Companies turn to the public cloud for various reasons, but it has become the defacto standard for every size organization—from startup, to SMB, to the enterprise.
While each cloud vendor puts their own name on their offering, they are all, for the most part, using the same open source software—meaning they all use the same products. So having settled on the cloud, which cloud provider should an organization use? Amazon has the most customers, but the Google Cloud Platform (GCP) is another large player in this market.
Making an informed decision requires a basic understanding of what GCP is, how much it costs, what products they offer, and what certifications your staff could get—especially as compared to the arguable industry leader.
What is the Google Cloud Platform (GCP)?
The Google Cloud Platform (GCP) is a suite of cloud services that offers server space on virtual machines, internal networks, VPN connections, disk storage, machine language SaaS (Software as a Service) applications, and even something called TPU (Tensor Processing Units).
Again, Google puts a different name on their machine learning tools, but neither Google nor Amazon invented this technology. A server in Google's data center uses the same Intel 8086 architecture as do servers in Amazon. What makes them different is pricing, features, and customer support.
What is Google Cloud Platform (GCP)? now.
Google Cloud pricing
As with other companies, Google's prices for virtual machines (VMs) vary with CPU type and memory. Monthly subscription fees can mount quickly if machines are not sized correctly or there is no mechanism to monitor prices carefully. The key is to task someone to become an expert at and use the Google cost calculator to monitor the budget.
Below is a brief list of monthly Google Cloud prices for the products the typical customer is most likely to use.
Standard environments are subject to quotas. Standard prices range from $0.05 to $0.40 per hour.
Flexible prices are per hour per vCPU ($0.0526), per GB of memory ($0.0071), and per GB of persistent disk storage ($0.0400).
Storage is $0.02 per GB/month or $0.01 per GB/month for long-term storage.
Streaming inserts are $0.01 per 200 MB.
$0.00002400 per vCPU. $0.00000250 per GB of memory. Network pricing is variable.
Pricing for VMs is complex with any vendor because of all the variables.
Users are charged by vCPU (virtual cores, versus physical CPU cores), GPUs (graphical process units that you can offload certain computation to), and memory size.
There are too many variations in prices to list them here, but they are all on the pricing calculator.
As with Amazon, with Google, you can commit to a fixed length contract to get a large price reduction. Or you can have dedicated, or what they call “preemptible instances,” whose storage is ephemeral, meaning it goes away when the server is stopped.
Unlike Amazon, Google does not have a spot market, which would let the user buy excess Google capacity at an online auction (with the proviso that the provider can shut that down at any time without advance notice). That is mainly used for volume and other testing or to surge capacity.
Memorystore for Redis
There are basic and standard tiers. Price is a function of tier, size, network speed, data center location, and elapsed time (not usage time). Prices range from $0.016 per GB per hour to $0.064.
Prices vary by size and architecture, meaning magnetic or solid state drives (SSD), and the speed of the I/O interface. SSD
Platinum level price is negotiated with the client.
Google Tensor Processing Unit (TPU) Pricing
TPU prices are high—ranging from $1.50 to $4.35 per TPU per hour.
Google says that customers can drive down expenses by running parts of the code on regular Compute instances and offloading the math-intensive code to TPUs.
GCP product and services offerings
The Google Cloud Platform includes more than 100 individual products—from AI and machine learning, to data analytics, to networking, storage, and security.
App Engine is a framework and platform for developing and hosting web apps that offers automatic scaling to answer increased demand. Instead of running apps in containers, you can create an application and run on an abstraction of it. This allows users to run an app without a VM or container.
Comparable to Amazon EC2.
BigQuery is a serverless, enterprise-level data warehouse. It is designed to help users set up their data warehouses quickly, so you can start analyzing and using the data, and it can analyze petabytes of data in minutes.
BigQuery uses a standard SQL dialect that is ANSI:2011 compliant in order to reduce the need for code rewrites. It also uses federated query, which means the platform can process external data sources without duplicating data.
Comparable to AWS Athena.
Build a True Data Lake with a Cloud Data Warehouse now.
Cloud Run lets you create containers without virtual machines. To see how that is useful, and can save money, consider how you would set up containers without Cloud Run or something like it. You would have to (A) spin up a virtual machine and then (B) spin up containers inside. So why do (B) when you only need (A)? This is inherently wasteful. Plus it runs counter to the whole idea of containers, which is a minimal operating system that does not need all the encryption software, file transfer software, a system log, etc., that a full-blown OS has.
But a container needs some place to run. Google provides that with Cloud Run.
Google Compute Engine delivers virtual machines. That is, it lets you pick from different operating systems and hardware sizes to create virtual machine instances. A virtual machine, of course, is a full-blown computer riding atop something called a hypervisor. It's really an abstraction of a computer, since it emulates a computer but does not have direct access to the screen or disk drive etc. Instead, the host operating system and hypervisor does that. This is how you take one individual computer hardware and divide it into multiple virtual computers.
VMs can be started and shut down as needed, so you can surge and contrast computing resources as the load on your application rises and falls. The end result is cost savings, since you are not paying for idle hardware.
Cloud Run is the same as Amazon EC2.
Memorystore for Redis
Cloud Memorystore for Redis is an open source, in-memory database, similar to SAP Hana or Apache Spark. The idea is that the database will run faster when there are no disk drives. This is because disk drivers have moving parts (e.g., the disk controller) that are not as fast as solid state storage (i.e., memory).
So Memorystore stores data in memory instead of on disk. But memory is expensive compared to disk storage, so it's not suitable for all tasks—unless their importance justifies their cost. For example, a typical medium sized VM has 8 to 32 GB of memory. That is not much compared to 1TB of data that you can add to your laptop for a mere $200 for an external drive.
Google Persistent Disk is block storage for VMs. It allows database blocks to be easily resized, backed up, and supported across multiple readers. It is also automatically encrypted, so users don’t have to worry about security for their cloud data.
You need a persistent disk, because when you shut down a virtual machine, the storage goes away. This is because the storage is the physical storage attached to the PC on which the VM runs.
Comparable to Amazon EBS.
GCP Support is available in various tiers. There is free, self-service support, and paid support.
For free support, Google refers you to StackOverflow, which Google engineers monitor. There are also Google Groups and Slack channels.
Paid support includes phone support and, optionally, a dedicated account manager for your account, depending on which tier you purchase.
Google Tensor Processing Unit (TPU)
Google Cloud TPU is an offering unique to Google, but not entirely since it is a proprietary form of GPUs (graphical processing units designed to handle large scale mathematics, which is important in machine learning).
TPUs (Tensor Processing Units) are based on technology invented by NVIDIA, who leads and even invented the market in graphics cards. The TPU is a graphics-card-like CPU, except it has hundreds or thousands or core instead of the normal 4 or 8 of regular CPUs.
TPUs use the same technology as the graphics card in your desktop or laptop to do very large scale mathematics.
Tensors are used to build neural networks to do voice and image recognition. Plus they provide the CPU power to solve massive problems, like figuring out the optimal shipping schedule for a worldwide shipping company or modelling computer networks and monitoring cybersecurity events.
You can do all of this with regular CPUs, but TPUs reduce the time to do that by an order of magnitude.
Google Cloud in a real-world business application
Travis Perkins is the UK’s leading construction materials supplier. The company operates more than 20 businesses and owns almost as many brands.
In 2017, Travis Perkins accelerated the migration of their data to the cloud using GCP. The company uses Google BigQuery as a cloud warehouse, and enjoys improved business intelligence from their big data due to built-in machine learning capabilities.
“Google BigQuery is helping us revolutionize our Business Intelligence, understand the data about our business, and drive improvements in our customer experience.” - David Todd, Group Data Director, Travis Perkins
With internal and external business intelligence data sources in an integrated, cloud data warehouse, the Travis Perkins team can analyze huge sets of unstructured data—from videos to social media posts—in order to uncover new correlations and patterns. Those kinds of unique insights drive competitive business advantages not otherwise available.
Google Cloud Platform certifications
Certification courses give your team a chance to thoroughly learn the platform, and certifications let engineers demonstrate their knowledge. Of course, there are plenty of excellent engineers who do not have any certifications, but it is a type of due diligence, good for audit compliance, marketing, SEC, and meeting privacy rules.
The names of available Google Certifications describe what they cover. Those are:
- Professional Cloud Architect
- Professional Data Engineer
- Professional Cloud Developer
- Professional Cloud Network Engineer
- Professional Cloud Security Engineer
- G Suite Certification
How to Build an End-to-End Cloud Data Integration Solution Using Talend and Google Cloud Platform now.
Before GCP: How is your data?
Consulting firms are available to help manage a cloud migration, but many companies manage a migration simply be letting their developers, architects, and administrators get early access to free trials and product tiers. The Google Cloud Platform is one increasingly popular option for keeping an organization competitive in the cloud.
A crucial step before any cloud migration—regardless of how much is migrating or which platform you are using—is making sure that the company’s data is integral, accurate, and secure. The benefit of a modern, cloud-navtive big data platform like Talend is the seamless integration with Google Cloud Platform.
See how Talend’s GCP integrations helped Travis Perkins accelerate their big data integrations and business intelligence insights:
So before you start migrating, unify your data environment, establish governed self-service, put machine learning to work, and more. Try Talend Data Fabric to get started.