Instructions:From the left column, select the
step in the indicated move of the MICE Model that best corresponds to each group of sentences in this introduction. The feedback is shown at the bottom.
Adapted from Mark Sederholm, "Techno-Economic Analysis of Software-Defined Data Center Networks," Master's thesis, Dept. of Elect. Eng., Aalto Univ., Espoo, Finland, 2017.
1 Introduction (steps within moves)
Step
Move 1: Current situation
Step 1
Step 2
Step 3
1Over the last half-century, data centers (DC) have evolved into large-scale, complex systems supporting a wide range of service offerings.
2More recently, we have witnessed an enormous change in service delivery patterns after service providers started to adopt cloud-computing practices in their DCs.
Step
Move 1: Current situation
Step 1
Step 2
Step 3
3Cloud computing represents an operative model that abstracts distributed computing resources from underlying infrastructure and delivered as a service, which can be rapidly provisioned and released with minimal management effort [1, 2].
4The cloud model is based on abstractions produced using various virtualization techniques.
5In the early development phases, these techniques primarily represented a mechanism that enables not only pooling of physical servers into a single logical computing resource but also sharing it with multiple tenants [3].
6Later, with the advent of virtualization technologies and identified benefits, its exploitation also spread to other parts of DCs.
7Adding abstraction layers into data center design brings new elasticity and automation capabilities, which help cloud service providers to significantly increase operational efficiency.
8With enhanced automation capabilities, cloud workloads can be instantiated, suspended and seamlessly moved between different servers using zero touch operations, thus driving down administration costs and significantly lowering the provider's operational expenses.
Step 1
Step 2
9However, the cloud service delivery process requires not only configuration of computer nodes but also storage and networking devices.
10Computing and storage have successfully leveraged virtualization techniques to automate their domains, whereas networking remains mainly stuck in the physical world, thereby slowing down cloud deployments [4].
11Although many virtualization primitives are currently available for networks, including Virtual Local Area Networks (VLANs) and Virtual Routing and Forwarding contexts (VRFs), they do not provide adequate abstractions upon which to build external control software [4].
12Thus, managing cloud networks often involves a significant amount of manual network management, which leads to high operational expenses and long delivery times [5].
Step
Move 1: Current situation
Step 1
Step 2
Step 3
13Nevertheless, network services are often enhanced with L4-to-L7 service functions, such as firewalls, load balancing and intrusion detection systems, which are tied together to form a service chain that delivers desired end-to-end service attributes.
14Provisioning of these services has traditionally been based on the deployment of proprietary hardware appliances for each network function and interconnecting them by following static, rigid configuration processes.
Step 1
Step 2
15However, this traditional network service provisioning method is completely incompatible with highly scalable cloud data center design and has become a barrier to rapid, flexible cloud service deployment [6].
Step
Move 3: Proposing a new approach
Step 1
Step 2
16A promising solution to these problems would be to decouple service provisioning from network infrastructure.
Step 1
Step 2
17However, this approach requires that the network infrastructure be opened and exposed to external control software.
Step
Move 1: Current situation
Step 1
Step 2
Step 3
18Recently, two paradigms for providing such an opportunity have received considerable attention from the networking industry: Software Defined Networking (SDN) and Network Function Virtualization (NFV).
19While SDN enhances network programmability by abstracting the data plane and logically centralizing network control [7], NFV focuses on virtualizing and decoupling network functions from hardware appliances, thus allowing them to be dynamically initiated on any platform.
20The new level of openness and programmability provided by these paradigms contributes to network automation, which plays a key role in the convergence of networks within the cloud service-provisioning ecosystem [8].
Step
Move 1: Current situation
Step 1
Step 2
Step 3
21An important component in the cloud service-provisioning ecosystem is the service orchestrator.
22The orchestrator automates some of the repetitive tasks and coordinates the end-to-end service delivery process, ensuring appropriate service levels and configurations [9].
23As the service is ordered, the orchestrator translates high-level service description into multiple provisioning requests, coordinated across multiple resource domains.
Step 1
Step 2
24However, setting up an orchestration engine involves costs in terms of integration and set-up time.
Step 1
Step 2
25Although numerous studies have investigated the performance aspects of orchestration in service delivery, few of these have focused on it from a techno-economic perspective.
Step
Move 4: Your solution
Step 1
Step 2
Step 3
Step 4
Step 5
26Therefore, the aim of this thesis is to evaluate the economic feasibility of adopting an orchestrator for service delivery processes used in a highly virtualized data center environment that leverages SDN and NFV.
Step
Move 4: Your solution
Step 1
Step 2
Step 3
Step 4
Step 5
27The feasibility of the orchestrator will be evaluated using a set of orchestrated service delivery simulations conducted in an emulated data center environment.
28The results will be analyzed and compared with the traditional service delivery methods, which are usually performed by manually provisioning end-to-end services
Step
Move 4: Your solution
Step 1
Step 2
Step 3
Step 4
Step 5
29In order to facilitate understanding about related domains, this thesis will provide a brief overview of each technology.
30The thesis is organized as follows.
31Chapter 2 defines cloud computing.
32Chapter 3 introduces the NFV framework which will be used as the reference architecture for the use cases.
33Chapter 4 briefly covers the concept of service function chaining and examines its core components.
34Chapter 5 provides an overview of the SDN concept and architecture.
35Finally, Chapters 6 and 7 will introduce the use case(s), results and analysis.