The cloud has become standard practice: From small businesses to large firms, the principle of cloud computing has been understood by companies and many IT departments are already using cloud-based services. Hardly any company can afford to forego cloud computing in today's digital world, where more and more aspects of IT are becoming software-defined. The public cloud in particular has established itself as the technical foundation for the business models of many startups.
One of the best known examples of success with the cloud is the Netflix video streaming service. Streaming of the online video library has been multiplied by a thousand in the past eight years. Scaling up on this scale would scarcely have been possible without the use of a cloud service.
Whether public cloud, private cloud or hybrid cloud, in the end most companies will land on a multicloud solution – a combination of several cloud services that behave like a single large cloud from the user perspective.
Multicloud takes things a step further than a hybrid cloud. A hybrid cloud extends a private cloud – on-premise or hosted – with a public cloud. For this combination to work smoothly, companies generally rely on the same public cloud platform that they use for their private cloud. This means that with a private cloud based on VMware infrastructure, a company will seek out a public cloud service that also relies on VMware.
In contrast, a multicloud knows no boundaries: It combines many more cloud services with different deployment models such as private or public cloud, and various expressions such as Infrastructure as a Service (IaaS) and Platform as a Service (PaaS), into a single platform. Different cloud technologies such as VMware or Amazon AWS can also be combined. This makes the hybrid cloud a kind of midway point from the private cloud and public cloud to a multicloud.
Supply chains in the automotive industry are often used as an example when explaining how a multicloud works: Vehicle manufacturers rely on multiple suppliers, who sometimes even supply some of the same components. These are put together by the manufacturer into a vehicle. The multicloud works in a similar fashion: It orchestrates a wide range of cloud providers and services to combine them into one big cloud.
The goal of a multicloud is to use applications from various services and infrastructures – based on load and need – at the same time or depending on the pricing models of the various cloud services. According to a survey by IT consultancy Crisp Research last year, multicloud is the model of choice for 36 percent of German small businesses surveyed. More than a third of German SMEs thus plan to use cloud services from more than one provider.
A study by market research and consultancy firm IDC, also from last year, indicated that 86 percent of US companies surveyed were already planning a multicloud approach to be able to cover their IT needs over the coming two years.
Multicloud offers numerous advantages over other deployment models for cloud services, such as a hybrid cloud:
Always the right cloud: There is no cloud service or cloud infrastructure suitable for all uses. A cloud service that is perfect for one application may be absolutely wrong for another. With a multicloud strategy, a company can always use the best cloud service for each purpose, in accordance with the best of breed approach.
No vendor lock-in: Using a range of cloud providers and services provides protection from vendor lock-in, overdependency on a single provider. Former cloud provider Nirvanix serves as an example of the importance of spreading risk: This US-based cloud service, a partner to IBM – and ranked by Gartner as a top cloud provider – shut down its services just weeks after giving notice. Customers were hard put to move their data to other cloud providers and adapt their systems in such a short time.
Fail-safe reliability: Even highly available cloud platforms such as those offered by Amazon, Google and Microsoft can go down – as has been shown repeatedly in the past few years. A company that relies on various providers and services prevents such an outage from impacting multiple areas of the business.
Performance: The smaller the distance between the data center and the company, the shorter the latency periods. A multicloud offers the possibility of using a nearby data center for those applications where speed is of the essence, while other applications can be directed through a more distant center, which may also be less costly to use.
Despite these benefits, multicloud also has some disadvantages, compared with other cloud deployment models:
More error sources: The more cloud providers and services are used, the more interfaces and therefore potential sources of error. Using multiple interfaces also offers cybercriminals more points of attack.
Resource drain: As the number of cloud services rises, so does the cost in terms of time and money needed to administer them. Cloud providers' constantly changing price models also need to be tracked, so that management of the cloud can be adapted accordingly.
Data protection: Every cloud provider has its own data protection rules. These rules have to be individually compared with the customer firm's own compliance requirements. Businesses need to be careful, particularly when it comes to customer data. But the data protection drawback can also be turned into an advantage – where each set of compliance requirements is matched to the right cloud service.
This is also why true multicloud management rarely exists so far, if by that we mean management in which very different clouds blend seamlessly, and workloads for example are automatically shifted from one cloud provider to another as needed.
The interplay of cloud services and various applications is multifaceted: IT managers have to bring all the workloads in the company under a single roof and connect them with uniform interfaces. Some of these are traditional business applications that often still run as on-premise software on local servers and workstations. Many of these applications are unable to support even horizontal scaling, meaning that they do not run on several machines in parallel. And other applications that are outward-facing often run on the public cloud or on mobile devices.
Service Level Agreements and the already mentioned data protection represent additional tangles to deal with.
Managing the complex layers of permission rights is another major undertaking that should not be underestimated. Along with managing access rights for individual cloud services, the active directory services often used by companies and their corresponding rights must be connected to the cloud and adapted to the relevant services. Access rights for outside workers are another layer to consider.
Such considerations quickly push a multicloud to the edges of manageability, so that only the bravest of IT managers dare attempt it.
The heart of a multicloud is a cloud management platform (CMP) that administers and combines the various cloud environments. Diverse in-house and external applications run on these clouds. The CMP manages these environments and makes them automatically available – users have no need to worry about which private or public cloud or which hybrid cloud model is being used in the background, in which data center.
Companies should prepare extensively before implementing a multicloud strategy. This is the only way to optimize all the important aspects, such as data security, availability, stability, scalability and cost, to fully leverage the advantages of the multicloud.
Organize communication streams: Running applications on multiple cloud services requires sophisticated cross-cloud communication. Proper orchestration across various clouds at every level must be ensured – from availability to monitoring to backup. Ensuring communication with acceptable latency periods and maximum security – such as through encryption – must be clarified with each individual cloud provider.
Mobility: One advantage of the multicloud is that individual applications can be shifted back and forth between cloud services as needed, for example to spread workload or geographic distribution. Such shifting cannot be allowed to impact the availability of an application. This means that applications must be prepared to run without major adjustments in other clouds.
Central tools: Companies need to know where and how the central infrastructures needed for a multicloud environment – such as the cloud management system, monitoring and reporting – will be operated. Will these be in-house, or will management be purchased as a service?
Backend connections: Companies need to prepare the connection of the multicloud environment with their own backend systems, and manage access.
Availability: What availability is required by the individual services? This can be boosted, for example via global load distribution. The greater the availability of a service, the greater the costs.
Security: Security management is key. What data should be managed in the cloud? Other questions include which cloud services will store which data where, whether and when data is encrypted, and how access protection is managed. Nexinto also highlights the importance of deciding whether Internet access will take place centrally through one cloud or locally through all clouds.
Orchestration: If a company is already using a cloud management platform, will it support a multicloud strategy?
There are no turnkey solutions for a multicloud – the different cloud providers, services and technologies are too complex and varied. There are products from Cloud Technology Partners, Flexiant, Nexinto, Rightscale and SlipStream in the market, for example.
In most cases, a multicloud strategy cannot be implemented without an outside partner such as a systems retailer: According to Crisp Research, almost two-thirds of German small businesses rely on outside support to implement their cloud projects (see chart, above left). Seven percent of companies leave such projects entirely to an outside service provider – and this number is trending upward.