瓦记录磁盘数据组织方式设计与实现

发布时间：2018-06-12 00:21

本文选题：磁盘 + 瓦记录　；参考：《华中科技大学》2012年硕士论文

【摘要】：根据摩尔定律，磁存储容量按照年60%的复合增长率递增。然而，受到超顺磁效应的影响，存储面密度无法超越1Tb/in2的限制。为了进一步提升磁存储面密度，热辅助记录（heat-assisted recording, HAMR）和晶格记录（bit patterned media, BPM）成为研究热点。但是由于HAMR技术需要磁头技术的突破，，BPM技术需要磁介质材料技术的突破，两个技术都无法实用化。瓦存储技术采用磁道堆叠的方式存储数据，不需要对磁头和介质本身进行修正，就可以将存储密度提高2-3倍，成为最有可能实用化的技术之一。然而瓦记录由于采用覆盖写的技术而导致随机写性能严重衰减。如何解决此问题成为当今研究的热点。采用日志方式修改地址映射逻辑，可有效的解决随机写性能衰减的问题。将瓦记录磁盘划分为很多存储单元，当一个存储单元中的数据需要修改时，将修改追加到单元内记录的末尾；当存储单元的空间写满后，直接重新分配一个新的存储单元，并将旧单元中的有效数据移动至新单元。存储设备上的软件维护了一个从逻辑地址到物理地址的映射表，并且执行将垃圾存储单元回收整理的工作。提出了一种可行的瓦记录数据组织策略，逻辑地址被映射为段、区、道、块。通过虚拟块设备的方式，建立瓦记录模拟器。并将传统虚拟磁盘和瓦记录虚拟磁盘进行性能横向对比。实验结果表明，采用提出的新型数据组织方式，瓦记录磁盘在随机读写小数据时性能与普通磁盘相当甚至略有提升，在顺序读写大数据时性能仅有10%左右的下降。
[Abstract]:According to Moore's law, the magnetic storage capacity increases with the compound growth rate of 60%. However, under the effect of superparamagnetic effect, the density of storage surface can not exceed the limit of 1Tb/in2. In order to further improve the density of magnetic storage surface (heat-assisted recording, HAMR) and lattice record (bit patterned media, BPM) become the hot spot of research. But because HAMR technology needs the breakthrough of magnetic head technology, BPM technology needs the breakthrough of magnetic medium material technology, the two technologies can not be applied. The tile storage technology stores data using the way of stacking of magnetic track. It does not need to correct the magnetic head and medium itself, so the storage density can be increased by 2-3 times to become the most likely practical technique. However, Watt record has severely weakened random write performance due to overwrite technology. How to solve this problem has become a hot research topic nowadays.
Modifying the address mapping logic by log method can effectively solve the problem of random write performance degradation. The tile record disk is divided into many storage units. When the data in a storage unit needs to be modified, the modification is added to the end of the record in the unit. When the space of the storage unit is full, a new memory is reassigned. The storage unit moves the valid data in the old unit to the new unit. The software on the storage device maintains a mapping table from the logical address to the physical address, and performs the work to reclaim the garbage storage unit. A feasible tile record data organization strategy is proposed, and the logical address is mapped into segments, zones, paths and blocks. In the way of virtual block device, a tile record simulator is set up, and the performance of traditional virtual disk and tile record virtual disk is contrasted.
The experimental results show that the performance of tile recording disk in random reading and writing of small data is even slightly improved by using the proposed new data organization method. The performance of the disk is only about 10% when reading and writing large data.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP333.35

【参考文献】