sata_promise performance observations

sata_promise performance observations

[Brad Campbell]

G'day all,

Interesting observation with 3 SATA150TX4 cards and 10 Maxtor Maxline-II 250GB SATA drives.

If I run a concurrent dd if=/dev/sda > /dev/null and dd if=/dev/sdb > /dev/null I get the expected 
115MB/s transfer rates in vmstat.

If I do a dd if=/dev/sda > /dev/null and dd if=/dev/sde > /dev/null I get about 80MB/s

sda, sde & sdi I get
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----

  3  0  22540   2560 463908   5320    0    0 85294    16 1102   761 11 88  0  1
  3  1  22540   1480 464852   5320    0    0 86493    25  987   512 13 86  0  0
  2  2  22540   2176 464304   5268    0    0 86774    18 1091   721 12 87  0  1
  2  2  22540   1488 464860   5368    0    0 85562    18  921   416 13 87  0  0


When streaming from drives not on the same controller, I lose about 30mb/s transfer rate. In fact, 
it seems to max out at about 86MB/s no matter what combination of drives I use if I stream from more 
than one controller.

However, if I use sda, sbb & sdc I get

procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----

  0  3  22512   1768 464448   5304    0    0 121661    18 1257   566  9 59  0 32
  0  3  22512   1536 464696   5316    0    0 122903    14 1262   578 10 59  0 31
  0  3  22512   2112 464228   5268    0    0 121655    14 1255   578  9 60  0 32

How bizzare. I have latency set to 32 on all controllers, but I have tried as low as 8 and as high 
as 128 to no avail.

I have readahead set to 4096 on all devices.

Anyone have any ideas?

Not that this is an issue for me (I'm only using it as a fileserver over GB eth so its not like 
100MB/s is actually required, just an interesting observation)

Regards,
Brad

Re: sata_promise performance observations

[Eric Mudama]

arbitration between the multiple PCI cards?  A single card that
doesn't have to compete for PCI resources would, in theory, be
faster...

(Don't know much about this chunk of hardware, 'tis only a guess)


On Sat, 23 Oct 2004 12:36:54 +0400, Brad Campbell <brad@wasp.net.au> wrote:
> G'day all,
> 
> Interesting observation with 3 SATA150TX4 cards and 10 Maxtor Maxline-II 250GB SATA drives.
> 
> If I run a concurrent dd if=/dev/sda > /dev/null and dd if=/dev/sdb > /dev/null I get the expected
> 115MB/s transfer rates in vmstat.
> 
> If I do a dd if=/dev/sda > /dev/null and dd if=/dev/sde > /dev/null I get about 80MB/s
> 
> sda, sde & sdi I get
> procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
> 
>   3  0  22540   2560 463908   5320    0    0 85294    16 1102   761 11 88  0  1
>   3  1  22540   1480 464852   5320    0    0 86493    25  987   512 13 86  0  0
>   2  2  22540   2176 464304   5268    0    0 86774    18 1091   721 12 87  0  1
>   2  2  22540   1488 464860   5368    0    0 85562    18  921   416 13 87  0  0
> 
> When streaming from drives not on the same controller, I lose about 30mb/s transfer rate. In fact,
> it seems to max out at about 86MB/s no matter what combination of drives I use if I stream from more
> than one controller.
> 
> However, if I use sda, sbb & sdc I get
> 
> procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
> 
>   0  3  22512   1768 464448   5304    0    0 121661    18 1257   566  9 59  0 32
>   0  3  22512   1536 464696   5316    0    0 122903    14 1262   578 10 59  0 31
>   0  3  22512   2112 464228   5268    0    0 121655    14 1255   578  9 60  0 32
> 
> How bizzare. I have latency set to 32 on all controllers, but I have tried as low as 8 and as high
> as 128 to no avail.
> 
> I have readahead set to 4096 on all devices.
> 
> Anyone have any ideas?
> 
> Not that this is an issue for me (I'm only using it as a fileserver over GB eth so its not like
> 100MB/s is actually required, just an interesting observation)
> 
> Regards,
> Brad
> -
> To unsubscribe from this list: send the line "unsubscribe linux-ide" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>