What Advantages Google Cloud Has Over AWS

Date: Oct 31, 2022

Abstract: This article will focus on where GCP is superior to AWS and use cases where the author believes GCP may be a better choice.

network protocol stack

The networking stack is a highlight of GCP, using Google's global ultra-low latency internal network. GCP uses software-defined networking (SDN), codenamed Andromeda globally, that delivers incredible performance, especially for low-latency microservices and big data processing.

Virtual Private Clouds (VPCs) in GCP are global, you can define them in different regions if you choose, and the entire network is software-defined with a lot of flexibility. Your load balancer works at the edge location, providing global load balancing and automatic scaling.

With GCP, you can build a global infrastructure with Geo distributed data very easily. This is very difficult for other cloud providers.

developer experience

While I'm primarily an AWS user, as a developer I must admit I prefer working with GCP. Its cli is great, it's consistent, fast and easy to use. You can also easily access alpha and beta features.

GCP's console experience is probably the best of all cloud providers, especially with cloud shell, where you can get a terminal directly from your browser and securely connect to a virtual machine (VM) from your browser, while No need to set up any SSH keys, it's perfect!

GCP VMs are very fast to spin up, much faster than AWS, which makes horizontal scaling especially sensitive. Its pricing is fair, and you can customize the amount of CPU and RAM you need, which is very convenient! GCP allows almost all instance types to connect to GPUs. This can turn any standard or custom instance into a machine learning (ML) enabled VM.

With Cloud Identity, GCP's identity management works brilliantly. It integrates with G suite and provides single sign-on (SSO), so there is no need to use other cloud providers' very popular solutions like OneLogin.

Finally, most services offer emulators. This is great, I can test all the applications with my laptop at once without using any 3rd party tools or complicated integrations.

Pub/Sub (publish/subscribe)

AWS provides many services for messaging such as SQS, SNS, Kinesis, Event Bridge, Kafka, etc. whereas GCP only provides Pub/Sub. To be honest, you don't need anything else, it's a very good and cheap service for a variety of use cases from data streaming to microservices. It's a global service that scales to handle huge amounts of data, and it's very fast.

Pub/Sub is very easy to integrate and use, and it supports many clients and protocols. It also provides two modes for consumers: push and pull. Best of all, it's very cost-effective and completely serverless!

database

Google has a particular focus on data, and they are very good at managing and scaling big data, providing flexible solutions for each use case.

Especially they offer 3 solutions that I don't think other competitors have, and these are big data solutions. Companies, along with other cloud providers, are building data lakes to be more cost-effective by storing large amounts of data in cheap storage like S3. They use traditional frameworks like Spark on Electronic Medical Records (EMR) to process it and optimize it to be able to query it from S3 using formats like Parquet.

Maintaining a data lake is complex, especially when data changes frequently. This can become unmanageable and eventually the cost will rise. Wouldn't it be great if we could store big data in a scalable and cost-effective database? It will be much easier. GCP has some good options. While object storage has always been cheap, these 3 solutions can be used for big data as long as they're not too bulky.



Big Table

Big Table is a fully managed NoSQL database. It can be compared to AWS DynamoDB, but they are different. DynamoDB is a NoSQL that scales to handle millions of transactions, but can only store 400Kb per item, and its goal is not to handle big data.

Big Table, on the other hand, is a petabyte-scale database. It provides consistent sub-10ms latency, so it's very fast and reliable, and it's also easy to scale and cost-effective.

Big Query

BigQuery is GCP's gold product, and since it's such a big product, it's hard to explain what it is. It's defined as: A serverless, highly scalable, and cost-effective cloud data warehouse designed to help you make fast, informed decisions so you can easily transform your business.

The closest AWS products are Redshift and Redshift Spectrum. BigQuery is serverless and scalable to query large amounts of data, it has built-in ML and BI models for a variety of use cases. What I like about BigQuery is that you can do anything with it, you can store logs or billing information. It has higher latency than BigTable, but is also a bit cheaper.

As a data warehouse for BI, Redshift might be better, but for artificial intelligence (AI) and machine learning (ML), BigQuery is better.



Spanner

Cloud Spanner is a fully managed, scalable relational database service for regional and global application data. I don't think there is a similar database in other cloud providers. It's massive, but also totally relevant. It allows you to use regular SQL at scale with strong consistent transactions.

Do you remember the tradeoff between SQL and NoSQL? Now they're gone, you can use SQL and scale globally, but it's not cheap.



ML/AI

Google has the best machine learning platform. It provides tools for all types of users and use cases. The number of services is huge, from low-level virtual machines for deep learning to high-level APIs.

With SageMaker, AWS is slowly catching up and has gotten very close to GCP, but GCP still provides a newer and more accurate toolset. It provides virtual machines specifically for deep learning, better integration with Kubernetes and machine learning training, etc.

Kubernetes

There is not much to say about Kubernetes, GCP has advantages over other cloud providers. GCP is cheaper, newer, faster and easier to use than other cloud providers. GKE is probably the best cloud service in the world due to its flexibility and price advantage. It allows easy migration from on-premises to cloud. It's secure and easy to set up, offers great autoscaling, and is easy to monitor.

The best part is that GCP empowers Kubernetes and provides a friendly ecosystem to run almost any workload, from microservices or data flow to big data pipelines. The Kubernetes ecosystem is huge and all of these tools have been validated and tested in GCP.

AWS is more focused on serverless and GCP is focused on Kubernetes, both technologies are great.

cost

In general, GCP is cheaper than other cloud providers because it always depends on what service you use and how you use it. If you use Kubernetes, GCP is the clear winner in terms of cost efficiency.

It's also the clear winner when it comes to compute and storage costs. GCP provides a better way to subsidize long-term usage, and snap-up virtual machines are very cheap.

The price of a GKE cluster running on a Snapshot VM is hard to match.



Example

AWS is still the best cloud provider with more mature offerings and more services than GCP. It also has a huge user base and better support. If you are in doubt, use AWS. Amazon has done a great job of catching up with GCP's machine learning capabilities, and has also lowered the cost of some services. But I still think that GCP might be a better choice for some of the following use cases:

Machine learning, especially deep learning or when using Kubernetes.

Thanks to Pub/Sub and DataFlow for big data stream processing. Thanks to the network stack, GCP has lower latency and pipelines run faster and less expensively. For batch processing, both providers are equally good.

Distributed real-time systems. If your microservices require extremely low latency, Google SDN + pub/sub is a good solution. For example Go microservice + gRPC runs very fast. Also, Akka is great for GCP.

Kubernetes. This is the main advantage of GCP, and GKE is a great tool if you want to run portable infrastructure cost-effectively. For serverless, AWS might be a better choice.

Global big data database. If you don't want to use a data lake and want to store big data at scale, then Spanner or Big Table are amazing databases that can make your life easier.

In short, if you want to run fast, low-latency microservices on Kubernetes or you have a lot of data, consider using GCP.





The most important asset is the developers

It is highly recommended that you try out the service and develop a small proof-of-concept (POC) on both platforms to gain experience on both platforms. Both providers have a free tier. Don't just consider reports from consultants, you need to judge for yourself and try both platforms.

I personally like Kubernetes, it makes your code portable across platforms, making switching between them a lot easier.

If you are an AWS user, please read the platform overview first and then check the best practices. After that, read the guide for AWS professionals.

GCP is also very easy to secure and manage compared to AWS. Finally looking at all the services GCP has to offer, it's catching up fast.

We're at a pivotal moment in software development, so whatever platform you choose, it's going to be a great choice. Just remember what the most important asset is and invest in it: developers!