Re: High end server and storage for a PostgreSQL OLTP system - Mailing list pgsql-performance

From William Yu
Subject Re: High end server and storage for a PostgreSQL OLTP system
Date
Msg-id 420120BF.5050208@talisys.com
Whole thread Raw
In response to Re: High end server and storage for a PostgreSQL OLTP system  (Bruce Momjian <pgman@candle.pha.pa.us>)
List pgsql-performance
Bruce Momjian wrote:
> William Yu wrote:
>
>>>>You can get 64-bit Xeons also but it takes hit in the I/O department due
>>>>to the lack of a hardware I/O MMU which limits DMA transfers to
>>>>addresses below 4GB. This has a two-fold impact:
>>>>
>>>>1) transfering data to >4GB require first a transfer to <4GB and then a
>>>>copy to the final destination.
>>>>
>>>>2) You must allocate real memory 2X the address space of the devices to
>>>>act as bounce buffers. This is especially problematic for workstations
>>>>because if you put a 512MB Nvidia card in your computer for graphics
>>>>work -- you've just lost 1GB of memory. (I dunno how much the typical
>>>>SCSI/NIC/etc take up.)
>
>
> When you say "allocate real memory 2X" are you saying that if you have
> 16GB of RAM only 8GB is actually usable and the other 8GB is for
> bounce buffers, or is it just address space being used up?
>

It's 2x the memory space of the devices. E.g. a Nvidia Graphics card w/
512MB of RAM would require 1GB of memory to act as bounce buffers. And
it has to be real chunks of memory in 64-bit mode since DMA transfer
must drop it into real memory in order to then be copied to > 4GB.

pgsql-performance by date:

Previous
From: "Merlin Moncure"
Date:
Subject: Re: High end server and storage for a PostgreSQL OLTP system
Next
From: Josh Berkus
Date:
Subject: Re: horizontal partition