Re: Looking for the Oracle equivalent to MS SQL Server timestamp field

Maybe you right. I said that because thinking to a reason to use a reverse index on a generated key like a sequence. The key is reversed to break the tree, so the index became more selective. By example having 10000 number key who start by 1234 at the first branch level. Oracle have to scan all these entries before passing to next branch. But having a completely random number like a sysquid assure you a direct access.

I don't see your point about caching sequences. The insert was never a problem, can be done on schedule. The performance issue is always on data retrieval.

On Wed, 19 Feb 2003 03:31:01 GMT, "Jim Kennedy" <kennedy-downwithspammersfamily_at_attbi.com> wrote:

>I don't think an index on a sysguid will be faster. The key will be larger
>(a lot larger) and it doesn't make sense that it would be faster. In
>addition, inserting 1,000 rows each with a cached sequence is probably
>faster than 1,000 rows of sysguid.
>
>Jim
>"Donovan R." <mdonovan_at_hotmail.com> wrote in message
>news:eis55vg4du05gd4dps9pi3bvh12nk707hk_at_4ax.com...
>> try sysguid.
>>
>> Some people will recommend using sequences. A completly random number
>> like sysguid (Oracle claims that the generated number will be unique
>> on universe not only on your federation of servers) is more efficient
>> for a b-tree index(less branches), so even if the sysguid is a raw
>> data (is bigger than a number) the index will be fastest finally.
>>
>>
>> On 18 Feb 2003 11:36:24 -0800, aguptill_at_nxtrend.com (Arch) wrote:
>>
>> >I'm looking for the Oracle data type and possibly method to shadow the
>> >MS SQL Server timestamp data type.
>> >
>> >From SQL Server Online help a timestamp is defined as:
>> >timestamp is a data type that exposes automatically generated binary
>> >numbers, which are guaranteed to be unique within a database.
>> >timestamp is used typically as a mechanism for version-stamping table
>> >rows. The storage size is 8 bytes.
>> >
>> >I'm not necessarily concernd that the datatype stays binary but the
>> >funcationality is what I'm after. In particular is the notion of
>> >ensuring that I can tell if a record has been updated since the data
>> >was pulled that I'm now working with and possibly updating back to the
>> >db system.
>> >
>> >Thanks in advance.
>> >Arch
>>
>
Received on Wed Feb 19 2003 - 04:59:00 CET