1000x speed to Google Colab using Techila Distributed Computing Engine
Articles,  Blog

1000x speed to Google Colab using Techila Distributed Computing Engine


[Music] Have you seen the Techila Distributed Computing Engine (TDCE) demo where we are running Jupyter notebook interactively in Google scale using Techila Distributed Computing Engine and 1000 Google CPU cores? Have you asked yourself can I do this in my Colab? In this demo we will use the “Asian” code example that can be found in the Techila SDK. We are using 1000 CPU cores from the Google Cloud Platform. Combining this with TDCE allows the Notebook in our Colab to run nearly 1,000 times faster. You can click the “Run Cell” button, just as if you were running the code locally. TDCE parallelizes and distributes the work automatically to your TDCE system in your own Google account. You will receive the results directly to your Colab, just as if you have run the code locally, only a lot faster. Along with the 1000 CPU cores the Techila Distributed Computing Engine gave our Notebook a 911 fold acceleration. So instead of waiting two and a half days, we got the results in just four minutes! Did you ask yourself “How can I do this in my Colab?” In the previous tutorial we saw how to set up a Google Colab for use with TDCE. In this demo we scaled our system for more throughput. Techila is available in the GCP Marketplace. You can use it simply with a Google account. It has an easy Configuration Wizard user interface that makes it easy to adjust the throughput on demand 250 4-core Virtual Machines translates to 1000 processor cores. Techila supports also preemptable VMs which can save up to 80% in infrastructure costs. Click the “Start” button. The Techila Workers will start automatically in your own Google account. You can simply click the “Run” button in your Colab. You can develop, test, analyze, and iterate efficiently. TDCE manages the system automatically. Nobody likes waiting! [Music]

Leave a Reply

Your email address will not be published. Required fields are marked *