Why tokudb's defraction is good?

郭枫

unread,

Aug 26, 2015, 9:30:10 PM8/26/15

to tokudb-user

I saw a ppt shows that tokudb's defraction is very good, but it doesn't say why and how tokudb do defraction, any one knows?

Leif Walsh

unread,

Aug 27, 2015, 10:05:21 AM8/27/15

to 郭枫, tokudb-user

It fragments normally, but with large blocks, it still achieves range query speeds close to disk bandwidth, unlike small block size btrees like innodb, which slow down considerably due to fragmentation. It doesn't actively defragment.

--
Cheers,
Leif

On Wed, Aug 26, 2015 at 9:30 PM, 郭枫 <leafo...@gmail.com> wrote:

I saw a ppt shows that tokudb's defraction is very good, but it doesn't say why and how tokudb do defraction, any one knows?

--
You received this message because you are subscribed to the Google Groups "tokudb-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email to tokudb-user...@googlegroups.com.
To post to this group, send email to tokud...@googlegroups.com.
Visit this group at http://groups.google.com/group/tokudb-user.
For more options, visit https://groups.google.com/d/optout.

郭枫

unread,

Aug 27, 2015, 10:13:42 PM8/27/15

to Leif Walsh, tokudb-user

Thx~

郭枫

unread,

Aug 28, 2015, 3:40:46 AM8/28/15

to Leif Walsh, tokudb-user

Thx,but I'm still comfusing:

1.Why “ with large blocks, it still achieves range query speeds close to disk bandwidth”？

2.“tokudb_read_block_size” is 64 before compression,after compression，it maybe even smaller than InnoDB‘s default 16K，so it's not large~

2015-08-27 22:05 GMT+08:00 Leif Walsh <leif....@gmail.com>:

Leif Walsh

unread,

Aug 28, 2015, 10:03:33 AM8/28/15

to 郭枫, tokudb-user

For range queries, the read block size doesn't matter, the effective block size is 4MB before compression.