AWS, Azure, Cloud, GCP

Cloud comparison – intro

Under the series of articles I’m going to discuss the differences between main cloud compatitors:

  • Alibaba
  • AWS
  • Azure
  • Google (GCP)
  • IBM Cloud
  • Oracle

This list is based on Gartner’s magic quadrant (2019) and the order is alphabetical. Looking into Gartner’s data, the leader is AWS and just few steps behind is Azure. Next is GCP. All others are much behind. However from that what I saw, the progress done by IBM and Oracle in cloud services and its offering is significant.

Each service provider has own specific offering and also pricing. Some services are cheaper in one option other in second one so it is not possible to say “this provider is cheaper” or “this one is more expensive”. That will be the true in some context (set of services).

The set of services varies depending on provider. Our choice should be based on the needs and potential growth of the business. Nowadays enterprise are more often building hybrid cloud solutions (on-premise + 1 cloud in single solution which does not mean that they do not use multiple clouds). However the future is multicloud – one solution deployed to multiple clouds.

Free accounts

Each providers gives some credits for defined period of time to get familiar with service. And after a period of time (usually 30 days) an account remains with free access to some services. I won’t analyze here what is in scope of free tier – that would be really long post only on that.

Registration

Most of forms is easy to follow and typical personal data are required. Also payment method (credit card) is requested. However each provider claims that the money will not be taken until you will start using paid services. So far I did not pay for anything in trial period.

The only problematic one is AWS – it requires to give a lot of details including problematic 2 step verification (with automated phone call to give you a code). When I was registring I had to request support call as I was not receiving the data or when I received I got the information that they are invalid (multiple tries, my colleagues were affected as well).

Credits

Each provider offers some credits for defined period of time to test the services. They not always cover all the services so you need to check what is in them before you will start using (most of popular services is in, it is rather about some very specific domains e.g. some AI service might be excluded).

ProviderCredit periodCredit amount
Alibaba30 days – 1 year, up to service450-1300 USD (higher for enterprise only)
AWS12 monthsselected services only
Azure30 days200 USD (170 EUR)
Google1 year300 USD
IBM30 days200 USD (only for pay-as-you-go selected)
Oracle30 days300 USD

Please note that above table is valid in the day of publication. It can be changed at any moment without prior warning or notification.

As you can see from above summary, the most permissive in terms of testing is Google.

Regions

The concept of region is pretty simple – it is multiple data centers composed into availability zones (described later) in one physical region. A physical distance between regions is usually significant to continue providing services to customers in case of distaster (e.g. fire, long lack of electricity, earthquake, etc.).

Usualy one region is composed by multiple availability zones (typically 2 or 3). Oracle and Alibaba are exception – they have one or more.

Number and distribution of regions is very important for enterprise use as some of countries have very strict rules applied to the data (e.g. USA, UK, Switzerland, Germany or China).

Provider Number of regionsRegions in China
Alibaba21 existing9
AWS24 existing2
Azure54 existing, 4 in plans4
Google22 existing0
IBM6 existing0
Oracle21 existing, 14 in plans0

As you noticed from this table, most of cloud suppliers has many regions and data centers are spread across the world. Three of them has no region behind the Great Firewall. Alibaba is exception as they origin is China.

Oracle has the biggest growth plan and Azure has the most of regions with new surprising comming soon (e.g. Mexico, Israel, Spain).

Availability zones

Some of services for all providers are “global” and by design high available so deployed in several availability zones in region.

Alibaba, AWS, Google, IBM and Oracle supports availability zones in all regions (if more than 1 exists). Azure does not support availability zones for all regions, only in selected.

In addition AWS has Local Zone concept – a region extension in different location with high speed connection to the region. As new feature, it is available right now only in us-west-2 region and you have to opt in for it. Also it is applicable only for few services (VPC subnet, EC2, EBS, FSx and Application Load Balancer).

Edge locations

Edge location (called also point of presence) is a caching output of the cloud provider (usualy it is equal to CDN). Sometimes it is something more (e.g. DNS, DDoS protection etc).

The number of edge locations is much higher than availability zones (hundreds or even thousands).

Important missing features

Each cloud provider made a huge progress since I meet them for the first time (e.g. a year ago GCP has not supported Microsft SQL Server as database service – right now they do). Most of use cases is covered by virtual networks, virtual machines, SQL databases as a service, functions, queues, containers.

Main feature from my perspective which is missing at IBM’s, Google’s and Oracle’s cloud is support for container services. Also not all providers support VR and mixed reality but these topics are out of scopes of this series of posts.

Summary

Each option has benefits. We need to really carefully consider each option depending on the needs and costs. As pricing is dynamic and region dependent, I decided to skip it – generic comparison would not be fair. In next posts, I will go deeper into particular services. Stay tunned!

Leave a Reply

Your email address will not be published.