Cloud: Choosing the right platform for your ambitions
As AWS and Azure scoop up more of the cloud IaaS /PaaS market, what are organisations looking for? A research presentation from Computing Delta
What do organisations look for when choosing a cloud infrastructure or platform provider? Is it all about price, compliance, roadmap, security credentials, cutting-edge capabilities or other factors? Most likely a combination including many different requirements, overlaid by what is already in place, internal skills and relationships with the vendor and its resellers on the ground.
This article looks at some of those factors, referencing recent research by Computing Delta, and considers movements in the market for Infrastructure and Platform as a Service (IaaS/ PaaS).
Direction of travel: upwards
Cloud is going only one way: upwards. Analysts such as Gartner, Forrester and IDC are united in the opinion that cloud IaaS and PaaS will keep on growing at double-digit rates for the next three years, even if they disagree on the numbers.
As it is growing the market is also consolidating. Smaller players being bought out by big ones, with others being forced into specialist niches, and the cloudwashers of the past, those MSPs that simply rebranded existing hosting services with the trendy ‘cloud' label, are no longer actively pushing those services.
More so than ever, cloud IaaS/PaaS is a market that's dominated by two or three huge players, and those big players are engaged in a price war. Prices of VMs, storage and compute capacity continue to fall year on year, so consumers of these services are quids in. But of course, it's not just about list prices, or even prices in general. There are many other factors to consider.
The cloud market
So, what does the market for IaaS and PaaS currently look like? This is actually quite a difficult thing to work out. Vendors report financial results in different ways with some bundling in SaaS or on premises licensing, and analysts accord different weightings to the various data points. Which is to say, it's not an exact science.
Figures drawn from research including Wikibon, Synergy Research, Deloitte and others
That caveat aside, all agree that AWS is the biggest, commanding something like a third of the entire market. Some analysts put its share of the pie as high as 45 per cent. Next comes Microsoft Azure which is less than half the size of AWS, by most counts, but growing faster.
After that it gets a bit murkier. IBM used to be third, but Big Blue is in fifth place on our chart having been overtaken by fast-growing Google and Alibaba. Perhaps this year's purchase of Red Hat will redress this trend, helping it to plug into the still-growing appetite for hybrid cloud in big customers such as banks and governments, but we have Google Cloud Platform in third followed by Alibaba. China's cloud giant is much more prominent in East Asia, but we haven't seen so much of it In EMEA and the Americas so far, and its prospects won't have been helped by Trump's trade war.
Then we have Oracle, SAP, Salesforce and Workday, plus 'other', which includes Rackspace, HPE, VMware Cloud, Digital Ocean and many other vendors.
The biggest takeaway is we have half the market controlled by two vendors: Amazon and Microsoft. And that's a conservative estimate. Gartner puts the market share of the duo at 63 per cent, with that number edging upwards every year.
As part of Computing Delta's recent research into cloud we polled 200 IT decision makers, mainly in the UK. The first thing to note is the leadership position is reversed. We've found consistently over the years that Azure is more widely used than AWS, but there's always a question mark as to where the line between products like Office 365 and Dynamics 365 and Azure are drawn.
VMware cloud, which comes fourth, is another interesting one. VMware public cloud actually runs on AWS, but are public and private clouds and indeed virtualised servers being conflated here? Quite possibly.
However, the combined dominance of AWS and Azure is even more apparent here.
Nevertheless, the cloud giants are not suitable for every business or use case. Google Cloud Platform (GCP) and IBM Cloud offer their own specialisms, particularly in areas such as AI, microservices and containers. Traditional SaaS vendors such as Workday and Salesforce offering up their platforms as the foundation for new applications that offer easy integration into those CRM and HCM applications, and Oracle and SAP customers may well appreciate the stepwise move to the cloud offered by those vendors.
What counts in cloud?
This figure shows the priorities accorded to various factors by our 200 IT decision makers who were asked to select five out of 15 items. These are factors they use when deciding which cloud vendor to adopt.
Prices are certainly up there, in the shape of running costs, but so are characteristics such as uptime and availability and security credentials.
Fourth on the list is appropriate local UK / EU focus. This is important not just because legislation is localised but also because a one-size-fits-all is rarely going to be the best approach for an individual firm, and businesses require support that understands the unique needs of their market. The global hyperscalers sometimes fall down here.
In fifth place we have breadth of services, which is where the hyperscalers like Azure and AWS have an obvious advantage. Chances are if they don't offer some service this week, next week they will.
The top bases (availability, security, pricing) are well covered by the big players, although local support can be patchy, so the main battleground is here, in the long tail.
A clearly written SLA was considered of top-five importance by 28 per cent of our respondents.
On the right of the chart above table come the specialisations and customisations, which are not so crucial for the majority of customers but are nevertheless vital for an important few, such as legal services, finance, scientific research, health and others.
Likewise with cutting edge innovation. Many will be happy in the middle of the road, but machine learning, distributed container platforms and cloud-native are essential for some.
The quality of local support can be an important differentiator. Cloud providers scale by automating and commoditising services as far as they possibly can. The last thing they want to do is answer the phone for a support call. A common criticism of the big cloud providers is that it can be very hard to get through to a human being when you really need to talk to someone.
It's when you need that extra support, or to add in security features, or to bring in additional services you didn't know you needed, that costs start to stack up.
Cloud costs add up
The way that cloud costs soon start to rise was mentioned frequently. Initial price tags can be deceptive.
Unlike with most on-premises licensed software, basic cloud services come with a list price, but these are like the discounted milk, the loss leaders that supermarkets use to get you through the door. Managing cloud services is complex but those prices can make it seem simple.
"Default configurations are often insecure; costs mount quickly," said one respondent of AWS.
Licensing is another area where unexpected costs emerge. A CIO at a university complained that Microsoft treated students as employees for licensing purposes bumping up costs quite significantly because the service was leased on a per-employee basis.
Other costs can arise when scaling up, adding new services that weren't considered necessary at the time - or when forgetting to turn off VMs and storage before going on holiday for a fortnight.
Cloud strategies
Turning now to overall cloud strategies, the largest number said hybrid cloud played a part in their plans. One third had a cloud-first strategy, meaning that new projects go into the public cloud unless there is a good reason for them not to. Meanwhile, nine per cent said they were cloud-only, dispensing with their own hardware completely.
Then we have the matter of multiple clouds versus a single cloud vendor. Interestingly these numbers were pretty similar. Twenty-seven per cent had a multi-cloud strategy, a best of breed approach designed to optimise the use of cloud services and reduce the risk of lock-in. However, a slightly larger number was willing to trade these advantages for the simplicity of having just one vendor to deal with.
Multi-cloud is a hot topic at the moment. It's easy to forget that cloud is still relatively new and there aren't too many companies who have taken the step of migrating large workloads from one cloud vendor to another. Suffice to say, it's likely to be a painful process, particularly with PaaS, and one that most will wish to avoid. However, integrating between vendors' cloud platforms can also be challenging. Different clouds use different languages, databases and toolsets, then there's the problem of integrations and optimisations. Certainly it's not a project to be undertaken lightly.
But interoperability is now becoming a differentiator between clouds. Last month CRN Australia published an article entitled: AWS forbids partners even mentioning multi-cloud. If this article is accurate, one way AWS is looking to retain its lead is by what might be called a monotheistic approach: I am the first and last, there is no other cloud but me.
Acording to this article, AWS has effectively banned resellers from talking about "multi-cloud," "cross cloud," "any cloud," "every cloud," "or any other language that implies designing or supporting more than one cloud provider" in its co-branded literature.
Naturally, Google lost no time in jumping on that article. Customer operations president Robert Enslin insisted on Twitter that, unlike AWS, Google embraces openness and choice - taking the opportunity to promote GCP's new Anthos hybrid cloud platform at the same time.
Doubtless Google, or for that matter Microsoft, would do the same thing if they were in Amazon's position, but for now they are enjoying a rare moment occupying the moral high ground and allowing their sharing, caring sides to show.
Microsoft and Oracle recently unveiled a cloud interoperability partnership, in which services running on Azure will now be able to connect seamlessly with Oracle's databases in the cloud. Google has a similar deal with SAP. And last year Microsoft and Google, together with Facebook and Twitter, started a data portability scheme called the Data Transfer Project, designed to make transferring data between cloud platforms easier. Last month they were joined in this endeavor by Apple.
In theory, then, the decision over which CSP to go with should be much less fraught in the future, but these things take time. For a number of reasons, operating across clouds is still not an easy proposition. Because of this selecting a CSP wisely is vital, since it's likely to be a wide ranging and long-term relationship that's going to grow over time.
Security concerns decreasing
To finish off, a quick word on security. Once the number one barrier to the cloud, it appears that concerns are easing off a little. One-third of respondents said they were less worried than they had been in the past.
"The core platform vendors are very well qualified in security controls, and initial scaremongering over security concerns has not come to pass," said one.
On the other hand, 15 per cent were more worried, their concerns summed up by this respondent. "The number of applications moving to the cloud has increased. The number of attacks targeting cloud infrastructure has increased correspondingly. Cloud security is in its infancy and harder to manage."
The full results of this research are available on Delta, with in depth comparisons of the vendors.
Author spotlight
John Leonard
More from John Leonard
Ofcom fines TikTok £1.9m for failure to provide child safety information