Google is ramping up efforts to entice customers to its cloud compute platform more formally known as Google Cloud Platform (GCP). To that end, the company has now announced new Compute Engine machine types with up to 96 virtual CPUs and up to 624GB of memory. If that sounds like a lot, that's because it is, since the platform and its processing units are meant to drive applications such as satellite image analysis or media rendering and production through the platform's virtual machines (VM). The new machine types have been created specifically to beef up the efficiency and power while cutting back the number of virtual machines needed. That reduction in machines needed should, in turn, also serve to cut some of the total overhead for customers, effectively driving down the cost for customers who take advantage of the GCP's technology – particularly when they scale up.
Google also says that the increase in available cores is thanks to Intel's latest Xeon-branded Scalable Skylake processors and that the increase to the available processing power goes well beyond just the number of cores provided. In fact, the company boasts that by comparison to the stock scikit-learn, running instances via the new Intel Distribution for Python: scikit-learn on its cloud platform can result in performance that's up to 23 times faster. The company does say that the new machines are most efficient in terms of scaling applications up when all 96 virtual CPUs are used, with smaller gains noted from smaller implementations of the new machine type. However, customers can customize their purchase so those that have instances which only need a more conservative boost aren't wasting money on what they don't need or so that memory can be extended even further. It goes without saying that boosts in performance will also depend on the use-case. Having said that, at least one customer – namely, Tim Kelton, a Co-Founder at Descartes Labs – says that its company noted substantial improvement to performance despite working to get petabytes of global satellite imagery "science-ready." Specifically, Kelton reports a 38-percent improvement in image compression performance and a 23-percent increase in image expansion performance.
In the meantime, anybody who might be interested in trying out the new machine types doesn't have to buy in right away, either. The search giant is currently offering a promotional one-month free trial, including $300 in cloud credits to give new customers a chance to check out its new machine types for whatever instances they may need to run in the cloud. Unfortunately, the new virtual machine types aren't yet available in every region, although they are available on pretty much every continent. While the company is working to expand their availability, they are currently only available in nine of the GCP's regions.