Forums » SAP Quality Assurance Best Practices


  • June 28, 2017 12:40 PM IST

    How historically RDBMS paintings

    As inside the advent of conventional RDBMS reminiscence become very very high-priced and compared to that disk area changed into loads cheaper so the disk based totally RDBMS was born. As I appearance back from today - this became usually meant as to be an intermediate answer.

    Know-how the Buffer Cache and it's nature

    Despite the fact that information of traditional RDBMS is disk based totally - the handiest area to perform on that statistics is inside the CPU-registers, so the buffer cache is wanted to carry a subset of the statistics closer to the CPU with out the want of I/O on each block access;

    The buffer cache itself is nothing else then a small, digital and logical reminiscence window of the entire disk based totally information. records blocks so as to study into the buffer cache and are changing other cached statistics blocks (already flushed ones), blocks changed within the buffer cache will written all the way down to the disk on checkpoint
    and contiguously changed devoted statistics can be logged as a byte stream via the log creator.

    Because the buffer cache is a virtual window on the report block orientated statistics there's no capability of direct reminiscence access to the records or more specific to a selected row of a desk once loaded into the buffer cache. You need to arrange quite a few lists, semaphores and reminiscence deal with translation stuff to get a selected row from the buffer cache, because the unique identifier of a row the rowid. The rowid isn't a memory based totally construct however a report based one - it carries no direct data wherein a particular row is placed inside the buffer cache - the rows beginning cope with in the memory. a number of CPU-cyles are needed to translate this virtual file cache nature into the a reminiscence addressable one.

    Lower back to the intro, if you will resize the buffer cache to hold the complete statistics inside the cache
    you still have these kind of digital report based totally mechanisms; No direct memory-get admission to to a row - you deal nevertheless with a disk based totally behaving RDBMS; This isn't always in-memory databasing!
    Do now not mix up old stuff with new one.

    Actual In-memory databasing - SAP HANA

    Now as CPU and reminiscence has expanded with it's capacity/abilities with stellar increase even a larger amount of facts may be hold immediately completely in-reminiscence.

    Therefore on startup of a SAP HANA database all facts is loaded into memory - then there may be no need to test anymore if a statistics is already in reminiscence or a study from disk is essential. The statistics because of column stores (vertical colum smart storage, imply values of one attribute are saved sequential in reminiscence) is CPU-aligned;
    no digital steeply-priced calculation of LRU, logical block addresses ... however direct (pointer) addressing of statistics.

    Moreover with SAP HANA the data is dictionary compressed means the desk itself is modelled as a micro starschema, tables information includes handiest integers (CPU -friendly and compact) or bitmaps as information
    referencing the dictionary maintained values of the column and even extra the usage of native advanced capabilities of the CPU for instance SIMD (unmarried instruction, more than one data) is supported.

    The primary database storage now could be the RAM as opposed to the disks;

    With this in thoughts an SAP HANA is in a position too be multitudes quicker compared to conventional RDBMS
    even the statistics on the vintage style RDBMS would in shape absolutely within the buffer cache.