Hi Niels,
Here is something for DHT2:
DHT2:
* Why DHT2:
To address consistency and correctness of operations, complexity and
scale requirements in the gluster IO path, keeping performance
characteristics unchanging.
The issue with current DHT is that a directory is present in all
subvolumes of DHT, this leads to a few consistency issues when dealing
with entry operations on directories or files. It also leads to scaling
limitations when dealing with larger subvolume counts. Resolution of the
consistency and correctness issues is becoming more complex and also
involves some fine grained locking across the network, which would
increase the complexity of the solution and also potentially decrease
the performance of the file system.
Refer:
http://www.gluster.org/community/documentation/images/f/fd/Summit-dht2.odp
* Outline of approach:
To resolve some of the above inadequacies in the current DHT design,
DHT2 aims to simplify the on disk data, and retain a directory in a
single subvolume and hence address the various pitfalls that are present
in the current distribution model.
It also addresses some of the other problems in DHT w.r.t layouts stored
per directory and simplifies the same in order to retain consistency and
correctness properties.
In the move to retain a directory on just a subvolume of DHT2,
separation of namespace/inode and data becomes essential to make the
distribution of data even. As a result DHT2, introduces a metadata
sub-volumes and a data sub-volumes into gluster.
* Impacts:
- DHT2 changes the on disk format for gluster and hence does not provide
an upgrade path from older DHT based volumes. It is not intended to in
place convert older gluster volumes to newer gluster volumes as well.
- The on disk format changes also means changes to some other xlators
like quota and possibly bit-rot and change log. Impact to these xlators
are being analyzed
Shyam
On 01/20/2016 05:41 AM, Niels de Vos wrote:
Hi all,
on Saturday the 30th of January I am scheduled to give a presentation
titled "Gluster roadmap, recent improvements and upcoming features":
https://fosdem.org/2016/schedule/event/gluster_roadmap/
I would like to ask from all feature owners/developers to reply to this
email with a short description and a few keywords about their features.
My plan is to have at most one slide for each feature, so keep it short.
Thanks,
Niels
_______________________________________________
Gluster-devel mailing list
Gluster-devel@xxxxxxxxxxx
http://www.gluster.org/mailman/listinfo/gluster-devel
_______________________________________________
Gluster-devel mailing list
Gluster-devel@xxxxxxxxxxx
http://www.gluster.org/mailman/listinfo/gluster-devel