Memory and Disk Accesses - Indexing and Searching Techniques - Lecture Slides, Slides of Research Methodology

Download Memory and Disk Accesses - Indexing and Searching Techniques - Lecture Slides and more Slides Research Methodology in PDF only on Docsity! Objectives_template file:///C|/Documents%20and%20Settings/iitkrana1/My%20Documents/Google%20Talk%20Received%20Files/ist_data/lecture4/4_1.htm[6/14/2012 3:17:51 PM] Module 1: Basics and Background Lecture 4: Memory and Disk Accesses The Lecture Contains: Memory organisation Example of memory hierarchy Memory hierarchy Disks Disk access Disk capacity Disk access time Typical disk parameters Access times Objectives_template file:///C|/Documents%20and%20Settings/iitkrana1/My%20Documents/Google%20Talk%20Received%20Files/ist_data/lecture4/4_2.htm[6/14/2012 3:17:52 PM] Module 1: Basics and Background Lecture 4: Memory and Disk Accesses Memory organisation Faster storage is more costly Gap between CPU and main memory speed is increasing Good programs exploit locality of data: both spatially and temporally Objectives_template file:///C|/Documents%20and%20Settings/iitkrana1/My%20Documents/Google%20Talk%20Received%20Files/ist_data/lecture4/4_5.htm[6/14/2012 3:17:52 PM] Module 1: Basics and Background Lecture 4: Memory and Disk Accesses Memory hierarchy Smaller capacity but faster access memory acts as a subset of storage for larger capacity but slower access memory Generalization of the concept of "cache" Why does memory hierarchy work? Objectives_template file:///C|/Documents%20and%20Settings/iitkrana1/My%20Documents/Google%20Talk%20Received%20Files/ist_data/lecture4/4_6.htm[6/14/2012 3:17:53 PM] Module 1: Basics and Background Lecture 4: Memory and Disk Accesses Memory hierarchy Smaller capacity but faster access memory acts as a subset of storage for larger capacity but slower access memory Generalization of the concept of "cache" Why does memory hierarchy work? Locality of access: Programs tend to access data higher up in the hierarchy much more often Intelligent caching algorithms Net effect: Large capacity of memory with very fast access Objectives_template file:///C|/Documents%20and%20Settings/iitkrana1/My%20Documents/Google%20Talk%20Received%20Files/ist_data/lecture4/4_7.htm[6/14/2012 3:17:53 PM] Module 1: Basics and Background Lecture 4: Memory and Disk Accesses Disks Physically, disks consist of circular platters Both surfaces of a platter can be accessed Each surface contains concentric tracks Tracks are divided into sectors separated by gaps Aligned tracks form a cylinder Objectives_template file:///C|/Documents%20and%20Settings/iitkrana1/My%20Documents/Google%20Talk%20Received%20Files/ist_data/lecture4/4_10.htm[6/14/2012 3:17:53 PM] Module 1: Basics and Background Lecture 4: Memory and Disk Accesses Disk access time Smallest unit of information that can be read from or written to disk is a sector Block or page is a logical unit read from or written to by O/S Block consists of a contiguous sequence of sectors Objectives_template file:///C|/Documents%20and%20Settings/iitkrana1/My%20Documents/Google%20Talk%20Received%20Files/ist_data/lecture4/4_11.htm[6/14/2012 3:17:54 PM] Module 1: Basics and Background Lecture 4: Memory and Disk Accesses Disk access time Smallest unit of information that can be read from or written to disk is a sector Block or page is a logical unit read from or written to by O/S Block consists of a contiguous sequence of sectors Access time : Time to access a particular sector Seek time : Time to position arm heads over cylinder containing the target sector Typical seek time: 8 ms Rotational latency : (Average) time to rotate r/w head to the first bit of the sector (1 / 2) (1 / rpm) (60 s / 1 min) Transfer time : Time to read bits from the sector (1 / (#sectors / track)) (60 / rpm) Objectives_template file:///C|/Documents%20and%20Settings/iitkrana1/My%20Documents/Google%20Talk%20Received%20Files/ist_data/lecture4/4_12.htm[6/14/2012 3:17:54 PM] Module 1: Basics and Background Lecture 4: Memory and Disk Accesses Typical disk parameters Average seek times from 4-10 ms Rotational speeds are 60, 90, 120, 250 revolutions per second, i.e. ,3600, 5400, 7200, 15000 rpm respectively Sector sizes vary between 512 bytes and 1024 bytes 400 to 1000 sectors per track 20,000 to 50,000 tracks per surface 1 to 5 platters per disk Objectives_template file:///C|/Documents%20and%20Settings/iitkrana1/My%20Documents/Google%20Talk%20Received%20Files/ist_data/lecture4/4_15.htm[6/14/2012 3:17:54 PM] Module 1: Basics and Background Lecture 4: Memory and Disk Accesses Access times This disk access time is for random I/O Once the first bit is read, the rest (sequential I/O) is almost free (only 0.02 ms) Ratio of random I/O to sequential I/O is, therefore, (12 / 0.02) 600 times Bulk transfer rates are calculated more precisely using gaps Objectives_template file:///C|/Documents%20and%20Settings/iitkrana1/My%20Documents/Google%20Talk%20Received%20Files/ist_data/lecture4/4_16.htm[6/14/2012 3:17:54 PM] Module 1: Basics and Background Lecture 4: Memory and Disk Accesses Access times This disk access time is for random I/O Once the first bit is read, the rest (sequential I/O) is almost free (only 0.02 ms) Ratio of random I/O to sequential I/O is, therefore, (12 / 0.02) 600 times Bulk transfer rates are calculated more precisely using gaps Time to access a byte 60 ns for DRAM (used for main memory) and 4 ns for SRAM (used for cache memory) Ratio of disk access time to memory access time 40000 for SRAM and 2500 for DRAM Objectives_template file:///C|/Documents%20and%20Settings/iitkrana1/My%20Documents/Google%20Talk%20Received%20Files/ist_data/lecture4/4_17.htm[6/14/2012 3:17:54 PM] Module 1: Basics and Background Lecture 4: Memory and Disk Accesses Access times This disk access time is for random I/O Once the first bit is read, the rest (sequential I/O) is almost free (only 0.02 ms) Ratio of random I/O to sequential I/O is, therefore, (12 / 0.02) 600 times Bulk transfer rates are calculated more precisely using gaps Time to access a byte 60 ns for DRAM (used for main memory) and 4 ns for SRAM (used for cache memory) Ratio of disk access time to memory access time 40000 for SRAM and 2500 for DRAM Disk access time is dominated by seek time and rotational latency Sequential access algorithms exploit the (almost) free access time of later bits heavily Most algorithms aim to avoid random I/Os