What Car and Tire Manufacturers Can Teach Us about Cloud Scaling
Cloud for Scale.
Ever wonder why Auto Manufacturers don’t produce tires? Scale. The required manufacturing facilities, chemistry expertise and marketing channels required for cars and for tires is remarkably different, thus separating two highly connected industries. These economic forces of scale are what encourage Pirelli, Goodyear and Firestone to create alliances with auto manufacturers.
That same economic strategy is emulated by cloud offerings such as AWS, GCP and Azure in the mutually beneficial relationships they form with large corporations. To understand how large cloud providers will take advantage of economies of scale from companies who are moving their computation, storage and network requirements, we begin by examining their competing offerings.
For every need, there is a Cloud at-scale.
Azure Compute, AWS EC2 and GCP Compute Engine all offer a diverse range of computation infrastructure that can be optimized for the specific workloads they are running. Large media companies that render graphics and movies will require intense GPU loads, such as rendering the soft fur on Sully in Monsters Inc. or the Michael Bay action scenes in Transformers 12.
Cloud scaling and analytics.
Big data firms have been flocking to Spark for their analytical workloads which require high memory machines. Along those lines, deep learning has pushed Google to come up with a Tensor Processing Unit (TPU) available from Compute Engine, specifically designed to handle artificial intelligence computation requirements.
Renting out various specialized computation infrastructure for different workloads is especially advantageous as the steady drop in microchips due to Moore’s Law makes capital investments in high memory or high GPU machines unappealing.
Storage scaling on the Cloud.
These computation workloads are further complemented by inexpensive object storage to handle the various outputs. Azure Blob Storage, AWS S3 and Google Storage have made storing petabytes of Parquet and CSV trivial. Storing a few gigabytes of media renderings costs cents on the dollar, dropping the difficulties around creativity.
These storage features compliment data scientists that want to push the boundaries of the data their models consume, as a model is only as good as the data it is given. Data files, media renderings and machine learning models are all stored as file objects by cloud storage offerings thus having an object storage solution is remarkably useful for users.
Cybersecurity and the Cloud at-scale.
As these objects and compute infrastructure must be held safely, cybersecurity also benefits from economies of scale from these cloud providers. My comparison for modern day cybersecurity is akin to building a house, while preventing any mosquitoes from flying in, during construction.
This requires every door, window and light switch to be highly tested and installed properly. In the case of a cloud, this is every user login, every network connection and database to have logging, access policies and bandwidth accounted for. An on-premises data solution would have to build its own systems or rely on collections of 3rd party software that would not benefit from the broad scale, integrated usage of cloud providers.
Advantages of a future with the Cloud at-scale.
As cloud providers mature their offerings and scale out their relationships, their scale advantages will build. Their network, compute and storage offerings will drop in price and become safer. Their experience setting up clouds and customer relationships will build as well, developing more accurate advice as to the challenges companies will face.
