The digital transformation challenges CIOs to remodel their existing IT architectures providing their internal customers with a dynamic platform that stands for a better agility and fosters the companies’ innovation capacity. This change calls for a complete rethink of the historically implemented architecture concepts. Even if most of the current attempts are to migrate existing enterprise applications into the cloud, CIOs have to empower their IT teams to consider novel development architectures. Because modern applications and IoT services are innovative and cloud based.

Microservice: Background and Meaning

Typical application architectures, metaphorically speaking, remind of a “monolith”, a big massive stone that is made of one piece. The characteristics of both are the same: heavy, inflexible and not or not easy to modify.

Over the last decades, many, mostly monolithic applications have been developed. This means that an application includes all modules, libraries, and independencies that are necessary to ensure a smooth functionality. This architecture concept implicates a significant drawback. If only a small piece of the application needs to change, the whole application has to be compiled, tested and deployed again. This also implies for all parts of the application that don’t experience any changes. This comes at big costs taking manpower, time and IT resources and in most cases lead to delays. In addition, a monolith makes it difficult to ensure:

• Scalability
• Availability
• Agility
• Continuous Delivery

CIOs can meet these challenges by changing the application architecture from a big object to an architecture design composed of small independent objects. All parts are integrated with each other providing the overall functionality of the application. The change of one part doesn’t change the characteristics and functionality of other parts. This means that each part works as an independent process, respectively, service. This concept is also known as microservice architecture.

What is a Microservice?

A microservice is an encapsulated functionality and is developed and operated independently. So, it is a small autonomous software component (service) that provides a sub-function within a big distributed software application. Thus, a microservice can be developed and provided independently and scales autonomous.

Application architectures based on microservices, are modularized and thus can be extended with new functionalities easier and faster as well as better maintained during the application lifecycle.

Compared to traditional application architectures, modern cloud based architectures are following a microservice approach. This is because of the cloud characteristics cloud native application architectures have to be adapted to. This means that issues like scalability and high-availability have to be considered from the very beginning. The benefits of microservice architectures are related to the following characteristics:

• Better scalability: A sub-service of an application is able to scale autonomously if its functionality experienced a higher demand without affecting the remaining parts of the application.
• Higher availability of the entire application: A sub-service that experiences an error doesn’t affect the entire application but only the functionality it is representing. This means that a sub-failure necessarily doesn’t affect customer-facing functionality if the service represents a backend service.
• Better agility: Changes, improvements and extensions can be implemented independently from the entire application functionality without affecting other sub-services.
• Continuous delivery: These changes, improvements and extensions can be conducted on a regular basis without updating the whole application respectively without a major maintenance mode.

Another benefit of microservice architectures: A microservice can be used in more than one application. Developed once it can serve its functionality in several application architectures.

What Provider works with Microservices?

Today, a number of providers already understood the meaning of microservice architectures. However, in particular the big infrastructure players have their difficulties with this transformation. Startups respectively cloud native companies show how it works:

• Amazon Web Services
  From the very beginning Amazon AWS aligned its cloud infrastructure providing microservices (building blocks). Examples: Amazon S3, Amazon SNS, Amazon ELB, Amazon Kinesis, Amazon DynamoDB, Amazon Redshift
• Microsoft Azure
  From the very beginning the cloud platform consists of microservices. The Azure Service Fabric exists for a short time offering capabilities for the development of own microservices. Examples: Stream Analytics, Batch, Logic App, Event Hubs, Machine Learning, DocumentDB
• OpenStack in general
  The community extends the OpenStack portfolio with new microservices with each release mainly for infrastructure operations. Examples: Object Storage, Identity Service, Image Service, Telemetry, Elastic Map Reduce, Cloud Messaging
• IBM Bluemix
  IBM’s PaaS Bluemix provides an amount of microservices. These are offered directly by IBM or via external partners. Examples: Business Rules, MQ Light, Session Cache, Push, Cloudant NoSQL, Cognitive Insights
• Heroku/Salesforce
  Heroku’s PaaS offers “Elements”, a marketplace for ready external services that can be integrated as microservices in the own application. Examples: Redis, RabbitMQ, Sendgrid, Raygun.io, cine.io, StatusHub
• Giant Swarm
  Giant Swarm offers developers an infrastructure for the development, deployment and operations of microservice based application architectures. For this purpose, Giant Swarm is using technologies like Docker and CoreOS.
• cloudControl
  cloudControl’s PaaS offers “Add-ons”, a marketplace to extend self-developed applications with services from external partners. Examples: ElephantSQL, CloudAMQP, Loader.io, Searchify, Mailgun, Cloudinary

The providers, based on their microservice portfolios, are offering a programmable modular construction system of ready services that are accelerating the development of an application. These are ready building blocks (see hybrid and multi-cloud architectures), whose functionalities don’t have to be developed again. Instead they can be used directly as a “brick” within the own source code.

Example of a Microservice Architecture

Netflix, the video on demand provider, is not only a cloud computing pioneer and one of the absolute role models for IT architects. Under the direction of Arian Cockroft (now Battery Ventures) Netflix has developed an own powerful microservice architecture to operate its video platform high scalable and high available. Services include:

• Hystrix = Latency and fault tolerance
• Simian Army = High-availability
• Asgard = Application deployment
• Exhibitor = Monitoring, backup and recovery
• Ribbon = Inter process communication (RPC)
• Eureka = Load balancing and failover
• Zuul = Dynamic routing and monitoring
• Archaius = Configuration management
• Security_Monkey = Security and tracking services
• Zeno = In-memory framework

All microservices, Netflix encapsulates within its “Netflix OSS” that can be downloaded as open source from Github.

An example from Germany is Autoscout24. The automotive portal is facing the challenge to replace its 2000 servers that are distributed over 2 data centers, and the currently used technologies based on Microsoft, VMware and Oracle. The goal: a microservice architecture supported by a DevOps model to implement a continuous delivery approach. Thus, Autoscout24 wants to stop its monthly releases and instead provide improvements and extensions on a regular basis. Autoscout24 decided for the Amazon AWS cloud infrastructure and already started the migration phase.

Microservice: The Challenges

Despite the benefits, microservice architectures come along with several challenges. Besides the necessary cloud computing knowledge (concepts, technologies, et.al.) these are:

• A higher operational complexity since the services are very agile and movable.
• An additional complexity because of the development of a massive distributed system. This includes latency, availability and fault tolerance.
• Developer need operational knowledge = DevOps
• API management and integration play a major role.
• A complete end-to-end test is mandatory.
• Ensuring a holistic availability and consistency of the distributed data.
• Avoiding a high latency of the single services.

The Bottom Line: What CIOs should consider

Today, standard web applications (42 percent) still represent the major part at public IaaS platforms. By far mobile applications (22 percent), media streaming (17 percent) and analytics services (12 percent) follow. Enterprise applications (4 percent) and Internet of Things (IoT) services (3 percent) are still playing a minor part. The reason for the current segmentation: websites, backend services as well as content streaming (music, videos, etc.) are perfect for the public cloud. On the other hand enterprises are still sticking in the middle of their digital transformation and evaluate providers as well as technologies for the successful change. IoT projects are still in the beginning or among the idea generation. Thus in 2015, IoT workloads are only a small proportion on public cloud environments.

Until 2020 this ratio will significantly change. Along with the increasing cloud knowledge within the enterprises IT and the ever-expanding market maturity of public cloud environments for enterprise applications the proportion of this category will increase worldwide from 4 percent to 12 percent. Accordingly, the proportion of web and mobile applications as well as content streaming will decrease. Instead worldwide IoT workloads will almost represent a quarter (23 percent) on public IaaS platforms.

These influences are challenging CIOs to rethink their technical agenda and thinking about a strategy in order to enable their company to keep up with the shift of the market. Therefore they have to react to the end of the application lifecycle early enough by replacing old applications and look for modern application architectures. However, a competitive advantage only exists if things are done differently from the competition and not only better (operational excellence). This means that CIOs have to contribute significantly by developing new business models and develop new products like an IT factory. The wave of new services and applications in the context of the Internet of Things (IoT) is just one opportunity.

Microservice Architecture: The Impact

Microservice architectures support IT departments to respond faster to the requirements of specialty departments to ensure a faster time-to-market. In doing so, independent silos need to be destroyed and a digital umbrella should be stretched over the entire organization. This includes the introduction of the DevOps model to develop microservices in small and distributed teams. Modern development and collaboration tools are enabling this approach for worldwide-distributed teams. This helps to avoid shortage of skilled labor at certain countries by recruiting specialist from all over the world. So, a microservice team with the roles product manager, UX designer, developer, QA engineer and a DB admin could be established across the world, which accesses the cloud platform via pre-defined APIs. Another team composed of system, network and storage administrators operates the cloud platform.

Decision criteria for microservice architectures are:

• Better scalability of autonomous acting services.
• Faster response time to new technologies.
• Each microservice is a single product.
• The functionality of a single microservice can be used in several other applications.
• Employment of several distributed teams.
• Introduction of the continuous delivery model.
• Faster onboarding of new developers and employees.
• Microservices can be developed more easily and faster for a specific business purpose.
• Integration complexity can be reduced since a single service contains less functionality and thus less complexity.
• Errors can be isolated easier.
• Small things can be tested easier.

However, the introduction of microservice architectures is not only a change on the technical agenda. Rethinking the enterprise culture and the interdisciplinary communication are essential. This means that also the existing IT and development teams needs to be changed. This can happen either by internal trainings or external recruiting.

Worldwide many CIOs are still looking for the right answer to let their companies benefit from cloud computing capabilities. Basically all kinds of organizations are the right candidates for a bottom-up cloud strategy – the migration of existing applications and workloads into the cloud. However, this approach isn’t much innovative but it offers a relatively high value proposition with low risk. An analysis of market-ready cloud marketplaces show promising capabilities to implement the bottom-up cloud strategy in the near term.

The Cloud drives the evolution of IT purchasing

In the blog post “Top-Down vs. Bottom-Up Cloud Strategy: Two Ways – One Goal” two strategy approaches are discussed that can be used to benefit from public cloud infrastructure. One conclusion was that top-down strategies remain reserved for innovators. Bottom-up strategies are mainly realized in the context of existing workloads to move them into the cloud. CIOs of prestigious companies worldwide are still searching for best practices to heave their legacy enterprise workloads to the cloud.

Looking at the general purchasing behavior of IT resources unveils a disruptive change. Besides CIOs, IT infrastructure manager and IT buyers also department managers are asking for their right to be heard or are going their own ways. Driver behind this development: the public cloud. Its self-service leads to a vanishing significance of classical IT purchasing. Obtaining hardware and software licenses from distributors and resellers will become less important in the future. Self-service makes it convenient and easy at once to get access to infrastructure resources as well as software. Thus, distributors and resellers systematically have to rethink their business models. Those system houses and system integrators who still haven’t start their cloud transformation so far are at risk to disappear from the market in the next three to five years. So long!

Besides self-service the public cloud offers primarily one thing: Choice! More than ever before. On the one hand there is the ever-growing variety of sources of supply – solutions of providers. On the other hand the different deployment models the public cloud can be connected with. Hybrid and multi cloud scenarios are the reality.

The next step of evolution is well underway – cloud marketplaces. Implemented by the operators the right way they are offering IT buyers an ideal central platform for purchasing IT resources. At the same time they are supporting CIOs to push their cloud transformation with a bottom-up strategy.

Bottom-Up Cloud Strategy: Cloud Provider Marketplaces support the implementation

The bottom-up cloud strategy helps companies to move existing legacy or enterprise applications into the cloud to benefit from the cloud’s capabilities without thinking about innovations or to change the business model. It is mainly about efficiency, costs and flexibility.

In this strategy approach the infrastructure is not the central point but rather a means to the end. Finally, the software needs to be operated somewhere. In most cases the purpose is to continue to use the existing software in the cloud. At application level cloud marketplaces can support to be successful in the short term. More importantly they are facing current challenges and requirements of companies. These are:

• The distributed purchase of software across the entire organization is difficult.
• The demand for accessing software in the short term – e.g. for testing purposes – increases.
• Individual employees and departments are asking for a catalog of categorized and approved software solutions.
• A centrally organized cloud marketplace helps to work against shadow IT.

In light of the fact that still a vast number of valid software licenses are used in on premise infrastructure underlines the importance of Bring Your Own License (BYOL). BYOL is a concept by which a company continues using its existing software licenses legally at the cloud provider’s infrastructure.

In respect of supporting the bottom-up cloud strategy by a cloud marketplace, experience has shown that cloud provider owned and operated marketplaces like Amazon AWS Marketplace or Microsoft Azure Marketplace are playing an outstanding role. Both are offering the necessary technology and excellence to make it easy for customers and partners to decide for running applications in the cloud.

Technical advantage, simplicity and most of all an extensive choice are the key success factors of cloud marketplaces of public cloud providers. The AWS Marketplace already offers 2100+ solutions, the Azure Marketplace even 3000+ solutions. Some mouse clicks and the applications are deployed on the cloud infrastructure – incl. BYOL. Thus, the infrastructure is becoming easily usable for all users.

The providers are investing a lot in the development of their marketplaces – with good reason. Amazon AWS’ and Microsoft Azure’s own cloud marketplaces are having a strategic importance. These are the ideal tools to maneuver new customers to the infrastructure and to increase the revenue with existing customers.

Cloud marketplaces operated by public cloud providers are constantly becoming more popular – this with verifiable numbers. The marketplaces are having a big market maturity and are offering a big and wide variety of solutions. Against this backdrop, CIOs who are planning to migrate their existing applications into the cloud should intensively deal with cloud marketplaces. Because these kinds of marketplaces are more than just an add-on – they can support to accelerate the cloud migration.

Along with the steady growth of the cloud the question for appropriate cloud uses cases rises. After companies like Pinterest, Airbnb, Foursquare, Wooga, Netflix and many others have shown how cloud infrastructure and platforms can be used to create new or even disruptive business models, more and more CEOs and CIOs would like to benefit from the cloud characteristics. The issue: established companies run many legacy enterprise applications that cannot be moved into the cloud based on its existing form. For many decision makers this raises the question whether they should follow a top-down or bottom-up strategy.

Cloud Strategy: Top-Down vs. Bottom-Up

An IT strategy has the duty to support the corporate strategy at the best. Thus, in line with the increasing digitalization of society and economy the value proposition and meaning of IT significantly rises. This means that the impact of IT on the corporate strategy will become more important in the future. Assuming that cloud infrastructure, platforms and services are the technological fundament of the digital transformation, it is consequent that the cloud strategy has a direct impact on the IT strategy.

This raises the question how far cloud services are able to support the corporate strategy whether direct or indirect. It is not mandatory that this is reflecting in numbers. If a company – for instance – is able to let its employees work more flexible – based on a software-as-a-service (SaaS) solution – than it has done something for the productivity, which has a positive effect on the company. However, it is important to understand that cloud infrastructure and platforms just serve as a foundation on which companies get the capabilities to create innovation. The cloud is just a vehicle.

Two approaches can be used to get a better understanding about the impact of cloud computing on the corporate strategy:

• Top-Down Cloud Strategy
  During the top-down approach the possibilities of cloud computing are analyzed and a concrete use case is defined. So an innovation or idea is created that is enabled based on cloud computing. On this basis the cloud strategy is created.
• Bottom-Up Cloud Strategy
  During the bottom-up approach an existing use case is implemented regarding the possibilities of cloud computing. This means it is analyzed how the cloud can help to support the needs of the use case. Thereof the respective cloud strategy is extrapolated.

Mainly the top-down approach comes up with new business models or disruptive ideas. The development is done on the green field within the cloud and mostly belongs to innovators. The bottom-up approach follows the goal to move an existing system or application into the cloud or redevelop it there. In this case it is mostly about to keep an existing IT resource alive or in best case to optimize it.

Bottom-Up: Migration of Enterprise Applications

Existing companies prefer to follow a bottom-up strategy in order to quickly benefit from the cloud capabilities. However, the devil is in the details. Legacy or classical enterprise applications are not developed to run on a distributed infrastructure – thus a cloud infrastructure. This means that it is not natural for them to scale out and they are only able to scale up at the utmost by using e.g. several Java threads on one single system. If this system fails also the application is not available anymore. Every application that should run in the cloud thus also has to follow the characteristics of the cloud and needs to be developed for this purpose. The challenge: Companies still lack of appropriate staff with the right cloud skills. In addition, companies are intensively discussing “Data Gravity”. This is about the inertia respectively the difficulty of moving data. Either because of the size of the data volume or because of a legal condition that requires to store the data in the own environment.

Vendors have recognized the lack of knowledge as well as the “Data Gravity” and try to support the bottom-up strategy with new solutions. Based on the “NetApp Private Storage” NetApp allows companies to balance the “Data Gravity” between public cloud services and the own control level. Companies have to fulfill several governance and compliance policies and thus have to keep their data under control. One solution is to let the cloud services access the data in a hybrid cloud model without moving them. Thus in this scenario the data is not directly stored in the cloud of the provider. Instead the cloud services are accessing the data via a direct connection when processing them. NetApp enables this scenario in cooperation with Amazon Web Services. For example, Amazon EC2 instances can be used to process the data that is stored in an Equinix colocation data center – the connection is established via AWS Direct Connect.

Another challenge with not cloud ready enterprise applications in the public cloud is the data level – when the data should leave the cloud of a provider. The reason is that cloud native storage types (object storage, block storage) are not compatible with common on-premise storage communication protocols (iSCSI, NFS, CIFS). In cooperation with Amazon Web Services, NetApp Cloud ONTAP tries to find a remedy. As some kind of NAS storage the data is stored on an Amazon Elastic Block Storage (EBS) SSD. In this case Cloud ONTAP serves as a storage controller and ensures the access of not cloud ready enterprise applications to the data. Due to the compatibility to common communication protocols the data can be moved easier.

VMware vCloud Air targets at companies with existing enterprise applications. The vCloud AIR public cloud platform based on the vSphere technology and is compatible to on-premise vSphere environments. So existing workloads and virtual machines can be moved back and forth between VMware’s public cloud and a virtualized on-premise infrastructure.

ProfitBricks tries to support companies with its Live Vertical Scaling concept. In this case a single server can be vertically extended with further resources – like amount of CPU cores or RAM – without rebooting the server. Thus the performance of a running virtual server can be enhanced without making changes to the application. The best foundation for this is a LAMP stack (Linux, Apache, MySQL, PHP) since e.g. a MySQL database recognizes new resources without any adjustments and a reboot of the host system and is able to use the added performance immediately. To make this possible ProfitBricks did modifications at operating system and hypervisor (KVM) level – that are transparent for the user. The customers only have to use the provided reference operating system image that includes the Live Vertical Scaling functionality.

A random bottom-up use case of enterprise applications in the public cloud can be find at Amazon Web Services. However, this example also shows that the meaning of system integrators in the cloud is rising.

• Amazon Web Services & Kempinski Hotels
  The hotel chain Kempinski Hotels has migrated the majority of its core applications and departments among others finance, accounting and training to the Amazon cloud infrastructure. Together with system integrator Cloudreach a VPN connection was established between the own data center in Genf and the Amazon cloud over which the 81 hotels worldwide are now provided. Furthermore, Kempinski plans to completely shut down the own data center to move 100 percent of its IT infrastructure to the public cloud.

Top-Down: Greenfield Approach

In contrast to just keep enterprise applications the Greenfield approach follows the top-down strategy. In this case an application or a business model is developed from scratch and the system is matched to the requirements and characteristics of the cloud. In this context we are talking about a cloud native application that is considering the scalability and high-availability from the beginning. The application is able to independently start additional virtual machines if more performance is necessary (scalability) respectively to shut them down when they are not needed anymore. It is the same when a virtual machine fails. In this case the application also independently takes care that another virtual machine is starting as a substitute (high-availability). Thus, the application is able to work on any virtual machine of a cloud infrastructure. One reason is that any machine can fail at any time and a substitute needs to be started. In addition, also the data that are processed by an application are not stored at a single location anymore but are stored distributed over the cloud.

Startups and innovative companies are aware of this complexity what the following uses cases display.

• Amazon Web Services & Netflix
  Netflix is one of the lighthouse projects on the Amazon cloud. The streaming provider is using each characteristics of the cloud what is reflected by the high availability as well as the performance of the platform. As one of the pioneers on the Amazon infrastructure Netflix has developed its own tools – Netflix Simian Army – from the very beginning to master the complexity.

However, the recent past has shown that innovative business models not necessarily have to be implemented on a public cloud and that the greenfield approach not only belongs to startups.

• T-Systems & Runtastic
  Runtastic is an apps provider for persistence, strength & toning, health & wellness as well as fitness and helps users to reach their health and fitness goals. The company has a massive growth. After 100.000 downloads in 2010 and 50 million downloads in 2013, the number of downloads reached over 110 million until today. Furthermore, Runtastic counts 50 million users worldwide. Basically the numbers speak for an ideal public cloud scenario. However, due to technical reasons Runtastic decided for T-Systems and runs its infrastructure in two data centers in a colocation IaaS hybrid model.
• Claranet & Leica
  Last year camera manufacturer Leica has launched “Leica Fotopark”, an online photo service to manage, edit as well as print and share photos. Managed cloud provider Claranet is responsible for the development and operations of the infrastructure. “Leica Fotopark” is running on a scale out environment based on a converged infrastructure and a software defined storage. The agile operations model is based on the DevOps concept.

Greenfield vs. Enterprise Applications: The Bottom-Line

Whether a company decides for the top-down or bottom-up cloud strategy depends on its individual situation and the current state of knowledge. The fact is that both variants help to let the IT infrastructure, IT organization and the whole company become more agile and scalable. However, only a top-down approach leads to innovation and new business models. Nevertheless, one has to consider that e.g. for the development and operations of the Netflix platform an excellent understanding for cloud architectures is necessary, which is still few and far between on the current market.

Regardless of their strategy and especially with regard to the Internet of Things (IoT) and the essential Digital Infrastructure Fabric (DIF) companies should focus on a cloud infrastructure. Those offer the ideal preconditions for the backend operations of IoT solutions as well as the exchange for sensors, embedded systems and mobile applications. In addition, a few providers support with ready micro services to simplify the development and to accelerate the time to market. Furthermore, a worldwide spanning infrastructure of datacenters offers a global scalability and helps to expand to new countries quick.