Distributed memory

<h2 id="programming-distributed-memory-machines">Programming distributed memory machines</h2>
<p>The key issue in programming distributed memory systems is how to distribute the data over the memories. Depending on the problem solved, the data can be distributed statically, or it can be moved through the nodes. Data can be moved on demand, or data can be pushed to the new nodes in advance.
</p><p>As an example, if a problem can be described as a pipeline where data <i>x</i> is processed subsequently through functions <i>f</i>, <i>g</i>, <i>h</i>, etc. (the result is <i>h</i>(<i>g</i>(<i>f</i>(<i>x</i>)))), then this can be expressed as a distributed memory problem where the data is transmitted first to the node that performs <i>f</i> that passes the result onto the second node that computes <i>g</i>, and finally to the third node that computes <i>h</i>. This is also known as <a href="/facts/Systolic_array/uXg0ClQ0">systolic computation</a>.
</p><p>Data can be kept statically in nodes if most computations happen locally, and only changes on edges have to be reported to other nodes. An example of this is simulation where data is modeled using a grid, and each node simulates a small part of the larger grid. On every iteration, nodes inform all neighboring nodes of the new edge data.
</p>
<h2 id="distributed-shared-memory">Distributed shared memory</h2>
<p>Similarly, in <a href="/facts/Distributed_shared_memory/IboHJvt6">distributed shared memory</a> each node of a cluster has access to a large shared memory in addition to each node's limited non-shared private memory.
</p>
<h2 id="shared-memory-vs-distributed-memory-vs-distributed-shared-memory">Shared memory vs. distributed memory vs. distributed shared memory</h2>
<ul><li>The advantage of (distributed) shared memory is that it offers a unified address space in which all data can be found.</li>
<li>The advantage of distributed memory is that it excludes race conditions, and that it forces the programmer to think about data distribution.</li>
<li>The advantage of distributed (shared) memory is that it is easier to design a machine that scales with the algorithm</li></ul>
<p>Distributed shared memory hides the mechanism of communication, it does not hide the latency of communication.
</p>
<h2 id="see-also">See also</h2>
<ul><li><a href="/facts/Memory_virtualization/G7wM0Wb0">Memory virtualization</a></li>
<li><a href="/facts/Distributed_cache/ry6do0bs">Distributed cache</a></li></ul>

<h2 id="references">References</h2>

<ol>
<li id="fn:1"><p>Pardo, David; Matuszyk, Paweł J.; Puzyrev, Vladimir; Torres-Verdín, Carlos; Nam, Myung Jin; Calo, Victor M. (2021). "Parallel implementation". Modeling of Resistivity and Acoustic Borehole Logging Measurements Using Finite Element Methods. Elsevier. doi:10.1016/C2019-0-02722-7. ISBN 978-0-12-821454-1 – via ScienceDirect. Distributed memory refers to a computing system in which each processor has its memory. Computational tasks efficiently operate with local data, but when remote data is required, the task must communicate (using explicit messages) with remote processors to transfer data. This type of parallel computing is standard on supercomputers equipped with many thousands of computing nodes. <a href="978-0-12-821454-1" target="_blank">978-0-12-821454-1</a> <a href="#fnref:1" class="footnote-back-ref">↩</a></p></li>
</ol>

Distributed memory open-in-new

Distributed memory