qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Dr. David Alan Gilbert" <dgilbert@redhat.com>
To: Wei Wang <wei.w.wang@intel.com>
Cc: kevin.tian@intel.com, quintela@redhat.com, qemu-devel@nongnu.org,
	peterx@redhat.com, gloryxiao@tencent.com, yi.y.sun@intel.com
Subject: Re: [PATCH v1 1/2] migration/xbzrle: replace transferred xbzrle bytes with encoded bytes
Date: Tue, 21 Apr 2020 20:21:06 +0100	[thread overview]
Message-ID: <20200421192106.GM3029@work-vm> (raw)
In-Reply-To: <1587352003-3312-2-git-send-email-wei.w.wang@intel.com>

* Wei Wang (wei.w.wang@intel.com) wrote:
> Like compressed_size which indicates how many bytes are compressed, we
> need encoded_size to understand how many bytes are encoded with xbzrle
> during migration.
> 
> Replace the old xbzrle_counter.bytes, instead of adding a new counter,
> because we don't find a usage of xbzrle_counter.bytes currently, which
> includes 3 more bytes of the migration transfer protocol header (in
> addition to the encoding header). The encoded_size will further be used
> to calculate the encoding rate.
> 
> Signed-off-by: Yi Sun <yi.y.sun@intel.com>
> Signed-off-by: Wei Wang <wei.w.wang@intel.com>

Can you explain why these 3 bytes matter?  Certainly the 2 bytes of the
encoded_len are an overhead that's a cost of using XBZRLE; so if you're
trying to figure out whether xbzrle is worth it, then you should include
those 2 bytes in the cost.
That other byte, that holds ENCODING_FLAG_XBZRLE also seems to be pure
oerhead of XBZRLE; so your cost of using XBZRLE really does include
those 3 bytes.

SO to me it makes sense to include the 3 bytes as it currently does.

Dave

> ---
>  migration/migration.c |  2 +-
>  migration/ram.c       | 18 +++++++++---------
>  monitor/hmp-cmds.c    |  4 ++--
>  qapi/migration.json   |  6 +++---
>  4 files changed, 15 insertions(+), 15 deletions(-)
> 
> diff --git a/migration/migration.c b/migration/migration.c
> index 187ac0410c..8e7ee0d76b 100644
> --- a/migration/migration.c
> +++ b/migration/migration.c
> @@ -926,7 +926,7 @@ static void populate_ram_info(MigrationInfo *info, MigrationState *s)
>          info->has_xbzrle_cache = true;
>          info->xbzrle_cache = g_malloc0(sizeof(*info->xbzrle_cache));
>          info->xbzrle_cache->cache_size = migrate_xbzrle_cache_size();
> -        info->xbzrle_cache->bytes = xbzrle_counters.bytes;
> +        info->xbzrle_cache->encoded_size = xbzrle_counters.encoded_size;
>          info->xbzrle_cache->pages = xbzrle_counters.pages;
>          info->xbzrle_cache->cache_miss = xbzrle_counters.cache_miss;
>          info->xbzrle_cache->cache_miss_rate = xbzrle_counters.cache_miss_rate;
> diff --git a/migration/ram.c b/migration/ram.c
> index 04f13feb2e..bca5878f25 100644
> --- a/migration/ram.c
> +++ b/migration/ram.c
> @@ -677,7 +677,7 @@ static int save_xbzrle_page(RAMState *rs, uint8_t **current_data,
>                              ram_addr_t current_addr, RAMBlock *block,
>                              ram_addr_t offset, bool last_stage)
>  {
> -    int encoded_len = 0, bytes_xbzrle;
> +    int encoded_size = 0, bytes_xbzrle;
>      uint8_t *prev_cached_page;
>  
>      if (!cache_is_cached(XBZRLE.cache, current_addr,
> @@ -702,7 +702,7 @@ static int save_xbzrle_page(RAMState *rs, uint8_t **current_data,
>      memcpy(XBZRLE.current_buf, *current_data, TARGET_PAGE_SIZE);
>  
>      /* XBZRLE encoding (if there is no overflow) */
> -    encoded_len = xbzrle_encode_buffer(prev_cached_page, XBZRLE.current_buf,
> +    encoded_size = xbzrle_encode_buffer(prev_cached_page, XBZRLE.current_buf,
>                                         TARGET_PAGE_SIZE, XBZRLE.encoded_buf,
>                                         TARGET_PAGE_SIZE);
>  
> @@ -710,7 +710,7 @@ static int save_xbzrle_page(RAMState *rs, uint8_t **current_data,
>       * Update the cache contents, so that it corresponds to the data
>       * sent, in all cases except where we skip the page.
>       */
> -    if (!last_stage && encoded_len != 0) {
> +    if (!last_stage && encoded_size != 0) {
>          memcpy(prev_cached_page, XBZRLE.current_buf, TARGET_PAGE_SIZE);
>          /*
>           * In the case where we couldn't compress, ensure that the caller
> @@ -720,10 +720,10 @@ static int save_xbzrle_page(RAMState *rs, uint8_t **current_data,
>          *current_data = prev_cached_page;
>      }
>  
> -    if (encoded_len == 0) {
> +    if (encoded_size == 0) {
>          trace_save_xbzrle_page_skipping();
>          return 0;
> -    } else if (encoded_len == -1) {
> +    } else if (encoded_size == -1) {
>          trace_save_xbzrle_page_overflow();
>          xbzrle_counters.overflow++;
>          return -1;
> @@ -733,11 +733,11 @@ static int save_xbzrle_page(RAMState *rs, uint8_t **current_data,
>      bytes_xbzrle = save_page_header(rs, rs->f, block,
>                                      offset | RAM_SAVE_FLAG_XBZRLE);
>      qemu_put_byte(rs->f, ENCODING_FLAG_XBZRLE);
> -    qemu_put_be16(rs->f, encoded_len);
> -    qemu_put_buffer(rs->f, XBZRLE.encoded_buf, encoded_len);
> -    bytes_xbzrle += encoded_len + 1 + 2;
> +    qemu_put_be16(rs->f, encoded_size);
> +    qemu_put_buffer(rs->f, XBZRLE.encoded_buf, encoded_size);
> +    bytes_xbzrle += encoded_size + 1 + 2;
>      xbzrle_counters.pages++;
> -    xbzrle_counters.bytes += bytes_xbzrle;
> +    xbzrle_counters.encoded_size += encoded_size;
>      ram_counters.transferred += bytes_xbzrle;
>  
>      return 1;
> diff --git a/monitor/hmp-cmds.c b/monitor/hmp-cmds.c
> index 9b94e67879..6d3b35ca64 100644
> --- a/monitor/hmp-cmds.c
> +++ b/monitor/hmp-cmds.c
> @@ -295,8 +295,8 @@ void hmp_info_migrate(Monitor *mon, const QDict *qdict)
>      if (info->has_xbzrle_cache) {
>          monitor_printf(mon, "cache size: %" PRIu64 " bytes\n",
>                         info->xbzrle_cache->cache_size);
> -        monitor_printf(mon, "xbzrle transferred: %" PRIu64 " kbytes\n",
> -                       info->xbzrle_cache->bytes >> 10);
> +        monitor_printf(mon, "xbzrle encoded size: %" PRIu64 " kbytes\n",
> +                       info->xbzrle_cache->encoded_size >> 10);
>          monitor_printf(mon, "xbzrle pages: %" PRIu64 " pages\n",
>                         info->xbzrle_cache->pages);
>          monitor_printf(mon, "xbzrle cache miss: %" PRIu64 "\n",
> diff --git a/qapi/migration.json b/qapi/migration.json
> index eca2981d0a..bf195ff6ac 100644
> --- a/qapi/migration.json
> +++ b/qapi/migration.json
> @@ -62,7 +62,7 @@
>  #
>  # @cache-size: XBZRLE cache size
>  #
> -# @bytes: amount of bytes already transferred to the target VM
> +# @encoded_size: amount of bytes encoded
>  #
>  # @pages: amount of pages transferred to the target VM
>  #
> @@ -75,7 +75,7 @@
>  # Since: 1.2
>  ##
>  { 'struct': 'XBZRLECacheStats',
> -  'data': {'cache-size': 'int', 'bytes': 'int', 'pages': 'int',
> +  'data': {'cache-size': 'int', 'encoded_size': 'int', 'pages': 'int',
>             'cache-miss': 'int', 'cache-miss-rate': 'number',
>             'overflow': 'int' } }
>  
> @@ -333,7 +333,7 @@
>  #          },
>  #          "xbzrle-cache":{
>  #             "cache-size":67108864,
> -#             "bytes":20971520,
> +#             "encoded_size":20971520,
>  #             "pages":2444343,
>  #             "cache-miss":2244,
>  #             "cache-miss-rate":0.123,
> -- 
> 2.20.1
> 
--
Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK



  parent reply	other threads:[~2020-04-21 19:22 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-04-20  3:06 [PATCH v1 0/2] Migration xbzrle changes Wei Wang
2020-04-20  3:06 ` [PATCH v1 1/2] migration/xbzrle: replace transferred xbzrle bytes with encoded bytes Wei Wang
2020-04-20  9:29   ` Daniel P. Berrangé
2020-04-20  9:49     ` Wei Wang
2020-04-21 19:21   ` Dr. David Alan Gilbert [this message]
2020-04-22  2:51     ` Wei Wang
2020-04-24 10:47       ` Dr. David Alan Gilbert
2020-04-27  7:26         ` Wei Wang
2020-04-20  3:06 ` [PATCH v1 2/2] migration/xbzrle: add encoding rate Wei Wang
2020-04-20  9:30   ` Daniel P. Berrangé
2020-04-20 14:53   ` Eric Blake
2020-04-21  1:14     ` Wei Wang

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=20200421192106.GM3029@work-vm \
    --to=dgilbert@redhat.com \
    --cc=gloryxiao@tencent.com \
    --cc=kevin.tian@intel.com \
    --cc=peterx@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=quintela@redhat.com \
    --cc=wei.w.wang@intel.com \
    --cc=yi.y.sun@intel.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).