Re: IO_CELL_OFFLOAD_RETURNED_BYTES stats

From: Ls Cheng <exriscer_at_gmail.com>
Date: Sat, 26 Apr 2014 00:46:18 +0200
Message-ID: <CAJ2-Qb9zBMeb_P5RYDFLt1EFJ3AxS+m-QN9eMx5f4JG8CaYR0g_at_mail.gmail.com>



Hi

I thought IO_CELL_OFFLOAD_RETURNED_BYTES shows the bytes returned by smart scan processing from the cells, was reading a old documentation, doc bug. Just looked the newest doc and it is as you have stated, regular path I/O.

Thanks

On Fri, Apr 25, 2014 at 6:11 PM, David Fitzjarrell <oratune_at_yahoo.com>wrote:

> The io_cell_offload_eligible_bytes column reports the bytes of data that
> qualify for offload. This is the volume of data that can be offloaded to
> the storage cells during query execution. The
> io_cell_offload_returned_bytes column reports the number of bytes returned
> by the regular I/O path. These are the bytes that were not offloaded to the
> cells.
>
>
> David Fitzjarrell
> Primary author, "Oracle Exadata Survival Guide"
> On Friday, April 25, 2014 9:39 AM, Ls Cheng <exriscer_at_gmail.com> wrote:
> Hi
>
> I am check a query statistics in Exadata and I noticed that
> v$sql.IO_CELL_OFFLOAD_ELIGIBLE_BYTES is cero but
> IO_CELL_OFFLOAD_RETURNED_BYTES has a pretty high value.
>
> How can this possible?
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> *selectexecutions,disk_reads*16384/(1024*1024)
> PHY_MB,IO_CELL_OFFLOAD_ELIGIBLE_BYTES/1024/1024 as "OFFLOADED_MB",
> IO_CELL_OFFLOAD_RETURNED_BYTES/1024/1024 as
> "OFFLOAD_RT_MB",IO_INTERCONNECT_BYTES/1024/1024 as
> "ICONNECT_MB",PHYSICAL_READ_REQUESTS ,PHYSICAL_READ_BYTES/1024/1024
> PHY_R_MB , PHYSICAL_WRITE_REQUESTS
> ,PHYSICAL_WRITE_BYTES/1024/1024 PHY_W_MBfrom v$sqlwhere sql_id =
> '4j92kqhrpcpr6'EXECUTIONS PHY_MB OFFLOADED_MB OFFLOAD_RT_MB ICONNECT_MB
> PHYSICAL_READ_REQUESTS PHY_R_MB PHYSICAL_WRITE_REQUESTS PHY_W_MB
> ---------- ---------- ------------ ------------- -----------
> ---------------------- ---------- -----------------------
> ---------- 1 130755.328 0 259544.156
> 520041.828 2801860 130598.375 641580
> 129899.297 *
>
> *Thanks*
>
>
>
>
>
>
>

--
http://www.freelists.org/webpage/oracle-l
Received on Sat Apr 26 2014 - 00:46:18 CEST

Original text of this message