ in 2 . The Information Storage Industry Consortium targets this density for 2015, enabling 7 TB and more in a single 2.5” disk at a cost of about $3/TB [14], [15]. While BPMR, HAMR, and MAMR offer daunting challenges at the level of device engineering, the bulk of the challenges and opportunity for shingled writing lie at the systems architecture level. The major, but significant, func- tional difference of shingled writing from current disk drives is that in-place overwrites of data in a track destroy the data in subsequent tracks. III. I NTEGRATION OF SWD S INTO S YSTEMS SWDs are an ideal replacement for tape for backup and archival data, thanks to the largely-sequential nature of writes to these devices. Moreover, the ability of SWDs to be read randomly makes them more attractive than tapes for archival and backup. However, to be most useful and to provide a mass market, SWDs must remain capable of serving as primary random-access storage devices. Kasiraj et al. propose organizing the disk into ...