public inbox for pbs-devel@lists.proxmox.com
 help / color / mirror / Atom feed
From: Thomas Lamprecht <t.lamprecht@proxmox.com>
To: Proxmox Backup Server development discussion
	<pbs-devel@lists.proxmox.com>,
	Dominik Csapak <d.csapak@proxmox.com>
Subject: Re: [pbs-devel] [PATCH proxmox-backup 2/2] tape/helpers/snapshot_reader: sort chunks by inode (per index)
Date: Fri, 18 Jun 2021 17:02:32 +0200	[thread overview]
Message-ID: <8b8aaf00-d4f7-471b-7bed-9f6ddbd9a634@proxmox.com> (raw)
In-Reply-To: <20210618092910.4518-2-d.csapak@proxmox.com>

On 18.06.21 11:29, Dominik Csapak wrote:
> sort the chunks we want to backup to tape by inode, to gain some
> speed on spinning disks. this is done per index, not globally.
> 
> costs a bit memory, but not too much, about 16 bytes per chunk which
> would mean ~4MiB for a 1TiB index with 4MiB chunks.
> 
> Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
> ---
> this resulted in a speedup in my setup of between 20 and 30%
> (single spinner with random snapshots, from 17-26MiB/s to 30-40MiB/s)
> we already do this for verification, but got no real feedback on it,
> so either it does not make that much of a difference in the real world,
> or it is not that visible on verification (since that varies very much
> anyway)

or, IMO more likely, people just don't give feedback if things are working out
OK-ish ;-)

> 
>  src/tape/helpers/snapshot_reader.rs | 20 +++++++++++++-------
>  1 file changed, 13 insertions(+), 7 deletions(-)
> 
> diff --git a/src/tape/helpers/snapshot_reader.rs b/src/tape/helpers/snapshot_reader.rs
> index 7b272e37..416c88c1 100644
> --- a/src/tape/helpers/snapshot_reader.rs
> +++ b/src/tape/helpers/snapshot_reader.rs
> @@ -107,7 +107,7 @@ impl SnapshotReader {
>  pub struct SnapshotChunkIterator<'a> {
>      snapshot_reader: &'a SnapshotReader,
>      todo_list: Vec<String>,
> -    current_index: Option<(Arc<Box<dyn IndexFile>>, usize)>,
> +    current_index: Option<(Arc<Box<dyn IndexFile + Send>>, usize, Vec<(usize, u64)>)>,
>  }
>  
>  impl <'a> Iterator for SnapshotChunkIterator<'a> {
> @@ -119,20 +119,26 @@ impl <'a> Iterator for SnapshotChunkIterator<'a> {
>                  if self.current_index.is_none() {
>                      if let Some(filename) = self.todo_list.pop() {
>                          let file = self.snapshot_reader.open_file(&filename)?;
> -                        let index: Box<dyn IndexFile> = match archive_type(&filename)? {
> +                        let index: Box<dyn IndexFile + Send> = match archive_type(&filename)? {
>                              ArchiveType::FixedIndex => Box::new(FixedIndexReader::new(file)?),
>                              ArchiveType::DynamicIndex => Box::new(DynamicIndexReader::new(file)?),
>                              _ => bail!("SnapshotChunkIterator: got unknown file type - internal error"),
>                          };
> -                        self.current_index = Some((Arc::new(index), 0));
> +
> +                        let datastore =
> +                            DataStore::lookup_datastore(self.snapshot_reader.datastore_name())?;
> +                        let order = datastore.get_chunks_in_order(&index, |_| false, |_| Ok(()))?;
> +
> +                        self.current_index = Some((Arc::new(index), 0, order));
>                      } else {
>                          return Ok(None);
>                      }
>                  }
> -                let (index, pos) = self.current_index.take().unwrap();
> -                if pos < index.index_count() {
> -                    let digest = *index.index_digest(pos).unwrap();
> -                    self.current_index = Some((index, pos + 1));
> +                let (index, pos, list) = self.current_index.take().unwrap();
> +                if pos < list.len() {
> +                    let (real_pos, _) = list[pos];
> +                    let digest = *index.index_digest(real_pos).unwrap();
> +                    self.current_index = Some((index, pos + 1, list));
>                      return Ok(Some(digest));
>                  } else {
>                      // pop next index
> 





  reply	other threads:[~2021-06-18 15:02 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-06-18  9:29 [pbs-devel] [PATCH proxmox-backup 1/2] backup/datastore: refactor chunk inode sorting to the datastore Dominik Csapak
2021-06-18  9:29 ` [pbs-devel] [PATCH proxmox-backup 2/2] tape/helpers/snapshot_reader: sort chunks by inode (per index) Dominik Csapak
2021-06-18 15:02   ` Thomas Lamprecht [this message]
2021-07-12  4:23 ` [pbs-devel] applied-series: [PATCH proxmox-backup 1/2] backup/datastore: refactor chunk inode sorting to the datastore Thomas Lamprecht

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=8b8aaf00-d4f7-471b-7bed-9f6ddbd9a634@proxmox.com \
    --to=t.lamprecht@proxmox.com \
    --cc=d.csapak@proxmox.com \
    --cc=pbs-devel@lists.proxmox.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
Service provided by Proxmox Server Solutions GmbH | Privacy | Legal