Google plans to release pre-emptible virtual machines, which provide short-term capacity at a fixed low price. Simultaneously, the search engine giant declared plans to cut the prices of its compute engines by up to 30%.
Google offers a cheaper solution
Since the launch of the company’s Compute Engine in 2013, VM prices have decreased by more than 50%. In addition, Google has asserted that now its Cloud Platform is 40% cheaper for most workloads compared to other public cloud services.
According to Google, the reduction in the prices of virtual machines in the U.S. amounts to 30% for micro, 20% for standard configuration, 15% each for high memory and small services, and 5% for high CPUs. Pre-emptible VMs are similar to normal virtual machines, barring their availability that depends on system supply and demand. Moreover, they utilize leftover capacity in the company’s data centers for their functioning, which enables Google to offer such services at significantly low prices. Recently, the search company stated that pre-emptible VMs will cost 70% less than an average regular VM.
In a blog post, Paul Nash, the company’s senior product manager, stated that there exists a “variety of computer tasks that fit nicely into this pricing model.” Presently available in beta, the pre-emptible VM service could be useful for fault-tolerant workloads and can easily be delivered to a number of virtual machines. Moreover, jobs that need plenty of computational power, for instance, simulation and modelling, data analytics, genomics, etc., could run intermittently and can also be carried out if one or more devices being used get disconnected. Further, users will be provided with a 30-second time window to save their work, as pre-emptible VMs cannot work continuously for more than 24 hours.
How it’s different from Amazon’s service
The company’s pre-emptible service, to an extent, resembles Amazon’s Spot Instances in the way that both services have been created to facilitate interrupted jobs. However, the prices of the services offered by Google are fixed, while Amazon’s cost can vary depending on demand. The search giant suggested that the prices of its Cloud Platform will be set according to Moore’s Law.
Further, few of the company’s cloud users have already indicated plans to use the pre-emptible service. One such customer is Descartes Labs, a satellite imagery startup which recently needed 30,000 processors to handle a petabyte of images by NASA. Citadel is another user that is looking to use Google’s latest cloud services.
