Should your business retain its in-house data centre or outsource it to a service provider? What are the drivers and barriers when it comes to cloud infrastructure? What kinds of workloads are best suited to cloud deployment, and who are the leading cloud platform providers? We look at recent recent research on these important IT architecture questions.
Organisations seeking to exploit today's highly virtualised, flexible and scalable 'cloud' IT infrastructure have a number of deployment options, depending on the level of ownership and control they wish to retain over the server, storage and networking resources involved, and the data processed by these IT resources.
The most conservative approach is to house and manage the IT infrastructure in your organisation's own data centre, running whatever workloads are currently suitable for deployment on a private cloud platform. Established organisations will probably run such 'cloud-ready' workloads alongside existing enterprise applications during a period of experimentation and transition, as they dip their toes into the world of private cloud computing.
Many companies — startups and small businesses, for example — won't want the expense and hassle of running their own data centre. One solution is to house your servers, storage and networking gear in a colocation facility, but manage them (largely remotely) yourself. Alternatively you can retain your on-premises data centre and outsource its management to a service provider. A popular solution is the managed private cloud, in which both the physical infrastructure and its management are outsourced, leaving you to specify how the virtualised resources are deployed.
Businesses are often reluctant to entrust mission-critical workloads to the public, 'multi-tenant', cloud because of perceived security worries (among others). However, public cloud IT infrastructure is often deployed as part of a hybrid strategy, in order to cope with peaks in demand for workloads that primarily run in private clouds.
In this overview article, we'll explore just what's meant by Infrastructure as a Service (IaaS), what's driving organisations and businesses to adopt it or remain wary of it, what kinds of workloads are best suited to running on cloud platforms, what different platforms are available to deliver IaaS in private and public clouds, and who are the leading IaaS vendors.
Infrastructure as a Service (IaaS) traditionally sits at the base of a cloud computing pyramid that also comprises PaaS (Platform as a Service) and SaaS (Software as a Service), with the amount of customer control over the virtualised IT resources decreasing as you ascend the pyramid. In a typical IaaS deployment, the service provider manages everything below the hypervisor layer (the server, storage and networking hardware and its virtualisation), leaving the customer to handle operating system, middleware and application deployment on self-service virtual machines that can be 'spun up' on demand, usually via a web-based dashboard.
As noted above, IaaS deployment options include single tenant (private cloud) or multi-tenant (public cloud), with IT resources housed in the customer's data centre, in a third-party service provider's facility or operating as a hybrid private/public configuration.
Neat definitions rarely fit the real world, of course, and according to market researcher Forrester, the lines between SaaS, IaaS and PaaS are becoming increasingly blurred: SaaS vendors are moving into platforms with application extension tools (Salesforce.com's Force.com, for example); PaaS vendors are supporting IaaS configuration (Engine Yard, for example) or delivering IaaS directly (Microsoft Windows Azure, for example); and IaaS vendors are embracing PaaS-like abstract development layers (Amazon Web Services, for example).
Over the last couple of years, private and public IaaS have progressed along Gartner's Hype Cycle for Cloud Computing: both occupied the 'Peak of Inflated Expectations' in 2011, while by 2012 public IaaS was edging its way towards the 'Slope of Enlightenment', with private cloud computing following it into the 'Trough of Disillusionment'.
It would seem, then, that public IaaS is well on the way to delivering widespread productivity, with private clouds — which will generally require much more in-house expertise to deliver — apparently lagging some way behind.
Cloud drivers and barriers
Managing a company's in-house data centre is a challenging task, especially in a less-than-benign economic climate. In a recent Brocade-sponsored survey, cost and complexity are cited as the biggest management challenges, followed by reliability, performance and scale:
Little wonder, then, that Brocade's survey notes considerable enthusiasm among IT decision-makers for highly virtualised on-demand (cloud) data centres:
Interestingly, the survey also finds that some departments in organisations are exploring cloud solutions without the involvement of their IT departments — although most are seeking IT guidance:
Customer or end-user demand for cloud computing is also identified as a key driver in the Uptime Institute's 2013 Data Center Industry Survey, with 43 percent of respondents listing this compared to just 13 percent in 2012:
The main barrier to public cloud adoption in the Uptime Institute survey — as usual — is concern over security, followed by a lack of cloud computing skills and expertise:
The same survey shows greater uptake of private cloud than public cloud computing, with 44 percent of respondents currently deploying private clouds (down from 49% in 2012) versus 28 percent (up from 25% in 2012) opting for public cloud:
At the same time, the Uptime Institute finds that, after excluding third-party service providers, public cloud adoption among enterprises has increased between 2012 and 2013, from 10 percent to 17 percent. Somewhat counter-intuitively, larger companies (managing over 5,000 servers) are twice as likely to deploy public cloud solutions as smaller ones (managing under 1,000 servers).
This trend towards public cloud is echoed by the Uptime Institute's findings on data centre budgets: the largest year-on-year increases (>10%) are reported by third-party data centre operators (63% of respondents) compared to enterprises (25%); meanwhile, budget decreases are negligible among third parties but noticeable among enterprises:
If third-party data centre service providers are growing at the expense of in-house enterprise facilities, as these figures suggest, then companies are increasingly going to have obsolete, outdated, unused or otherwise 'comatose' servers on their hands. Removing and recycling this equipment will significantly improve the financial and energy efficiency of enterprise IT operations. According to the Uptime Institute's 2013 survey, only 13 percent of respondents believe that 10 percent or more of their servers are comatose, although the true figure may be higher as nearly half of the respondents lack an auditing procedure for identifying and removing unused systems.
What should go in the cloud?
If your business is evaluating private, public or hybrid cloud infrastructure, you need to consider the types of workloads you run, and whether these are suitable for deployment on a highly virtualised architecture that's designed for applications that can scale out rather than scale up. Companies will usually run some combination of cloud-native applications, websites and web apps, mainstream business and enterprise applications (some of them mission-critical), test-and-development workloads, and batch computing jobs such as HPC (High-Performance Computing) and Big Data analytics.
In 2011, PWC asked a sample of 489 senior executives and middle managers (including CIOs and CTOs) about the cloud-readiness of their organizations' workloads. At the time, only 14 percent of respondents estimated that more than half of their workloads were cloud-ready, but this figure rose to 51 percent in the 2014 timeframe:
Back in 2011, data storage and retrieval was the most prevalent type of workload deployed on cloud infrastructure. This was predicted to remain the most popular category in 2014, but was expected to be joined by significant amounts of other kinds of workload — transaction processing, batch computing, web services and high-performance computing/analytics:
A significant omission here is the category of mainstream business and enterprise applications, mostly running under Windows, that serve vital functions in many organisations. Rewriting legacy (Windows or Linux) software to run efficiently in the cloud is expensive and time-consuming, and organisations may be better off continuing to run some of the more complex applications in the traditional way.
However, solutions are available to 'cloudify' platform-native software in a painless manner, notably from Numecent with its cloudpaging technology and recent Native as a Service (NaaS) offering. Here, instead of running a Windows application such as Photoshop on a remote server and streaming the pixels to a client device, cloudpaging is used to stream the program itself over the network in such a way that it can load quickly on the client (using its CPU and GPU resources) and call up additional functionality as required.
The NaaS service, which also looks after licensing and has robust data security features, currently runs on Amazon Web Services (AWS) infrastructure and is aimed at ISVs and developers, but is expected to be tailored for service providers, telcos and enterprises, and be available as a self-hosted offering, in the future.
The alternative, of course, is to move away from platform-native software and replace it with cloud-native SaaS applications.
When it comes to creating a private cloud, the vital component is the virtualisation platform on which it's built. Factors to consider include the ease with which virtual machines can be moved between hypervisors (if you're planning on running more than one), and whether you can create a hybrid cloud by integrating with a public IaaS provider.
A recent survey from IT systems management company Opsview canvassed 420 IT decision-makers in large and medium-sized companies in North America and Europe. Not surprisingly, market leader VMware dominates the private cloud space in this survey, followed by Microsoft, giving proprietary platforms 78 percent of the preferred-vendor share. OpenStack is the best-supported open-source platform with just 12 percent:
When it comes to public cloud IaaS, another big beast, Amazon Web Services (AWS), dominates the scene. Together with Rackspace, these two service providers have 74 percent of the preferred-vendor share in Opsview's survey:
Increasingly, public cloud service providers will want to interoperate with enterprise private clouds in order to facilitate hybrid deployments. This will be easiest where the service provider and the enterprise use the same virtualisation platform. Rackspace, as the founder and main supporter of OpenStack, should have an advantage in the open-source space, while Microsoft, having added IaaS to its Windows Azure service, is well placed to improve its competitive position versus VMware.
The outlook for enterprise IT is definitely cloudy. For reasons of energy and cost efficiency, IT decision-makers will want to virtualise their data centres, remove or repurpose any 'comatose' servers, and in many cases outsource the data centre entirely. Many businesses will benefit from access to on-demand private, public or hybrid cloud infrastructure, although they will need to evaluate the cloud-readiness of their workloads and perhaps make some adjustments to the mix. The IaaS market is evolving rapidly and includes both proprietary and open-source platforms, which makes the choice of technology and service provider a challenge for enterprise IT architects. Given that a hybrid cloud architecture is likely to prove a good fit for many businesses, service providers that make it easy to integrate private and public cloud infrastructure are likely to prosper.