quickstep-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "J. M. Patel" <jmp.quicks...@gmail.com>
Subject Re: Quickstep paper
Date Thu, 22 Mar 2018 23:52:09 GMT
Here is a proposed abstract and title. Darts?

------------------
TITLE: Apache Quickstep: A high-performance data analytics platform for data-center-in-a-box
servers (aka. all modern servers)

ABSTRACT: Modern servers pack enough storage and computing power that just a decade ago was
spread across a modest-sized cluster. To exploit the full potential of modern servers, which
can be viewed as a data-center-in-a-box, one needs to fundamentally build into the underlying
data platform mechanisms for a) unbounded intra and inter-operator parallelism, b) sophisticated
query optimization techniques, and c) hybrid storage mechanisms. Apache (incubating) Quickstep
is data platform that is built from ground up, and in a principled way, to meet this need.
To keep the project focused, the project’s initial target is read-mostly in-memory data
warehousing workloads in single-node settings. This talk describes the architecture and design
of this initial version of Quickstep. 

This talk will also describe results from a performance evaluation comparing Quickstep to
newer systems like Apache Spark and traditional systems like PostgreSQL. Quickstep outperforms
these systems by over 10X in most cases. 

This talk will also discuss how the hardware trends point to higher storage and compute densities
in individual servers. The natural consequence of this trend is that focusing on the Quickstep
approach of “scaling-up” to exploit the full hardware parallelism is critical for the
future of high-performance data platforms. This scaling-up approach is complementary to “scaling-out,”
which is also on the Quickstep project roadmap, and this talk will also describe the initial
plan for a distributed version of Quickstep.
------------------

Cheers,
Jignesh

On 3/17/18, 7:06 PM, "Julian Hyde" <jhyde@apache.org> wrote:

    ApacheCon is for the community, so keynotes are chosen entirely on merit. Bigger audience
than the regular talks, but fewer minutes, and it would be good if there are larger themes
than simply what the project does.
    
    I recommend that you submit a keynote and a regular talk. If people enjoy the keynote
they will come to the regular talk to hear more details. Writing an abstract doesn’t take
very long.
    
    Julian
    
    
    > On Mar 17, 2018, at 4:59 PM, J. M. Patel <jmp.quickstep@gmail.com> wrote:
    > 
    > Hi Team: I'm starting to work on the abstract for this (will circulate it soon).

    > 
    > Apache Con has two options for talk submission: Keynote and Standard. I'm thinking
of submitting to the Keynote category. Any comments? I figured go-big-or-go-home.
    > 
    > Not sure if ApacheCon is like some of the other industry-style conferences in which
keynotes are favored from sponsors. 
    > 
    > Also, what do folks think of working on a release sometime early summer? We can try
to see if Marc can help if he has settled down from the recent excitement (https://techcrunch.com/2018/01/25/ford-acquires-autonomic-and-transloc-as-it-evolves-its-mobility-business/).

    > 
    > Cheers,
    > Jignesh 
    > 
    > On 2/13/18, 10:09 PM, "J. M. Patel" <jmp.quickstep@gmail.com> wrote:
    > 
    >    Good tips Julian! 
    > 
    >    Cheers,
    >    Jignesh  
    > 
    >    On 2/9/18, 6:00 PM, "Julian Hyde" <jhyde@apache.org> wrote:
    > 
    >        ApacheCon is a good opportunity to attract community, so I suggest you think
of the talk as a sales pitch. You need to convince members of the audience to use Quickstep
in their application or data engine, and then to start contributing. The fact that the Quickstep
is high-performance is necessary but not sufficient. You also need to demonstrate that it
is easy to use Quickstep in their project. From your talk by all means reference your research
and experiments to back up your assertions about speed etc.
    > 
    >        Julian
    > 
    > 
    >> On Feb 8, 2018, at 5:00 PM, J. M. Patel <jmp.quickstep@gmail.com> wrote:
    >> 
    >> Hello Quicksteppers: VLDB (research track) has accepted a paper on Quickstep.
The current (temporary) version of the paper is at: http://pages.cs.wisc.edu/~jignesh/publ/quickstep.pdf
If any of you have feedback, please reply to this email. 
    >> 
    >> 
    >> 
    >> Perhaps this can also serve as material that we can use for the ApacheCon submission?
    >> 
    >> 
    >> 
    >> Cheers,
    >> 
    >> Jignesh
    >> 
    >> 
    >> 
    > 
    > 
    > 
    > 
    > 
    > 
    
    



Mime
View raw message