摘要:
In a video server hard disk storage unit, video layout ensures average case bandwidth and case seek time during bounded-time information transfers, to maximize access of users S. The storage unit includes D hard disks whose platters are divided into Z zones. Each movie is divided into i blocks and is striped across all D disks such that movie block i is stored on disk (i mod D). In one embodiment, movie block i is placed in zone (i/D) mod Z of disk (i mod D). User streams are organized into slots containing N=S/D streams and the number of streams in each slot accessing a zone is ≤ (N/Z). Seeking is to the nearest end of a set of accesses, where total seek time for a set of N accesses is ≤ N*seek (1.5/N). Preferably two storage buffers/stream are used. Admission policy permits stream assignment to any slot, but streams do not transfer until the needed data block is on the disk needed by the other streams in the slot, and until the block is in the proper disk zone. A second embodiment optimizes disk seeks by placing movie block i in zone (((i/D) + (i mod D)) mod Z) of disk (i mod D) to reduce buffer unit requirements. The S streams progress in order 0 to S-1 through disks 0 to D-1, all streams accessing each disk from the same zone before moving through the disk in another zone.
摘要翻译:在视频服务器的硬盘存储单元,视频布局确保平均情况下的带宽和壳体中有界时间信息传输寻道时间,以最大化用户的访问S的存储单元包括ð硬盘谁的盘片被分成多个区域Z. 每个电影被分成i个块和在所有磁盘维搜索做电影块i被存储在磁盘上(I MOD D)被条带化。 在一个,实施例的电影块i在区域(i / D)盘的模Z(I MOD D)放置。 用户数据流被组织成包含N = S / D流和在每个时隙中访问区流的数量的时隙是≤(N / Z)。 求是对一组的访问,其中,总寻道时间为一组N个存取是≤ N的最近端*寻求(1.5 / N)。 优选地,两个存储缓冲区/流被使用。 招生政策允许流分配给任何插槽,但流不转移到所需的数据块是通过插槽中的其他流所需的磁盘上,并且直到块是正确的磁盘区。 第二实施例优化磁盘通过将电影块i中区(((I / D)+(I MOD D))模Z)盘(I MOD D),以减少缓冲单元要求寻道。 在S,以0到磁盘0流进展S-1至D-1,通过盘在另一区移动之前访问来自同一区中的每个磁盘的所有流。
摘要:
In a video server hard disk storage unit, video layout ensures average case bandwidth and case seek time during bounded-time information transfers, to maximize access of users S. The storage unit includes D hard disks whose platters are divided into Z zones. Each movie is divided into i blocks and is striped across all D disks such that movie block i is stored on disk (i mod D). In one embodiment, movie block i is placed in zone (i/D) mod Z of disk (i mod D). User streams are organized into slots containing N=S/D streams and the number of streams in each slot accessing a zone is ≤ (N/Z). Seeking is to the nearest end of a set of accesses, where total seek time for a set of N accesses is ≤ N*seek (1.5/N). Preferably two storage buffers/stream are used. Admission policy permits stream assignment to any slot, but streams do not transfer until the needed data block is on the disk needed by the other streams in the slot, and until the block is in the proper disk zone. A second embodiment optimizes disk seeks by placing movie block i in zone (((i/D) + (i mod D)) mod Z) of disk (i mod D) to reduce buffer unit requirements. The S streams progress in order 0 to S-1 through disks 0 to D-1, all streams accessing each disk from the same zone before moving through the disk in another zone.
摘要:
In a video server hard disk storage unit, video layout ensures average case bandwidth and case seek time during bounded-time information transfers, to maximize access of users S. The storage unit includes D hard disks whose platters are divided into Z zones. Each movie is divided into i blocks and is striped across all D disks such that movie block i is stored on disk (i mod D). In one embodiment, movie block i is placed in zone (i/D) mod Z of disk (i mod D). User streams are organized into slots containing N=S/D streams and the number of streams in each slot accessing a zone is ≤ (N/Z). Seeking is to the nearest end of a set of accesses, where total seek time for a set of N accesses is ≤ N*seek (1.5/N). Preferably two storage buffers/stream are used. Admission policy permits stream assignment to any slot, but streams do not transfer until the needed data block is on the disk needed by the other streams in the slot, and until the block is in the proper disk zone. A second embodiment optimizes disk seeks by placing movie block i in zone (((i/D) + (i mod D)) mod Z) of disk (i mod D) to reduce buffer unit requirements. The S streams progress in order 0 to S-1 through disks 0 to D-1, all streams accessing each disk from the same zone before moving through the disk in another zone.