Google Redefines Multi-Cloud Computing
I came across an interesting article by Bob O’Donnell from TECHnalysis Research, and could not resist sharing with you.
Ask any IT professional or tech industry observer to name the top trends they see, and you would inevitably hear many variations on the theme of cloud computing. In particular, the phrase multi-cloud computing is likely to pop up.
At its simplest level, multi-cloud computing simply means that a company is choosing to use several different public cloud vendors, such as Google’s GCP, Amazon’s AWS, Microsoft’s Azure, etc. Companies often choose this approach to avoid vendor lock-in, provide redundancy, and help deal with data sovereignty issues where certain data has to be stored in a given country for legal or regulatory purposes (and not every cloud provider has data centers in every region).
In addition, certain cloud providers have developed expertise in specific areas and companies are choosing to run workloads with these providers to take advantage of those unique capabilities.
With one of the many announcements that Google Cloud CEO Thomas Kurian made at the debut of their multi-week Google Cloud Next virtual event, it’s clear that Google is looking to expand the definition of multi-cloud. Specifically, Google announced a new offering that works across multiple datasets and workloads running on several different providers. The new BigQuery Omni capability lets companies leverage Google’s deep heritage in search and querying to run its unique analytics capabilities across data stored in Google Cloud, AWS, and, later this year, Azure.
At a high level, BigQuery Omni essentially enables a macro-level grouping of multiple cloud platforms, workloads and datasets under a single level of control, all while leveraging one of the widely acknowledged strengths that Google has on its competition. This federated, multi-cloud, data analytics architecture is exactly the kind of clever and aggressive move that’s helping Google start to win the kinds of big-name customers that the company also highlighted at this year’s Cloud Next, including 5G carriers Telefonica, Vodaphone, and now Verizon, as well as Deutsche Bank, Renault, Fox Sports, and others.
One of the key benefits of how BigQuery Omni works is that Google code runs natively on the different cloud platforms and accesses data locally from the storage resources within those platforms, thereby avoiding the expensive and time-consuming process of transferring data across platforms (otherwise known as data egress). Plus, the BigQuery Omni offering is structured such that users can utilize the same basic BigQuery interface (running on GCP) to create the SQL commands necessary to query the databases, and then those requests are computed locally within each environment. The results from multiple sources can then all be transferred back to a single pane-of-glass UI for easier analysis or stored within each platform to avoid any cross-cloud move of data.
Part of the reason this is all possible is that, since its inception over 10 years ago as a Google internal tool, BigQuery, with its Dremel query engine, has separated the compute elements from the storage elements. That architecture wasn’t originally built with multi-cloud computing in mind, but by combining it with Google Anthos’ ability to run or transfer workloads across different cloud platforms through an abstraction layer, the company was able to create a solution that essentially treats different cloud platforms as if they were different regions within a single platform.
The idea of turning multi-cloud computing into an extended version of a single cloud platform—which BigQuery Omni does—seems to offer a number of interesting benefits. First, from a user’s perspective, this meta-platform concept provides a consistent way to access multiple data sets across multiple platforms, all through a single interface. More importantly, it opens up the idea of thinking about cloud computing resources overall in a significantly more flexible manner. It’s not difficult to imagine, for example, that Google (or other cloud providers) could create other types of multi-platform solutions that let them leverage some of their unique IP. Using new types of machine learning or neural network training algorithms across multiple datasets stored in multiple locations, for instance, could be an interesting new option.
The bottom line is that by leveraging both its BigQuery and Anthos assets, Google has put together an intriguing new twist on multi-cloud computing. Offerings like BigQuery Omni have the potential to open up the cloud world to new types of unified macro-level meta-platform offerings that, ironically, can further break down the walls that exist between the different cloud platforms at the same time. It remains to be seen how effective they prove to be in the real world, but conceptually it breaks some interesting new ground.