Robert Treat wrote:
> On Tue, 2003-11-18 at 17:31, Sailesh Krishnamurthy wrote:
>>
>>One step at a time :-)
>>
>>Actually a big problem is figuring out new pieces for the
>>projects. Most of the items in the TODO list are way too much for a
>>class project - we gave 'em 3 weeks to make the Hash GroupedAgg work
>>for large numbers of unique values (by using a form of hybrid hashing).
>>Another thing I toyed with was having an implementation of a
>>Tid-List-Fetch .. sorting a TID-list from an index and fetching the
>>records of the relation off the sorted list for better IO
>>performance. AFAICT something like this isn't present yet .. can pgsql
>>do this already ?
> While some form of bitmapped indexing would be cool, other ideas might
> be to implement different buffer manager strategies. I was impressed by
> how quickly Jan was able to implement ARC over LRU, but there are a host
> of other strategies that could also be implemented.
Remember that interview with Jim Gray:
http://www.acmqueue.org/modules.php?name=Content&pa=showpage&pid=43
"Certainly we have to convert from random disk access to sequential
access patterns. Disks will give you 200 accesses per second, so if
you read a few kilobytes in each access, you're in the
megabyte-per-second realm, and it will take a year to read a
20-terabyte disk.
If you go to sequential access of larger chunks of the disk, you will
get 500 times more bandwidth—you can read or write the disk in a day.
So programmers have to start thinking of the disk as a sequential
device rather than a random access device."
Isn't a TID-List-Fetch implementation a crucial first step in the
right direction?
Mike Mascari
mascarm@mascari.com
