by John Edwards

7 steps to a well-architected cloud

Mar 20, 20199 mins
Cloud ComputingEnterprise ArchitectureIT Strategy

Efficient cloud architectures don't emerge out of thin air. The best environments are constructed piece-by-piece from the ground up.

cloud comput connect hand
Credit: Getty Images

Creating a strong and reliable cloud architecture is essential for long-term IT and enterprise success. Unfortunately, many cloud architectures are constructed haphazardly over several years, failing to meet the demands imposed by rapidly evolving technologies and business requirements.

If you’re building your first cloud architecture, or if your organization’s existing cloud structure is beginning to show signs of age and/or poor planning, the following 7 tips can help you get it on the fast track to productivity and efficiency.

[ Stay on budget with these 6 cloud cost management tips, learn the 5 fundamentals of effective cloud management and beware hidden cloud migration gotchas. | Get the latest cloud computing insights by signing up for our newsletter. ]

1. Know what you’re doing

A cloud architecture is a holistic structure that depends on symbiotic relationships formed between individual components, including front- and back-end platforms, a cloud-based delivery system and network resources, notes Michael Tavares, a research analyst at tech market analysis firm ABI Research. “If each of these components is lacking, we do not have a cloud architecture,” he says. “More importantly, a change in one component will result in changes in other components.”

You should never wing it, warns Thomas Boyles, director of platform services for Sauce Labs, a web and mobile applications testing cloud provider. Building a cloud infrastructure is a complex task, and you can’t account for scalability and durability after the fact. “Don’t make the mistake of just piecing it together and expecting it to work,” Boyles says. “You need to be extremely detailed and prescriptive about the design.”

Cloud leaders such as AWS, Google and Microsoft provide building blocks under the assumption that adopters will do their homework and read the user manual. “They’ve given you Legos, but make sure you use them the right way,” says Ennio Carboni, product manager of VMware’s CloudHealth cloud optimization platform. “Otherwise, you run the risk of unchecked cost spikes and environmental destabilization.”

The best way to achieve a well-architected cloud solution is through planning and foresight, advises Kyle Brown, an IBM Fellow and cloud architecture CTO for Cloud Labs. “There is no magic bullet,” he states. “Teams should define the architecture and plan incrementally so they can make iterative changes.”

2. Focus on application owners’ needs

Cloud services should take advantage of cloud-native capabilities, says Sekou Page, cloud architecture solutions lead at professional services and business consulting firm Capgemini. “Maximize the potential of the cloud on behalf to the customer to enhance performance, increase reliability and minimize cost.”

“There must be a radical focus on the needs of application owners — both developers and those responsible for sourcing and/or operating external software from ISVs,” declares Chip Childers, CTO of the Cloud Foundry Foundation, a nonprofit organization dedicated to advancing an open source, multi-cloud application platform. Childers notes that a large part of AWS’s success has been driven through a focus on customer needs. “The same should apply to any cloud provider, public or private,” he says.

Tavares believes that the entire enterprise should be surveyed before making any crucial decisions regarding storage, scalability, security and other services. “End users need to understand where and why cloud infrastructure is best managed and structured across an organization,” he says. “Only then can they avoid cloud sprawl — an eventuality when a company adopts disconnected systems for different divisions, such as HR, finance, operations, marketing and PR.”

Enterprise and IT culture matters, too. “You have to have the right level of buy-in and investment across the organization and across the user group for whom you’re deploying your cloud,” Boyles explains. A certain amount of in-house expertise is also necessary, potentially including cloud architects. “You need people who have done it before and, more importantly, who understand the use case and user persona,” he advises. “Understanding the needs of the user group is the key to good planning and design.”

“You can’t do it on your own,” adds Justin Stone, senior director of secure DevOps platforms, for global insurance firm Liberty Mutual. “A cloud journey requires multiple teams across an IT organization and a strong business connection,” he says.

3. Seek scalability and standardization

Cloud infrastructure components should be able to scale independently along three key dimensions: storage, compute (both memory and processing power) and network. “No matter the specific choices made, you must prepare the platform to scale in standardized ways across the critical capacity planning dimensions that matter to your users,” Childers recommends.

An organization’s cloud architecture should not be dependent on any single cloud service provider, warns Prof. William Rials of Tulane University’s School of Professional Advancement’s Applied Computing and Technology Program. “A well-defined cloud architecture should focus on open standards as much as possible to achieve maximum compatibility between different cloud service providers and the on-premises infrastructure.”

The rise of microservices is fueled in no small part by the rapid adoption of standardized software containers. “That technology allows developers to package their code into lightweight, platform-agnostic packages that can be easily moved across different infrastructure,” Tavares says. “Given the rise of the multi-cloud market, containers are an essential aspect of cloud infrastructure which allow for the individualized and coordinated development of apps across different vendors’ software.”

4. Innovate, don’t imitate

IT should never rely on legacy ways of thinking. “This includes the way they operate, secure, monitor, deploy and so on,” Page notes.

The temptation to “lift and shift” an existing on-premises architecture into the cloud, while hard to resist thanks to its immediate time and cost benefits, is almost always the wrong approach to building a cloud architecture. “Perhaps the biggest ‘no-no’ is to think of the cloud as no more than a ‘data center in the sky’ to which you just move applications … and declare victory,” says Miha Kralj, managing director of cloud strategy, architecture and delivery for business and tech consulting firm Accenture. “CIOs that lift and shift their applications with no roadmap for the future often find they are not getting the value they expected and, in some cases, fall into a technical debt where they are paying far more than they expected.”

Cloud innovation requires an IT team that’s willing to embrace new approaches and technologies. “Ensuring the right mix of skills is critical,” says Roland Barcia, an IBM distinguished engineer at IBM Cloud Labs. It’s wrong to blindly assume that a team that was originally assembled to build traditional data center solutions will be able to function as effectively when developing cloud-native solutions, he notes. “Instead, it will be important to retrain ops teams to be more engineer-like and develop more software-defined skills.”

External support is also valuable. “Partnering with cloud providers and software partners that understand your business and your goals is key to getting started,” Stone observes.

5. Vary your cloud storage diet

Focusing on one storage type is not an option, Tavares declares. Most major cloud vendors offer a variety of storage services. AWS, for example, provides Simple Storage (S3), Elastic Block Storage (EBS), an Elastic File System (EFS), an import/export large volume data transfer service and a Glacier archive backup and storage gateway. “No single storage option fits all situations,” he says. “Moreover, leveraging different cloud storage options for different datasets can bring performance, cost and functional benefits.”

6. Build resilient connections

A well-architected cloud supplies high bandwidth and low latency. “The network layer needs to be agile and provide end users the ability to move quickly and efficiently between servers and, in some instances, other clouds,” explains Mike Attar, CIO at IT service management company Future Tech Enterprise. “Monitoring and administrative functions should be easy in order to facilitate decisions on how to utilize the cloud resources accordingly.”

“A well-defined cloud architecture involves not only the cloud infrastructure but connectivity as well,” Rials notes. By default, organizations connect to cloud service providers via the internet. “Every major cloud service provider offers a private connection from your location to their cloud infrastructure,” he says.

Since a well-defined cloud architecture involves a multi-cloud approach, having separate connections to each cloud service provider may be cost prohibitive. “So it is recommended to establish network connectivity to a cloud connectively broker or cloud exchange that has network interjections to every major cloud service provider,” Rials advises.

7. Integrate security throughout

Cloud security services should never be siloed from other services. “Rather, they should be integrated into additional services, such as storage and those pertaining to improved scalability,” Tavares says.

Sound security principles should be embedded in every layer of the cloud architecture. “Organizations using cloud services should have a clear understanding of where the cloud service provider’s responsibility ends and the organization’s responsibility starts,” Rials says. Many new cloud adopters make the incorrect assumption that cybersecurity is the complete responsibility of the cloud service provider. “A well-defined cloud architecture has cybersecurity ingrained throughout the entire architecture,” he observes.

More on cloud computing