For Google Cloud Platform customers, Google has a recommendations engine for machine types and sizing, Google's Urs Hölzle said at the Structure conference on cloud computing on Tuesday.
"I hope in five years, 1 percent of Cloud customers knows the word 'machine type' and 99 percent say, 'I never thought about it," said Hölzle, Google's senior VP for technical infrastructure, said at the San Francisco conference.
Infrastructure, he said, "is one of those things where if you do it right, people don't care about it... This is something we're dealing with and not you."
The Google comments rhyme with what's increasingly being pitched by the large infrastructure as a service players. For instance, Amazon Web Services has its Lambda effort, also known as a serverless managed service. The general idea is that AWS handles the infrastructure and manages it. Customers focus on the function required.
Also: TechRepublic: AWS vs Microsoft Azure: Understanding the serverless application trend
Google tries in several places to keep the customer from thinking about it, he said. For instance, users see the same API whether they're using high-end hot storage or low-end archival storage at less than a cent a month per gigabyte.
Asked where the bottleneck is today, Hölzle said that currently, it's cheap, fast storage. Additionally, he said, compute is a pressure point because of machine learning. "Tomorrow, if NVRAM happens, it will be networking," he said.