[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAKLKtzfRysmG8Wj5_Lmjy2xH+-DJZ=Fwr7JydNJRc5A4+SnL0g@mail.gmail.com>
Date: Mon, 28 May 2012 20:00:07 +0530
From: Saugata Das <saugata.das@...aro.org>
To: Jeff Moyer <jmoyer@...hat.com>
Cc: Saugata Das <saugata.das@...ricsson.com>,
linux-ext4@...r.kernel.org, linux-fsdevel@...r.kernel.org,
patches@...aro.org, arnd.bergmann@...aro.org, venkat@...aro.org,
lporzio@...ron.com, tytso@....edu, adilger@...ger.ca,
hch@...radead.org, bharrosh@...asas.com, deepak.saxena@...aro.org
Subject: Re: [PATCH v2 1/2] block: add BH_Meta flag
On 22 May 2012 22:07, Jeff Moyer <jmoyer@...hat.com> wrote:
> Saugata Das <saugata.das@...ricsson.com> writes:
>
>> From: Saugata Das <saugata.das@...aro.org>
>>
>> Today, storage devices like eMMC has special features like data tagging
>> (introduced in MMC-4.5 version) in order to improve performance of some
>> specific writes. On MMC stack, data tagging is used for all writes which
>> has REQ_META flag set. This patch adds the capability to add REQ_META flag
>> during meta data write.
>
> Presumably you're doing this to get better performance for some
> workload, yes? Could you please let us know what workloads you tested
> and how this patch set helps? In other words, what benchmarking did you
> perform and what were the results? Do you expect this to make some
> other workloads perform worse?
>
Sorry for the late reply.
The use cases, we are trying to improve are the file writes to eMMC by
redirecting the meta-data to a dedicated location within eMMC which
provides high write performance. With the proposed patch, I see that
for a 128KB file write, ~11% data writes are meta-data write. If we
consider, eMMC provides atleast double the performance for such writes
then this fetches ~5% improvement in overall file write performance.
Lower the file size, higher will be the gain. Note that eMMC
specification does not say how the memory devices will implement this
data tag feature, so the improvement seen will vary between devices
from different vendors.
There is no negative impact on any other use cases.
> Cheers,
> Jeff
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists