A web search on cloud adoption rates generates a dazzling array of charts and stats that for years have been telling a familiar story: some links simply aggregate numbers to say the same. And while there is some variation on the theme, i.e. relatively more enterprises are now using public cloud than private cloud infrastructure, by 2018 the share of total cloud workloads that will be SaaS will increase, while the share that are IaaS will decrease, and by 2015, end user spending on public cloud services could be reach more than $180 billion on a global basis, the general consensus is that cloud deployment is growing more quickly than traditional IT, resulting in an expectation that hybrid IT models will dominate for some time to come. How quickly? According to InsightaaS forecasting, by 2020, spending on cloud and traditional IT will be roughly equivalent in the Canadian marketplace, creating the ideal hybrid scenario.
Indeed, new methods of data collection and analysis have pushed the dark arts into a realm of sophistication and complexity that was unimagined in market forecasting of years past. No matter if the numbers tell a different tale, in fact their variability is evidence of a hearty research industry that feeds on change and flourishes on differentiation. For the vendor community, this wealth of information provides endless fuel for strategic market campaigning, and even validation for current market positioning. But for the end user organization, the value of market slicing and dicing is less clear – except as a reminder that the cloud train has left and it’s time to get on board. But herein lies the challenge: what steps can the end user organization take to ensure the success of cloud deployment in the murky world of hybrid IT, which is likely the most complex in terms of management?
To understand more about best practices in cloud deployment, InsightaaS spoke recently with CenturyLink CTO Aamir Hussain about two of the key adoption questions: how does the user determine the most appropriate IT deployment model for the business and its workloads, and what techniques can be employed to support data portability and system interoperability, which are critical to IT’s role as a business enabler rather than cost centre? With advanced academic training in electrical engineering, and a diverse work background in data, security, voice, video and wireless technologies developed through senior leadership roles at Liberty Global, Covad, TELUS and Qwest, Hussain is well equipped to offer guidance: in addition to his position at CenturyLink where he has led the product development and technology division since 2014, Hussain sits on the boards of several startup and non-profit organizations, is technical advisor to technology companies and holds 11 patents.
Cloud computing is often touted as a technology approach that reduces complexity for the adopting organization. While SaaS download from easy to use online portals is virtually instantaneous, PaaS is designed to automate configuration and provisioning of development environments, while IaaS aims at scaling compute capacity up, down or out and on-demand based on established thresholds, without additional human intervention. In reality, though, decisions around the acquisition and management of cloud environments and alignment with business goals are anything but simple, involving as they do, a need to opt for on-premise private cloud, dedicated hosted cloud, or public cloud resources – or any combination of all of the above. According to Hussain, the first step is to understand what the business needs are: “cloud is a big word, but at the end of the day, businesses are trying to optimize their core and increase their agility.” Cost is one consideration in cloud adoption, but as Hussain explained, the user organization is also typically looking for services that enable the development of competitive products and services, while allowing focus on core business without large investments of time and resources in IT infrastructure.
To achieve the optimal mix that will support business goals, Hussain argued that organizations must first understand learn to appreciate “the value of hybrid IT and how things will move to cloud over time.” As companies begin to shift away from development on legacy systems running on large mainframe environments, and to gather compute and networking resources for hosting on scalable external cloud platforms, Hussain believes they need to understand what decisions and behaviours they need to change or make to ensure cloud transition is successful. In addition, they need to research what external services are available in the marketplace and how these would map to their individual business needs. Customer education is a critical, first exercise.
A next step is to identify which applications and workloads will work on cloud and which won’t. According to Hussain, “all types of workloads can live in the cloud, but the cloud has to adjust and customers have to understand what type of clouds they want for running the application.” At an operational level, this means workload evaluation is an important means to determining what cloud delivery model is most appropriate. High compute, high availability and low latency applications, for example, may be better suited to a dedicated, bare metal offering in public cloud since sharing in multi-tenant infrastructure can mean that convergence on a VM can take time.
Companies also need to understand whether the application is self-aware and resilient enough to “work within clouds of clouds and in a distributed fashion,” Hussain added. It is the responsibility of the application designers within the user organization (as opposed to the provider), he argued, to ensure that the app can be run in the cloud: “if I am running an Oracle environment and developing different types of script to pull different types of data, I have to make sure that I can take all that and put it in a cloud where it will run and where it will be available when I need it to be.” For example, “applications that are single threaded, that are not able to read out traffic if a link drops, that are not aware of where they are running, that were written in a language that is twenty or thirty years old, such as Fortran – in languages that are not cloud sensitive” may need to remain on-premise, until rewritten in a modern language such as CC++, Python or JAVA. “The real limitation,” Hussain explained, “is how the application itself is written. Is it cloud aware and would it run in a distributed network? Is it able to run when processes are stopped and restarted, or when the application needs to move to another server? Can the application continue to run when spun up on another VM or in another cloud – can it work in a logical fashion in a distributed environment?”
The relative need for security and isolation of sensitive data associated with a particular application will also play a role in determining cloud deployment – critical or sensitive data, for example, may demand use of dedicated resources as opposed to multi-tenant cloud. A more recent consideration in cloud acquisition is the buyer, which in recent years has begun to shift from the CIO exclusively to encompass a broader user group. As Hussain noted, developers represent an increasingly significant customer set for the CenturyLink platform: for these, the provider’s ability to support scale, to offer database-as-a-service, automated provisioning of OS and memory, as well as services that help them migrate workloads from development to production environments or from in-house and back to cloud again, are important criteria in the decision to contract public PaaS services.
In the management of hybrid scenarios, a whole other set of criteria associated with interoperability and data portability come into play as means to ensure optimal IT delivery to the business. Hussain divides this discussion into two questions based on intended outcomes: how do organizations build seamless migration of workloads between their own private infrastructure and public/private/or multi-tenant environments; and how do they ensure that applications and data can interoperate between different provider clouds? On the first question, a management interface for a service such as DR-as-a-service can allow the cloud provider to migrate a workload in real time and with high availability without affecting what’s running on-premise. This can be managed through a number of mechanisms, but to deliver this capability, CenturyLink acquired Edmonton-based, DRaaS provider DataGardens last year and incorporated this functionality into its platform, which had the added benefit of enabling CenturyLink to quickly onboard customers to its cloud services.
Interoperabiltiy between clouds that will allow the user organization to realize additional value through cloud orchestration, on the other hand, depends on the type of APIs offered by the provider. Established clouds, such as Microsoft’s Azure or IBM’s Bluemix PaaS, have APIs that are clearly defined and typically OpenStack to ease integration for customers. “The APIs are key; all our APIs are OpenStack, and we now have API extensions available for the major platforms,” Hussain added. A second strategy for integration also originates with the provider, who may build out a service ecosystem through partnerships with other providers that have built their applications to be run on a partner’s public cloud environment. CenturyLink, for example, maintains an integration team that has onboarded close to a hundred partners and will add more if a customer requires a specialized integration. A third tactic involves creation of a virtual LAN interface that the customer can manage and which can later be disconnected.
Would be cloud adopters, then, who are looking for educational support and the flexibility needed to entertain hybrid options, as well as the ability to orchestrate clouds of clouds, are well advised to do due diligence in their provider selection. CenturyLink, which offers colocation, hosting and IT managed services in addition to a variety of cloud packages, positions its services mix as a ‘platform’ that can deliver on all these requirements: the platform supports sales which works to understand specific customer needs, the operations team, which has a single view of the customer, and the customer, who is able to source not only cloud, but also a range of data, voice, video, MPLs and CDN networking applications.
Another key element in CenturyLink’s value proposition is its strong heritage in communications, which has allowed the company to enter the cloud business with an intent to bring networking to the cloud – an approach that has enabled the delivery of strong functionality in orchestration, easy provisioning and scalability. Hussain explained: “It used to be that a customer would go to a cloud company and buy IT, to a networking company and buy networking, and go to a hosting company and buy hosting services, but because we have network, colo, cloud and managed applications, we bring this together in the platform we are continuing to build.” To develop its cloud services, the company has engaged in a number of acquisitions – Tier 3, AppFog, Orchestrate and DataGardens – and is now also working to enhance networking functionality. “When you have things running in the cloud, access networks are really important – access has to be perfect,” Hussain explained, “so we invest heavily in building more capacity, rolling out more fibre, and on innovation in our infrastructure to create more bandwidth that we can use to provide high quality service delivery guarantees to our customers.” In addition, the company is investing in network virtualization services, including SDN and NFV, which are now available to customers out of 43 CenturyLink data centres. While this combination of delivery components may offer the simplification that cloud buyers now look to, the company’s breadth of service delivery can provide the flexibility customers need as they negotiate the complexity that hybrid IT presents.