CS4624 Disks

• RAID: Redundant Array of Inexpensive Disks
  
  
  • reliability vs. expense
    
  • higher throughput by bit parallel operation (e.g., 3 MB from disk to 15 MB from array)
  
  
• Disk retrieval unit
  
  
  • decrease seeks and transfers by transferring more at a time (granularity)
    
  • size tradeoff vs. fragmentation (end of file, after deletion)
  
  
• Disk use
  
• Disk layout / allocation on disk (scattering)
  
  
  • contiguous allocation
    
  • constrained allocation (within some local or global bound on separation)
    
  • random/scattered
  
  
• Disk scheduling
  
  
  • round robin: need large buffers else get large delays as no. involved increases
    
  • elevator: pick up everything at this level
    
  • GSS: group sweeping scheme, so handle all in a group in one scan
    
  • servicing policies: local vs. global