Cloud Computing for Large-Scale Complex IT Systems

The notion of cloud computing , where computing infrastructure, platforms, and software application services are offered at low cost from remote very-large-scale data centres accessed over the internet, is one that has recently received large amounts of attention in the IT industry. There have been predictions that this 'utility computing' will predominate in future with organisations discarding their internal servers in favour of applications accessible in the cloud . To service users, clouds clearly offer advantages in scalability, may reduce the costs of application management, and may reduce overall hardware costs. To service providers, they offer the opportunity to leverage existing data-centre infrastructure and to take advantage of the economies of scale available exclusively to purchasers of extremely large volumes of hardware and network capacity. While we think that some predictions of the costs savings from cloud adoption are optimistic, we are confident that cloud computing offers real business benefits. These will mean that more and more business and public-sector organisations will migrate some of their applications from dedicated servers to private or public clouds.This proposal extends the work of the existing EPSRC-funded Large-Scale Complex IT Systems (LSCITS) Initiative's research programme. The interests of the LSCITS Initiative in cloud computing are twofold: we argue that there is a need to provide analytical and predictive methods and tools that support decision making about the costs, benefits and risks of migrating applications to the cloud; and we believe that, because cloud data centres are instances of LSCITS, there are research challenges in their setting-up, operation, management and evolution while providing guaranteed levels of performance and dependability, quantified environmental impact, support for the development of scalable applications; and a range of resource pricing models.This proposal envisages a situation where enterprises are satisfied that cloud computing is a viable option for them and they wish to plan for a situation where a significant fraction of their LSCITS are deployed on the cloud. The deployment could be wholly outsourced on external shared enterprise clouds, such as those provided by IBM or HP or on a dedicated in-house managed facility, such as a hypothetical NHS cloud. Most likely, large organisations will have a mix of facilities using dedicated servers, an in-house cloud and external cloud providers. We will not tackle immediate short-term (2009) problems of cloud computing. Rather, we will focus on two, longer-term research themes, which are applicable to both data-centre design and application migration:1. Modelling and Simulation. We aim to develop simulation tools and techniques to help document, analyse and predict the attributes and behaviour of LSCITS where some or all of these systems are deployed in the cloud. These techniques will support application system and application portfolio modelling as well as the modelling of risks, pricing and the environmental impact of cloud-based systems.2. Dependability and Quality of Service We will explore how aspects of LSCITS such as dependability, resilience, and performance, are affected by the new benefits and vulnerabilities that arise from deploying systems in the cloud; we will explore methods for quantifying the degree to which these types of requirements are achieved and we will devise new theory and tools to help researchers and practitioners reason about the dependability and quality of service delivered by the cloud. We believe that these are important areas because cloud providers and enterprise users of cloud resources will require tools and techniques to inform their decision-making about the costs and risks of deploying business-critical systems on the cloud, and about the related issues of management, maintenance, and ongoing evolution.

Cloud computing is already important economically and looks set to grow very significantly over the next few years. Research by IDC suggests that the market for cloud-computing services was $16bn in 2008 and will rise to $42bn/year by 2012. Starting from conservative assumptions on the cost reduction process, Etro (2009) suggests that the diffusion of cloud computing will provide a positive contribution to the annual European growth rate of 0.2%, and will contribute to create about a million new jobs through the development of a few hundred thousand new SMEs in the EU. Although it is difficult to distinguish trustworthy data from hype, it is obvious that growth in this area will be very significant and that cloud computing is of major industrial significance. We anticipate that the project will have a major impact on three industrial communities (the industrial beneficiaries): 1. Commercial cloud-computing providers such as HP and IBM, with whom we have already established relationships. The work proposed here on price and risk modelling and on self-star systems is particularly relevant to this community. 2. Cloud users, particularly larger companies with an established IT infrastructure, who may be interested in moving some of their services to the cloud. The work proposed on modelling cloud applications and application portfolios will be particularly relevant for this class of user. 3. Cloud consultants, who will support industrial companies in the migration of their applications to the cloud. Realistically, the work that we will do will not be packaged in such a way that it can be used without extensive support. While this is available in-house in some of the larger companies, we anticipate that much of this support will be provided by consultants (mostly SMEs) who will provide services to application users. As well as these industrial beneficiaries, government agencies are also potential beneficiaries. Our current discussions in the LSCITS project with the Home Office have indicated that government data centres may move to a cloud model, with dedicated government clouds, so the work on cloud service management and on application migration is highly relevant. We will maintain our relationship with the Home Office to track development in this area and so be in a position to present the work to decision makers. The key benefit of this project is quite clearly economic but there are, in addition, potential environmental benefits that may ensue. At this stage, it is practically impossible to reliability monetise these savings, partially because we need research like that proposed here to help understand what percentage of applications can move to the cloud. Nevertheless, Nicholas Carr, author of a highly influential book on the shift to cloud computing, claims that in an average organisation's IT facilities, 80% of server capacity is wasted, 65% of storage capacity is wasted, and 70% of IT staffing costs are devoted to ongoing maintenance of legacy applications. Cloud computing offers the prospect of drastic reductions in these levels of wastage, and both the economic and environmental impacts of such reductions are likely to be highly significant. Furthermore, dynamic pricing of cloud services increases economic efficiency, because it reduces the risk to the seller of lost sales (from fixed prices set too high) or lost revenues (from prices fixed too low); and for the buyer it offers the ability to accurately signal the buyer's importance of that transaction, via the price that the buyer bids. The primary environmental benefits are likely to be due to the economies of scale and efficient allocation of scarce resources that underlie the cloud computing rationale. Economies of scale mean that larger data-centres are usually significantly more energy-efficient per unit of cloud resource provided, and efficient-allocation policies attempt to minimise the amount of infrastructure that lies idle.