Re: repoSpanner sprint report

[Randy Barlow]

Greetings!

On Mon, 2019-10-14 at 21:05 +0300, Benson Muite wrote:
>  The tool seems useful. I wish there was a high level description of
>  the design. Seems like many issues are common with distributed nosql
>  databases. Are there any common techniques one can use to have good
>  performance? 
We should expand the project's documentation. The performance issues I
have identified as being most significant in repoSpanner are unique to
repoSpanner, so at the moment I don't think we're quite at the point of
having similar issues to nosql databases in terms of performance.
Perhaps we will once we've removed the current bottlenecks.

>  Is the typical use case in a single data center or for repositories
>  across multiple data centers? 
The intention of the project is to allow it to be used in
geographically diverse datacenters. In particular, we plan to use this
to bring the CentOS and Fedora dist-git projects together into a single
git forge. CentOS is already using it, and that deployment is currently
in three different datacenters in the United States (Phoenix, Raleigh,
and Chapel Hill).