Source-linked AI summary
Decision Support Tools for Cloud Migration in the Enterprise
Ali Khajeh-Hosseini, Ian Sommerville, Jurgen Bogaerts, Pradeep Teregowda
TL;DR
Enterprise cloud migration decisions are difficult because costs, benefits, risks, existing systems, and organizational changes must be considered together. The paper presents a cost-modeling tool and a benefits-and-risks spreadsheet, evaluating them through two case studies. The tools informed decision makers about cloud costs, benefits, and risks, while the cost model remained limited to infrastructure costs.
Problem
Enterprise cloud migration requires assessing costs, benefits, risks, existing-system integration, and organizational changes, but available decision-support tools are often marketing-oriented or proprietary.
Method
The paper develops a public-IaaS cost-modeling tool and an enterprise benefits-and-risks spreadsheet, then evaluates them in two case studies.
Results
The tools informed decision makers about the costs, benefits, and risks of using public IaaS clouds.
Takeaways & Limitations
The tools provide decision support by combining infrastructure cost estimation with enterprise-oriented benefits and risks assessment.
Takeaways & Limitations
The paper models only infrastructure costs, so full cloud-migration costs require project-management and software-cost estimation techniques.
Abstract
from arXiv · showhide
This paper describes two tools that aim to support decision making during the migration of IT systems to the cloud. The first is a modeling tool that produces cost estimates of using public IaaS clouds. The tool enables IT architects to model their applications, data and infrastructure requirements in addition to their computational resource usage patterns. The tool can be used to compare the cost of different cloud providers, deployment options and usage scenarios. The second tool is a spreadsheet that outlines the benefits and risks of using IaaS clouds from an enterprise perspective; this tool provides a starting point for risk assessment. Two case studies were used to evaluate the tools. The tools were useful as they informed decision makers about the costs, benefits and risks of using the cloud.
I. INTRODUCTION
Enterprise cloud migration decisions are difficult because existing systems, costs, benefits, risks, and organizational changes must be considered together. The paper addresses this need with two impartial decision-support tools evaluated in two case studies.
- Existing enterprise systems often require brownfield integration with a range of systems, complicating migration to public IaaS clouds.
- Cloud migration decisions require evaluating benefits, risks, costs, and organizational and socio-technical changes.
- Available guidance is often provided through provider marketing materials or proprietary tools tied to expensive consultancy contracts.
- The paper describes and evaluates two impartial tools intended to support enterprise cloud migration decisions.
- The tools were evaluated through two case studies representing different types of systems.
II. COST MODELING
The cost-modeling tool represents applications, data, infrastructure, and resource usage to estimate infrastructure costs across cloud deployment choices. It handles variable usage through elasticity patterns, simulates resource consumption over time, and reports detailed cost breakdowns.
- Cloud infrastructure cost estimation is complicated by uncertain resource consumption and deployment-dependent charges such as inter-cloud data transfer.
- The tool models software, data, virtual machines, storage, databases, remote nodes, and communication paths using an extended UML deployment model.
- Users specify providers and resource requirements including virtual-machine hours, storage, storage I/O requests, and data transfers.
- Elasticity patterns express temporary or permanent changes to baseline resource usage, including peaks, drops, and linear or exponential trends.
- The simulator processes each node and edge month by month, multiplies resource usage by provider-specific unit prices, and produces time-varying cost reports.
- Group-level cost breakdowns let architects compare deployment options, including duplicating system components across clouds for increased availability.
III. BENEFITS AND RISKS ASSESSMENT
The assessment tool brings together enterprise benefits and risks of public IaaS clouds to support decision making beyond direct cost calculations. It organizes evidence into categories, consequences, mitigations, and indicators, while allowing users to rate items from their own perspective.
- Assessment rationale: Enterprise assessment must consider indirect benefits and risks alongside costs, including flexibility, business continuity, compliance, customer relationships, and public image.Some indirect benefits, such as improved time-to-market, are difficult or meaningless to quantify directly.
- Evidence base: Benefits and risks were identified from over 50 academic papers and industry reports and categorized as organizational, legal, security, technical, or financial.The presented tables include only a subset of the identified benefits and risks; complete tables were made available separately.
- Assessment tool: The spreadsheet provides a starting point for risk assessment by listing risks, potential consequences, mitigation approaches, and indicators.Users can rate each item from unimportant to very important using a Likert scale.
- Organizational risks: The identified risks include migration-related morale and productivity effects, extra management effort across several clouds, provider changes, organizational resistance, and incident-handling mismatches.These organizational risks can create anxiety, hidden management costs, service disruption concerns, or limited responses during incidents.
- Legal and security risks: Legal and security risks include unusable software licenses, non-compliance with data and industry regulations, unauthorized access, denial-of-service attacks, and interception of management or transit data.The listed consequences include regulatory non-compliance, unavailable resources, increased usage bills, and possible infrastructure manipulation.
- Mitigation approaches: Suggested mitigations include involving experts and stakeholders, checking provider SLAs and licensing agreements, using multiple providers, selecting compliant jurisdictions, and encrypting data.These approaches are paired with indicators such as job-uncertainty rumors, insufficient incident information, physical software locks, undisclosed data-center countries, and unsupported encryption.
A. Digital Library and Search Engine
The digital library case study used the tools to compare cloud-provider costs and assess enterprise-relevant benefits and risks for a large, growing system.
- System context: CiteSeerx indexes over 1.5 million documents, receives around 2 million hits per day, and requires around 2TB of storage.The system is deployed on 15 servers in a university data center.
- Cost modeling: AWS was found to be the cheapest cloud provider among the alternatives investigated for CiteSeerx.
- Cost modeling: AWS costs were distributed across VMs at 66%, data transferred out at 20%, storage at 10%, and storage I/O requests at 4%.This breakdown identifies potential cost-optimization targets, including switching off unused VMs.
- Cost modeling: Using AWS S3 was estimated to cost around $4,000 more than EBS over three years because of storage I/O request costs.The estimate may be inaccurate because S3 I/O requests are difficult to infer from Linux iostat disk-I/O figures, and providers price storage I/O differently.
- Benefits and risks: The benefits-and-risks exercise took around one hour and identified 7 benefits and 13 risks as important.
B. R&D Division of a Media Corporation
The R&D division case study compared AWS deployment options for an enterprise environment and used the assessment spreadsheet to examine technical and organizational benefits and risks.
- System context: The R&D division belongs to a media corporation with over 20,000 employees worldwide and manages applications on 9 servers and 2 network storage systems.Its systems form part of a larger interconnected corporate IT environment.
- Cost modeling: The cost study compared non-elastic AWS instances, elastic AWS instances, and an elastic setup using one small instance per user.The proposed thin-client arrangement included session servers and compute servers for centralized work and heavy processing.
- Cost modeling: The deployment options were evaluated using the cost modeling tool against the AWS-EU cost table.
- Cost modeling: Over three years, the elastic setup cost around $70,000 less than the non-elastic setup.Using 4 high-memory quadruple-extra-large instances cost around $4,000 less than using 120 small instances.
C. Discussion
The case studies reveal that stakeholders assign different importance to cloud benefits and risks, including differing views within the same enterprise. Technical benefits often matter strongly, while risk priorities vary by organizational context and division.
- Benefits: Technical benefits were more important than organizational and financial benefits for the digital library and the media corporation’s R&D division.The digital library emphasized handling volatile demand and growth, while R&D emphasized simplified provisioning and anywhere/anytime resource access.
- Benefits: The media corporation’s corporate IT department considered financial and organizational benefits more important than technical benefits.
- Risks: Financial, technical, and security risks were the main concerns in the digital library case, whereas legal and organizational risks were not important there.The case involved a small enterprise handling non-sensitive data.
- Risks: Risk priorities differed within the media corporation: R&D appreciated the importance of different risk types, while corporate IT focused mainly on organizational and legal risks.Corporate IT was not highly concerned with financial risks because budgets were probably managed by local divisions.
- Stakeholder perspectives: The case studies show that cloud migration decisions must consider multiple stakeholder perspectives, including differing views among divisions within one enterprise.
V. RELATED WORK
Related work includes case studies and cloud cost-modeling research, but the paper positions its tool as a practical model spanning infrastructure, deployment, elasticity, providers, and price changes. Its benefits-and-risks tool synthesizes a broader enterprise-oriented assessment than security-focused prior work.
- Evaluation: The paper’s case-study approach addressed limited academic investigation of factors involved in cloud migration and evaluated the decision-support tools in multiple settings.
- Cost modeling: Prior research investigated cloud costs through individual case studies, user-oriented cost modeling, and performance-versus-cost trade-offs across clouds.
- Cost modeling: The paper distinguishes its cost-modeling tool by supporting practical models of infrastructure, deployment options, elasticity patterns, cloud-provider cost models, and changes to those models.
- Benefits and risks: The paper’s benefits-and-risks review builds on sparse academic and industry work and provides a starting point for enterprise risk assessment.
- Benefits and risks: Unlike prior work focused mostly on security risks, the paper argues that other risk categories should also be considered during cloud migration decisions.
VI. CONCLUSION
The paper presents cost-modeling and benefits-and-risks tools for public-IaaS migration decisions and evaluates them across a small technical system and a corporate enterprise setting. The tools were useful to participating enterprises, but the analysis covers infrastructure costs rather than the full cost of migration.
- Contribution: The paper describes two tools: a cost model for applications, data, infrastructure, and usage patterns, and a benefits-and-risks spreadsheet for enterprise risk assessment.
- Evaluation: The tools were evaluated with a small-team technical system and a corporate enterprise system embedded in a larger interconnected IT environment.
- Limitation: The paper discusses only infrastructure costs, so full cloud-migration costs require complementary project-management and software-cost-estimation techniques.
- Findings: The case-study enterprises found the tools useful because they informed decisions about public-IaaS costs, benefits, and risks.
- Future work: The authors plan longitudinal follow-up on migration decisions and their effects on systems and organizations, while making the tools available through ShopForCloud.com.