From: Stefan Hajnoczi <stefanha@gmail.com>
To: "Benoît Canet" <benoit@irqsave.net>
Cc: kwolf@redhat.com, qemu-devel@nongnu.org, stefanha@redhat.com
Subject: Re: [Qemu-devel] [PATCH V2 for-1.6 2/5] block: Modify the throttling code to implement the leaky bucket algorithm.
Date: Tue, 23 Jul 2013 16:59:07 +0200 [thread overview]
Message-ID: <20130723145906.GB14190@stefanha-thinkpad.redhat.com> (raw)
In-Reply-To: <1374582067-9063-3-git-send-email-benoit@irqsave.net>
On Tue, Jul 23, 2013 at 02:21:04PM +0200, Benoît Canet wrote:
> @@ -152,6 +254,10 @@ void bdrv_io_limits_enable(BlockDriverState *bs)
> qemu_co_queue_init(&bs->throttled_reqs);
> bs->block_timer = qemu_new_timer_ns(vm_clock, bdrv_block_timer, bs);
> bs->io_limits_enabled = true;
> + bs->previous_leak = qemu_get_clock_ns(rt_clock);
> + qemu_mod_timer(bs->block_timer,
> + qemu_get_clock_ns(vm_clock) +
> + BLOCK_IO_THROTTLE_PERIOD);
> }
I don't fully understand the patch yet but:
1. Can we activate the timer only when requests are actually pending?
Imagine a host with 1000 guests, even a 1 second timer becomes
wasteful.
2. You don't vary the wait time, does this mean a throttled request must
wait for max 1 second? If yes, then it introduces a big variance on
request latency.
> +/* This function check if the correct bandwith threshold has been exceeded
> + *
> + * @is_write: true if the current IO is a write, false if it's a read
> + * @ret: true if threshold has been exceeded else false
> + */
> +static bool bdrv_is_bps_threshold_exceeded(BlockDriverState *bs, bool is_write)
> +{
> + /* limit is on total read + write bps : do the sum and compare with total
> + * threshold
> + */
> + if (bs->io_limits.bps[BLOCK_IO_LIMIT_TOTAL]) {
> + int64_t bytes = bs->leaky_buckets.bytes[0] +
> + bs->leaky_buckets.bytes[1];
BLOCK_IO_LIMIT_READ, BLOCK_IO_LIMIT_WRITE instead of 0/1?
> + return bs->io_limits.bps_threshold[BLOCK_IO_LIMIT_TOTAL] < bytes;
> + }
> +
> + /* check wether the threshold corresponding to the current io type (read,
> + * write has been exceeded
s/write/write)/
> + */
> + if (bs->io_limits.bps[is_write]) {
> + return bs->io_limits.bps_threshold[is_write] <
> + bs->leaky_buckets.bytes[is_write];
> + }
> +
> + /* no limit */
> + return false;
> +}
> +
> +/* This function check if the correct iops threshold has been exceeded
> + *
> + * @is_write: true if the current IO is a write, false if it's a read
> + * @ret: true if threshold has been exceeded else false
> + */
> +static bool bdrv_is_iops_threshold_exceeded(BlockDriverState *bs, bool is_write)
> +{
Same comments apply as to bdrv_is_bps_threshold_exceeded().
> @@ -280,10 +280,25 @@ static int parse_block_error_action(const char *buf, bool is_read)
> }
> }
>
> +static bool invalid(int64_t limit)
Please choose a more specific function name like check_io_limit().
next prev parent reply other threads:[~2013-07-23 14:59 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-07-23 12:21 [Qemu-devel] [PATCH V2 for-1.6 0/5] Leaky bucket throttling and features Benoît Canet
2013-07-23 12:21 ` [Qemu-devel] [PATCH V2 for-1.6 1/5] block: Repair the throttling code Benoît Canet
2013-07-23 14:31 ` Stefan Hajnoczi
2013-07-23 12:21 ` [Qemu-devel] [PATCH V2 for-1.6 2/5] block: Modify the throttling code to implement the leaky bucket algorithm Benoît Canet
2013-07-23 14:59 ` Stefan Hajnoczi [this message]
2013-07-23 15:14 ` Benoît Canet
2013-07-23 15:20 ` Benoît Canet
2013-07-23 12:21 ` [Qemu-devel] [PATCH V2 for-1.6 3/5] block: Add support for throttling burst threshold in QMP and the command line Benoît Canet
2013-07-23 12:21 ` [Qemu-devel] [PATCH V2 for-1.6 4/5] block: Add iops_sector_count to do the iops accounting for a given io size Benoît Canet
2013-07-23 12:21 ` [Qemu-devel] [PATCH V2 for-1.6 5/5] block: Add throttling percentage metrics Benoît Canet
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20130723145906.GB14190@stefanha-thinkpad.redhat.com \
--to=stefanha@gmail.com \
--cc=benoit@irqsave.net \
--cc=kwolf@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=stefanha@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).