<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Archiving and Interchange DTD v1.0 20120330//EN" "JATS-archivearticle1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink">
  <front>
    <journal-meta />
    <article-meta>
      <title-group>
        <article-title>Composable, Hardware-Accelerated Executors (Lightning Talk)</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <string-name>Felipe Aramburu</string-name>
          <email>felipe@voltrondata.com</email>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <contrib contrib-type="editor">
          <string-name>Composable Data Management Systems, Hardware Accelerated Executors</string-name>
        </contrib>
        <aff id="aff0">
          <label>0</label>
          <institution>Voltron Data Inc.</institution>
          ,
          <addr-line>650 Castro Street Suite 120, PMB 96571, San Francisco, CA 94041</addr-line>
          ,
          <country country="US">USA</country>
        </aff>
      </contrib-group>
      <pub-date>
        <year>2022</year>
      </pub-date>
      <abstract>
        <p>By design, a composable data system architecture provides the flexibility to use many execution paradigms. At Voltron Data, to prove this out, we are developing an execution system that uses the same logical plans on both CPU and GPU executors: (1) GPU executor is RAPIDS cuDF from NVIDIA, and (2) the CPU executor is Velox from Meta. The number of FLOPs is steadily increasing and the costs for memory is decreasing quickly [1], [2].</p>
      </abstract>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>-</title>
      <p>LGOBE</p>
    </sec>
  </body>
  <back>
    <ref-list />
  </back>
</article>