Open Stack

Hi,

My name is Renat Akhmerov. I decided to run for Mistral PTL for Mitaka release cycle.

This is the first time when I’m doing it officially after Mistral was accepted into Big Tent.
In fact, I’ve been driving the project for just a little less than 2 years by now and I've put a lot of
my energy into this initiative by designing its architecture, coding (including initial PoC versions),
reviewing nearly every single patch coming in and presenting Mistral at every conference that
I could including OpenStack summits. Of course, I wasn’t doing it alone and I couldn’t find enough
words to express how thankful I am to folks from Mirantis, StackStorm, Huawei, Alcatel-Lucent,
HP, Ericsson and other companies. You all have done a great work and I’m proud to be a part of
such a great team.

Although a lot has been done and we certainly have achievements in the form of users who
use Mistral in production, there’s a lot more ahead. And below is what I think we need to focus on
during Mitaka cycle.

HA and maturity

Making Mistral truly stable and mature technology capable of running in HA mode. I have to admit
that so far we haven’t been paying enough attention to high-load testing and tuning. And my belief
that it’s a high time we started doing it. Some of the issues are known to us and we know how we
should be fixing them. Some have to be discovered. In my opinion, what we’re missing now is a
comprehensive understanding of, believe it or not, how Mistral works :) This may sound strange,
but I really mean is that we need to know in very tiny detail how every Mistral transaction works
in terms of potential race conditions, isolation level, concurrency model etc. In my strong opinion,
this is a prerequisite for everything else. Having said that, I am going to bring more expertise
on the project to fill this gap: either by attracting corresponding people and by planning more
time for the current team members to work on that.

Apart from that I find it very important to stop developing two many new features in workflow engine
and do a proper refactoring of it. In my strong opinion, Mistral engine started suffering from
squeezing more and more functionality into it. It’s generally normal but I believe that we need to
simplify the code base by cleaning it up wisely and at the same time improving test coverage
accounting for all kind of corner cases and negative scenarios.

Use cases

This is probably the most tricky part about this project and I believe I personally should have done
much better job of clearly explaining Mistral value for the industry. I plan to change the situation
drastically by providing battle proven scenarios where it’s hard or nearly impossible to avoid using
Mistral. Also recording screencasts and writing cookbooks is part of the plan.

UI

Thanks to engineers from Huawei and Alcatel Lucent who’ve done a good job in Liberty to move Mistral
UI to a much better state. Most of basic CRUD functionality is there and this work keeps going on.
However, I still see a lot of ways how to advance Mistral UI and make it really remarkable.
For example, one specific thing that I’d really like to work on is workflow graph visualisation
(for both editing and monitoring running workflows).

I find it very important particularly because having good UI would help us to build even larger
community around the project. Just because it would be easier to deliver a message of what the
project goal is.

What else

Other things that I’d like to pay attention to:
Solving guest VMs access problem
More intellectual task scheduling mechanism accounting for workflow priorities (FIFO but on workflow level)
New REST API (don’t confuse with DSL, or workflow language) on which we’ve almost agreed within the team
Improving significantly CLI so that it becomes truly convenient and fun to use

Eventually, my goal is to build a really useful and beautiful technology

Renat Akhmerov
@ Mirantis Inc.



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20150915/32fd17d6/attachment.html>