<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Sammed Kagi on Medium]]></title>
        <description><![CDATA[Stories by Sammed Kagi on Medium]]></description>
        <link>https://medium.com/@sammedkagi?source=rss-ca3afb9c1d56------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*_09a895sy8plbewn.</url>
            <title>Stories by Sammed Kagi on Medium</title>
            <link>https://medium.com/@sammedkagi?source=rss-ca3afb9c1d56------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 07:47:28 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@sammedkagi/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[GPU Accelerated Databases]]></title>
            <link>https://becominghuman.ai/gpu-accelerated-databases-459b0e476a6e?source=rss-ca3afb9c1d56------2</link>
            <guid isPermaLink="false">https://medium.com/p/459b0e476a6e</guid>
            <category><![CDATA[gpu]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[database]]></category>
            <category><![CDATA[gpu-accelerated-databases]]></category>
            <dc:creator><![CDATA[Sammed Kagi]]></dc:creator>
            <pubDate>Sun, 24 Mar 2019 05:54:38 GMT</pubDate>
            <atom:updated>2019-03-25T15:15:18.656Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/678/1*CeHbhTKzOldz9ri5SLXHGQ.jpeg" /></figure><p><strong>What is a GPU?</strong><br>GPU stands for Graphics Processing Units which was initially built for rendering images, videos, 3D animations etc. A GPU uses many cores and works by parallelizing the task or multi-tasking and perform quick math operations, whereas a CPU uses very few cores and does the work sequentially. By this, we understand that GPU is faster than the CPU. After realizing its potential it has been put to use in multiple fields which require high computation in a short duration of time.</p><p><strong>Evolution of Database:</strong><br>If we look at the 1990s and 2000s, the data warehouses, relational database management systems were prominent and analytics were being readily available. But as we are scaling and growing in the world the data growth is exponential and analytics were getting expensive. Then the distributed storage systems were introduced which resulted in massive amounts of affordable storage but were not efficient performance wise. With the invention of in-memory databases for quick processing but the memory was a bottleneck in this case. So, today the GPU accelerated databases use their power of parallel computation to solve most of the problems.</p><p><strong>GPUs in Databases:</strong><br>With the advent of Deep Learning and Neural Networks, the requirement of high-end computation and the importance of GPU has been realized to a great extent. Another field where GPU can be used is in Databases. The rapid growth in the size of the datasets and the pressing need to process and query, these huge datasets has led to the usage of GPUs in Databases and these databases are called GPU accelerated Databases. In layman terms, GPUs that can run database queries or instructions on multi-billion row datasets in milliseconds.</p><p>The architecture in the GPU databases is mostly Master-Slave architecture. In this a central CPU based node farms subqueries (Nested SQL queries) to other database instances which are GPU accelerated. The node farms are the masters and the database instances, receiving the subqueries, are the slaves. The slaves will now execute the subqueries and since they are on separate servers this job will be done parallelly, send the result back to the node farms which combine the results and sends it to the client.</p><p>The advantages we have for using GPU accelerated database is real-time analysis and visualization of huge datasets can be done. There is a performance increase of over 100 times when compared to traditional RDBMS, NoSQL, In-Mem databases and around 75% reduction in infrastructure costs if used in industries.</p><p>The following are a few of the many applications of GPUs:<br>1. <strong>Geospatial Analysis:</strong><br>Geospatial analysis utilizes a lot of data which are positioned by space and time. This data can become very complex since it is stored in buckets depending upon the resolution required and hence GPU is ideal for processing such information and high-performance visualization capabilities <br>2. <strong>Machine Learning and Predictive Analysis of huge data:</strong><br>As time goes by we find the size of datasets increasing exponentially with an increase in the complexity of data. Loading the entire datasets to execute learning algorithms would be difficult. So, in GPU accelerated databases we can create a User Defined Functions framework to run the code directly within the database. This is called in-database analysis or processing. An in-database analysis is highly flexible means of performing complex, advanced analysis at scale via grid computing and linking the idle processing power of every computer in the network to take on a single job, divided into multiple tasks.</p><p>There are a few Commercial GPU-optimized databases which are focussed on data analytics applications. To name a few:<br>1. <a href="https://www.brytlyt.com/">Brytlyt</a>: Can cost effectively query multi-billion row datasets in seconds<br>2. <a href="https://blazingdb.com/#/">BlazingDB</a>: Run fast, simple SQL queries on huge datasets<br>3. <a href="https://www.kinetica.com/">Kinetica</a>: Perform standard SQL queries on billions of rows in microseconds and also visualize the results. This platform helps in executing machine learning models and ingesting large amounts of streaming data.<br>4. <a href="https://www.omnisci.com/">OmniSci</a>: Compiles SQL queries into machine code compatible with Nvidia Corp GPUs, Intel Corp, X86, IBM Corp and Power CPUs.<br>5. <a href="https://sqream.com/">SQream</a>: Real-time columnar database processing</p><p><strong>GPU Accelerated databases in Blockchain:</strong><br>One of the bases of Blockchain technology is consensus protocols. Consensus protocols involve highly parallel computations on relatively small amounts of data and this is where the GPUs work at their best. A few of the GPU database vendors have sold their solutions, of utilizing the maximum extent of GPU, to cryptocurrency-mining market for handling consensus protocols and other blockchain related workloads.</p><p><strong>Limitations of GPU-accelerated databases:</strong><br>As promising as it sounds but implementing a GPU accelerated database is tough. This is because GPUs heavily rely on the fact the queries can be parallelized and produce the same output as that of sequential execution. They are ill-suited for accelerating any database processing which cannot be parallelized, or which involve floating point and other numeric processing. It also does not support where there is a requirement of data to be moved back and forth across the system from the main central processing units.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=459b0e476a6e" width="1" height="1" alt=""><hr><p><a href="https://becominghuman.ai/gpu-accelerated-databases-459b0e476a6e">GPU Accelerated Databases</a> was originally published in <a href="https://becominghuman.ai/">Becoming Human: Artificial Intelligence Magazine</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>