Google Cloud Platform's big pitch to enterprises is likely to resonate well. Why? Google used a mashup of tried-and-true messages that enterprises have bought into before. The keynote from Google Cloud Platform's conference in San Francisco featured chief Diane Greene, Alphabet Chairman Eric Schmidt and a series of execs and innovators to outline new tools for everything from infrastructure management to big data to analytics and machine learning. Overall, Google's enterprise cloud subtext was clear. The company wants enterprise customers to know that its cloud efforts aren't simply a beta.
Google Cloud announced Monday that it has acquired Qwiklabs, a startup that offers lab-learning environments for cloud platforms and infrastructure software vendors. Qwiklabs, founded in 2012, offers step-by-step instructions for using popular cloud services. It also lets users test different use cases and train teams. More than half-a-million users have collectively spent over 5 million hours on the platform. "With Qwiklabs, we're closing the IP skills gap in the cloud," Jason Martin, director of professional services for Google Cloud, wrote in a blog post.
Google has announced that early next year it will be adding graphics processing units (GPUs) as a service to its Cloud Platform in order to better compete with its rivals. GPU as a service is already available on Amazon Web Services, Microsoft Azure and IBM's Bluemix, but Google's Cloud Platform will seek to differentiate itself by offering a variety of GPUs for customers to choose from. The company will allow users the option to use two AMD FirePro S9300s or either a Nvidia Tesla P100 or K80. Google has also decided to change the way it prices GPU usage to make it more affordable for its customers. Generally users pay by the hour when using GPU as a service but the company has instead opted to price by the minute so that even smaller organizations and business can make use of its services.
Firstly, follow the setup guide to install the Google Cloud Machine Learning SDK. This will also ask you to install TensorFlow. Make sure to specify the correct number of training classes (--num_classes) and number of samples in your validation set (--valid_batch_size). This will differ depending on the number of files you've downloaded and how the data has been divided. Check the training source for other flags you can specify. Your trained model will be exported to /tmp/model/00000001 by default.
With over 10,000 attendees, the event resembled Google I/O – the most popular developer conference hosted by Google. Having gone through a major branding exercise that consolidated Google Cloud Platform (GCP) and G-Suite under Google Cloud, the company wanted to send a strong message that it is serious about the cloud. Google was successful in establishing itself as a credible player in the cloud market. Google's top executives didn't miss the opportunity to pitch its cloud platform to enterprises. Though Amazon and Microsoft are ahead of the game, it is using some of its core strengths to deliver a differentiated platform to customers.