Thank you for your response, Tejun. We have figured some things out since I first e-mailed. We have been using IOMeter running on a remote Windows system which connects to a Dynamo module running on the Linux target. The Dynamo module uses a file open command with an O_DIRECT flag to minimize the effects of caching. Unfortunately, this causes the IO to be totally synchronous which is kinda bad for this type of benchmarking. I have another question for you. Where can I find information on how to get the AHCI driver to recognize our new AHCI controller? Thank you! Bob -----Original Message----- >From: Tejun Heo <tj@xxxxxxxxxx> >Sent: Jul 31, 2008 11:06 PM >To: the4hoffmans@xxxxxxxxxxxxx >Cc: jgarzik@xxxxxxxxx, linux-ide@xxxxxxxxxxxxxxx, blward@xxxxxxxxxx >Subject: Re: Issue with AHCI driver > >the4hoffmans@xxxxxxxxxxxxx wrote: >> Sorry about the multiple e-mails Jeff, but our IT group has seen fit >> to deny me the ability to send e-mail to vger.kernel.org so I'm >> trying again from my own personal account. >> ------------------------------------------------------------------- >> >> >> My name is Bob Hoffman. I am a software engineer for Micron >> Technology in Minneapolis. I have been working on AHCI drivers. We >> recently installed kernel build 2.6.19.0 (unpatched) of Linux on an >> Intel motherboard system which includes the ICH9 SATA controller. We >> have been running some benchmarks and found that the SATA harddrives >> looked slow. When I attached a SATA analyzer it shows that we are >> only getting one command "in flight" or "outstanding" at a time, and >> it is always slot (or tag) zero. The AHCI driver doesn't seem to be >> taking advantage of the 32 deep queue. >> >> I checked the queue depth using the info at >> http://linux-ata.org/faq.html#io32 and it shows "NCQ (depth 31/32)". >> I'm wondering if there is something else I need to know. >> >> I've copied our "Linux guru", Bob Ward, who knows more about the >> kernel stuff than I do. >> >> Thank you for any help you can provide. Bob (Hoffman) ; ) -- To >> unsubscribe from this list: send the line "unsubscribe linux-ide" in >> the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo >> info at http://vger.kernel.org/majordomo-info.html >> > >What kind of workload are you testing? If you're doing multiple random >IOs, using deadline io scheduler will give you a better result. Can't >say whether that would help on any realistic workload tho. > >-- >tejun -- To unsubscribe from this list: send the line "unsubscribe linux-ide" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
