The big cloud-off: AWS vs Azure

This year there was a clear winner with Computing Delta respondents

The cloud infrastructure market

The cloud infrastructure market is consolidating around three US players: Amazon Web Services, Microsoft Azure and Google Cloud Platform. There are plenty of other cloud service providers, of course, but even the likes of IBM and China's giant Alibaba - a major player in East Asia - have a market share in the UK that could qualify them as ‘niche'. Meanwhile, the proportion of the IaaS/PaaS market captured by native European cloud services is tiny and declining even within Europe.

The big three

AWS is the largest, accounting for 33 per cent of the global cloud infrastructure service market, according to research by Synergy, compared with Azure's 21 per cent and Google's 10 per cent. However, while the overall pie continues to grow apace, AWS's share has remained pretty static. Not so Azure, whose share has grown 60 per cent in the last four years, largely at the expense of ‘others', that period having seen many firms including HPE and Fujitsu giving up the cloud infrastructure ghost.

Recent Computing Delta research among 180 UK IT leaders found that AWS and Azure dominated, as expected. But what sets the leaders apart from each other, and which factors influence IT leaders' choices?

AWS vs Azure

Each year, Computing Delta checks the take up and opinions of IT leaders on public cloud providers through a mix of online surveys and interviews. Here we present our comparision of AWS and Azure from reseach undertaken in December and January.

Market share

Despite their relative shares of the global market, Azure was far more likely to be taken into production than AWS among our respondents. Sixty-three per cent said they were using Microsoft's cloud in production versus 28 per cent for Amazon's service. The figures for those having trialled Azure and AWS were 68 per cent and 47 per cent, respectively.

Which of these IaaS/PaaS services have you taken into production?

N=180 UK IT leaders in organisations using IaaS/PaaS.

On the basis of these numbers, Azure is both the better known and the more likely to be adopted by those who have tried it, continuing a pattern we have seen for the past five years, during which time the preference for Azure has increased steadily. An important caveat here is that our sample was predominantly larger organisations, which are more likely to have an ongoing relationship with Microsoft. Also worth noting is Microsoft's strong presence in the UK.

"It's Microsoft and almost everything we do is Microsoft," said a software director in a mid-sized technology firm. "Costs are mostly reasonable, service has been moderately reliable - but setup is complex."

Other quotes from respondents who selected Azure included the following:

"MS Azure is better integrated into our development workflow"

"Azure performed better than AWS and is more aligned with existing staff skill sets"

"Azure has some must-haves such as Azure AD"

"Global coverage, mature, massive range of services, security and compliance, best for Microsoft workloads and migration of legacy infrastructure"

Having an existing relationship with AWS was not thought to be particularly advantageous in the way it might be with other vendors, as Amazon′s services tend to be hands-off without the consultancy wrapper. In addition, Amazon lacks the enterprise heritage of vendors such as Microsoft. That said, Microsoft will only usually negotiate with the largest customers.

What AWS provides is an extensive geographical coverage, 200-plus services and more in the way of configurability - provided you can work your way around the ecosystem. "It offers a wide range of services covering extensive regions; very powerful, mature and extensible," said a development manager at a software company.

"Massive provider, great support, great regional availability"

"Class leader in configurability"

"Our directors preferred AWS because of its reputation"

"Good for non AD stuff, like HPC"

AWS vs Azure: Global reach

Amazon and Microsoft define infrastructure metrics such as availability zones and geographic regions differently, so direct comparisons are not particularly helpful. Both are available in many parts of the industrialised world, but AWS is probably still the larger of the two.

In Europe, AWS has six availability zones (separate data centres within the same latency region) with two more on the way, while Azure has seven.

AWS vs Azure: Pricing

Comparing the prices of large cloud vendors' services is extremely difficult. Their pricing permutations and combinations are carefully optimised to maximise profits while keeping customers on the right side of walking away, and are full of loss leaders, special deals, opaque terms and hidden extras.

Focusing on virtual server list prices, in the UK, on-demand prices for a low-end Azure Bs-Series B1LS VM running Ubuntu Linux with 1 vCPU, 0.5 GiB RAM and 4GiB temporary storage costs £0.039 per hour. An equivalent Windows VM is £0.08/hour. Savings of around 42 per cent and 62 per cent, respectively are available on one-year and three-year reserved options (reserved instances or RIs). These prices go all the way up to £25.11 an hour for a M128M instance with 3,800 GiB RAM and 14 TB storage, although savings of 90 per cent are possible through spot pricing.

AWS EC2 instances start at $0.047/hour for the t4gnano with 0.5 GiB RAM and 2 vCPUs, rising to $28.712 for a p3.16xlarge with 488 GiB RAM, 64 vCPUs and 25 Gigabit bandwidth.

These list prices are broadly similar with the differences being very much in the detail, for example in the discounts available for reserved instances (up to 75 per cent for AWS), and the ability to exchange RIs for other services, in which AWS is more flexible than Azure.

Then there are the inevitable ‘extras' which can be a trap for the unwary.

"Default configurations often insecure; costs mount quickly," said one respondent of AWS, while a CIO at a university accused Microsoft of treating students as employees for licensing purposes, thus pushing up the running costs of its services.

Azure

"It can be difficult to calculate what a service will cost before actually using it"

"Azure is reasonably priced but I find AWS cheaper and more customisable"

"Azure is expensive and it's hard to migrate existing licences"

"Excellent service, wish it wasn't so expensive"

AWS

"Inexpensive, very wide range of services"

"Wide range of services, but high costs"

"It works, but very much on its own terms. Expensive and suffers from Amazon backlash"

"Reliable and massive scale but too many offerings to be clear on what provides best value"

With the increasing popularity of the multi-cloud model, a current battleground is around egress costs. Previously AWS was accused of levying high prices for removing data from its ecosystem, but last year it bowed to the inevitable and extended its free tier before egress charges, also known as ‘data transfers out to the internet', kick in. For many smaller users there is now less to choose between them on basic data transfer costs although AWS's bandwidth pricing in particular is complex to navigate. We heard more complaints about AWS's egress charges than for Azure, although those impressions may come from before the free tier extension.

Data egress charges

*From S3

With good reason, operatives who understand and can effectively manage cloud costs are highly in demand. It is easy for organisations to lose control of their spend with the big cloud providers.

AWS

"Incredible breadth of capabilities, lacks coherence and feeling of being 'joined up'. Egress and Az-Az bandwidth far too expensive"

"Expensive to extract data, not easy to move away to another provider"

"Poor speed of egress of data"

Azure

"We have price and lock-in concerns"

Free tier

Free tiers are the way cloud vendors lure individuals and small-time users, in the hope they will become big time paying users. Some free services are perpetual, others are limited to a year and restricted in their terms of use. Both AWS and Azure offer plenty of free services, with AWS's the more extensive.

AWS's Free Tier includes 750 hours of Linux and Windows EC2 t2.micro or t3.micro instances each month for one year. To stay within the free tier, users must only deploy EC2 micro instances. Other freebies include 5GiB of S3 cloud storage, 25 GiB storage in Amazon DynamoDB, one million requests to Lambda serverless functions, and more, up to 100 packages in fact.

The Azure free account also lasts for 12 months. It covers around 40 services, a $200 credit, and is available to all new customers who haven't previously had an Azure free account. Linux and Windows B1 VMS are free for up to 750 hours per month, there's 5 GiB of locally redundant storage (LRS) file storage, 250 GiB S0 instance with 10 database transaction units… and so on.

AWS vs Azure: Support

Cloud hyperscalers have tended to take a hands-off approach, relying heavily on MSPs and resellers, self-service portals community support with both providing well-used developer and support forums. Neither scored highly for the price of support, or for the ease of obtaining direct human support which only comes with the higher end paid-for plans. However, unlike in previous years, both AWS and Azure received some praise for their support.

"Amazing 24/7 human support, said an IT manager in the manufacturing sector.

Amazon's AWS premium support starts at $29.00 per month up to many thousands, or as a percentage (5%-10%) of monthly AWS charges. By coincidence, Azure premium support also starts at $29 per month for Developer, rising to $1,000 for Professional Direct.

Which of the following terms describe your experience with your IaaS/PaaS?

Orange = AWS, Blue = Azure

Our cloud services provider is/has …

Orange = AWS, Blue = Azure

Neither Azure nor AWS scored well on factors such as lock-in concerns, specialising in the customers' sectors or locale, or the quality of local support, although Azure was slightly ahead in all but lock-in.

AWS vs Azure: Complexity

Both services were criticised for being complex to use, operate and manage. AWS has more services and is less integrated than Azure, which can mean a steep learning curve for beginners, but Azure's licensing and pricing is particularly opaque.

The as-a-service model moves costs into Opex, which smooths the cost curve, but only in the absence of hidden costs and unexpected spikes, such as when invisible thresholds are crossed. We heard more criticism of Azure than AWS in this regard.

The issue of complexity is an important one, because it means that mastering the services requires skills and experience. When similar services have different names and different ways of working between platforms, this is another way that organisations can find themselves locked in. Cross-cloud skills are rare and expensive.

AWS

"Early platform so got a head start; reliable and scalable; costs mostly reasonable; setup is complex"

"Not that easy for a new user to understand"

"It's complicated, especially around managing costs"

"AWS can be cheaper, but it's often much harder to use and manage"

Azure

"It's difficult to control costs"

"Easily deployed, but there's a lack of control"

"Very complex - but well documented"

"Expensive and complex"

"Can be complex to manage billing, some services are expensive"

"Pricing can be unclear and difficult to estimate"

"Continually improving means dealing with a high rate of change. Infrastructure has a few oddities."

"They keep moving the goal posts!"

AWS vs Azure: Roadmap

AWS is constantly rolling out new services and expanding into new areas of the globe. However, its piecemeal advances mean that it scored lower than Azure in regard to understanding of its roadmap, possibly just because Microsoft's gameplan of ‘integrate everything' is easier to follow.

Many of the recent announcements on the AWS 'what's new' page concern the Elastic Kubernetes Service (EKS), showing that cloud native is an area of focus. Some respondents felt that AWS was a little behind others, particularly Google, on AI/ML, indeed its last announcement in this area was 2020. Among announcements at AWS Reinvent in December were Private 5G, Sustainability goals, Control Tower to manage regions, IP address manager, EBS snapshots to reduce storage costs, and Kinesis data streams on demand.

Azure announcement's from Microsoft's Ignite event in September included expanded security features in Azure AD, updates to various AI services, industry specific features for Azure Synapse Analytics, developer tool improvements and chaos engineering tools, AKS (Kubernetes) solutions for Oracle WebLogic Server and IBM WebSphere Liberty and updated migration tools.

Conclusion

The two leading public cloud services are actually quite different. AWS continues to expand outwards, adding services and locations without seemingly that much focus on integrating them together, whereas integration is very much part of the Microsoft plan, and it can be hard to see where IaaS/PaaS stops and SaaS starts. Indeed, the distinction between the types of cloud grows more blurry every year.

Historically, AWS has been favoured by developers, and may still be best for many open source projects and cloud native, although licensing rows may have tarnished its image somewhat, as have news stories about antitrust, recent downtime, treatment of workers and the antics of its CEO - all of which were mentioned by our respondents. It has also been famed for innovation, but signs are the other clouds may have caught up here, especially around AI.

AWS was praised for its scalability, the range of its offerings and its availability. It was generally thought to be reasonably priced, and the support is fine, so long as you are willing to pay for it. Among many of our respondents it was still the go-to cloud, but maybe not the only one anymore.

As already mentioned, our audience, made up of UK Computing readers, may not be typical of global cloud users everywhere, but consistently, for five years, more have said they use Azure than AWS, and that gap is widening.

Some respondents had gone multi-cloud and were using AWS for development or HPC and Azure for enterprise, but not the other way around. Increasingly, Azure is seen as the 'enterprise cloud'.

Azure's integration with the existing Microsoft stack, including identity management through AD, was far and away its key asset and one that AWS cannot hope to emulate. This addresses, one of the big cloud problems, skills, and also the strategy adopted by many to run a hybrid cloud.

Microsoft is seemingly moving in every direction at once including AI, DevOps, cloud native and is stealing some of AWS's ground from beneath its feet, all while somehow avoiding the bad press around anti-trust, tax and privacy despite being just as guilty as AWS or Google in many cases.

We were genuinely surprised to note that, this year, Azure was the clear winner.

Computing's Deskflix: Future of Hybrid and Multi-Cloud event is happening on February 22nd. Register today!