Why does the interface matter? Cloud provider consoles compared

Originally written for InfoWorld.

The big draw of cloud services is the fact that all resources are consumable via APIs. Providing a consistent interface that developers can access is the first step towards offloading the “undifferentiated heavy lifting” to a cloud provider. Why build parts of your application infrastructure when you can make use of a service built, maintained and continually updated by experts?

The challenges of providing a good API center around the design of the API methods themselves, having good quality documentation, code samples and libraries for major languages. What becomes more difficult is when you have to complement the programmatic API with a web and/or mobile UI.

Every cloud provider offers a web based console which allows you to manage all aspects of your cloud account. In the early days, this console could be very simple – just a few products, billing and user management was all that was necessary. But as the number of products has expanded and the complexity of those products has increased, having a single console to manage it all has become a major area of focus.

What are the challenges? #

The main thing is product range. AWS has 18 top level categories of products, ranging from Compute to Storage and from AI to Application Services. Each of these has up to 15 products within them, not to mention the AWS Marketplace, support and billing systems. This vast number of top level services all have sub services and features, have interconnected components and all are in use to different degrees by different customers. Google Cloud and Microsoft Azure have similar challenges with the range of products they offer.

Providing a full set of infrastructure services is complex, and building an interface to manage them has a number of things it must address:

• Discoverability. How do you categorize the products and ensure that customers know what they are called, especially with the prevalence of acronyms like ELB, GCE, CloudML, VPN and “code names” like StorSimple, HDInsight, Aurora and Firebase.
• Navigation. With potentially hundreds of different services and varying levels of hierarchy, a simple tab interface or range of icons is going to quickly break down. Search is going to be crucial, as is being able to quickly access recent services.
• Geographies. Different services have different regional availability, pricing and features. How do you make it clear which context you’re operating in and how does that work when services operate on a global, regional and zone level
• Consistency. With such complexity inherent in the product portfolio, having a consistent design language across the UI is important to reduce the overhead of actually using the interface itself. Different teams will likely be responsible for their own products, but having them all within a single console means you should not have different layouts, button styles, fonts, images, etc.

How do the cloud providers stack up? #

AWS #

The AWS console has had a very similar design since it was initially launched. Minor design tweaks and the addition of search show gradual improvements but it is still very basic in design. AWS is known for its complexity and the UI doesn’t try to hide that.

You can switch between accounts and regions, although the currently selected option doesn’t particularly stand out. There’s no real way to search resources globally across the account, which can make it difficult to find what you might be looking for, but you can pin services to the top bar to make them easy to access.

The UI is very consistent, something which has improved over the years e.g. billing and support moving away from the old Amazon.com style into AWS proper.

Amazon has a reputation for being efficient and frugal, words which you could use to describe the AWS interface as well. It does the job with a minimalistic design, which is probably perfect for a technical audience.

Google Cloud Platform #

The GCP console has changed radically several times over the past few years as their cloud portfolio has expanded. GCP works on the basis of projects, which were originally entirely self contained but recently gained the ability to be linked under custom organisational and billing hierarchies. GCP also doesn’t use geographies in the same way as AWS. Instead, specific resources have zone or regional properties but everything is managed together so that you have a global view of the infrastructure. This makes it easier to get a top down view of all environments but is quite different if you’re used to them being logically separate like on AWS.

As you might expect from Google, search is presented as a central part of the web console. It is not limited just to products – you can search many of your resources as well e.g. for compute instance names, even across projects.

The design language is becoming consistent across Google products but completing this is clearly a big project! Even within GCP, you can still see older products like BigQuery using different UI designs, or acquired products like Stack Driver not really fitting in properly. The support ticket interface is also poorly implemented compared to the main GCP console. Rearchitecting and merging legacy components is difficult, even for Google!

From a design perspective, the GCP console is very nice to use and builds in a lot of functionality for ease of use e.g. menu pinning as well as for power users e.g. keyboard shortcuts and search. Google are iterating quickly to make improvements but there’s still some work to do with older areas of the platform.

Microsoft Azure #

The new Azure portal is a major improvement on the old portal, but for me continues to be one of the worst interfaces I’ve had to use. Even the login authentication is confusing with the many different login screens and URLs that Microsoft seems to adopt for its online services.

Whilst the Azure portal design style is consistent across many of the other Microsoft services, the model of navigating horizontally through a product hierarchy is poorly implemented, buggy and really breaks down when trying to use a small screen on a laptop or tablet. You can see the similarities between the Azure interface and Windows itself, which probably means that Windows users will be more comfortable and familiar with how it works.

With so many products and options, it is very difficult to find what you’re looking for, especially when the text labels disappear and are replaced with a tiny icon that you must memorise! At least the older UI has been retired now, because I remember that billing was still in the older interface until recently, which made figuring out how to perform simple actions like downloading an invoice very confusing.

Why does the interface matter? #

With more and more resources being deployed to the cloud, being able to manage them is already a critical part of many businesses.

Cloud providers risk losing a competitive bid because their products are too complex to use. Unlike in the old world of enterprise software, the end user is now more often than not the buyer and decision maker. Developers will pick the provider with the best management interface, which includes the web UI as well as the API. With design a central part of consumer products, it’s not unreasonable to expect good quality design as part of business products too.

As an example of this, I was recently making a payroll tax payment through the Silicon Valley Bank online banking interface. Their UI is the worst I have ever used – it is confusing, has totally different designs depending on which section you are in and has poorly implemented workflows which drop you into different views repeatedly and unnecessarily. This resulted in a payment not being made on time and interest charges being levied, all because the UI wasn’t clear about the status of a payment pending approval. Banking is the ultimate legacy industry and it shows the opportunity for challenger banks. I am an investor in the UK consumer banking startup, Monzo, and the contrast in user experience is stark.

When critical resources are being managed through a third party interface, you want to ensure that interface provides accurate information that is easy to understand, with design that make it easy to manage and avoid mistakes. This is a principle that we take seriously at my company, Server Density, when presenting critical monitoring data through alerts and graphing. You don’t want to be bogged down in trying to navigate a poorly designed UI in an emergency. Making a mistake with online banking has financial consequences just as much as making a mistake with a cloud provider interface could result in downtime, deleted data and loss of revenue.

The interface matters.