In this thesis we propose a Content-Based Differentiated Resource Management Mechanism for web cluster. The major difference of our system from other researches is that we used the URL string of content for request routing and admission control. In the system we implemented two important mechanisms for resource management. The Content-Based Load Balancing is used to share the service requests of each server node and the admission control avoids overload and provides differentiated web service for high priority content. We separate the content into Static Content and Dynamic Content. For Static content, we consider the influence of I/O resource and for dynamic content we consider the influence of CPU resource. When system running server nodes send their load information to distributor periodically, and distributor sorts these server nodes according their load. The I/O load of server is its network bandwidth usage and CPU load information is its CPU Utilization. In the period that the system updates the load information, requests are be forwarded to Light Load Set. CPU light load set are the servers, which have lower CPU Utilization, and I/O light load set are servers having lower network bandwidth usage. And we use load index of each request recorded in URL table and bucket of each node to distribute the traffic evenly. When overload, admission control drops the lower priority request to protect the high priority requests. With priority and differentiated service, our system can provide more resource to important content. Experiment results show that the proposed system has better performance than Linux Virtual Server(LVS) with Laver4 routing mechanism including RR(Round Robin) and WLC(Weight Least Connection).

Best effort networks fail to deliver the level of service emerging Internet applications demand. As a result many networks are being transformed to Quality of Service (QoS) networks, of which most are Differentiated Services (DiffServ) networks. While the deployment of such networks has been feasible, it is extremely difficult to overhaul the transport layer protocols such as Transmission Control Protocol (TCP) running on hundreds of millions of end nodes around the world. TCP, which has been designed to run on a best effort network, perform poorly in a DiffServ network. It fails to deliver the performance guarantees expected of DiffServ. In this thesis we investigate two aspects of TCP performance in a DiffServ network unaccounted for in previous studies. We develop a deterministic model of TCP that intrinsically captures flow aggregation, a key component of DiffServ. The other important aspect of TCP considered in this thesis is its' transient behavior. Using our deterministic model we derive a classical control system model of TCP applicable in a DiffServ network. Performance issues of TCP can potentially inhibit the adoption of DiffServ. A DiffServ network commonly use token buckets, that are placed at the edge of the network, to mark packets according to their conformance to Service Level Agreements (SLA). We propose two token bucket variants designed to mitigate TCP issues present in a DiffServ network. Our first proposal incorporates a packet queue alongside the token bucket. The other proposal introduces a feedback controller around the token bucket. We validate both analytically and experimentally the performance of the proposed token buckets. By confining our changes to the token bucket we avoid any changes at end-nodes. The proposed token buckets can also be incrementally deployed. Most part of the Internet still remains as a best effort network. However, most nodes run various QoS functions locally. We look at one such important QoS function, i.e. the ability to survive against flows that are non-responsive to congestion, the equivalent of a Denial of Service (DoS) attack. We analyze existing techniques and propose improvements.

Techn. Universiẗat, Diss., 2003--Braunschweig.

通訊與資訊科技的大幅進步，電信自由化帶來的激烈競爭，以及網際網路的蓬勃發展，刺激大量多媒體網路資訊的流通，為了因應此種趨勢，網路提供者已趨向合併數據及電信網路朝單一的All-IP網路方向發展。為了保證時效性服務在All-IP網路上的品質，網路服務品質(QoS)已成為All-IP網路的主要研究議題。不同的網路應用各有不同的特性與需求；對於那些比較不注重傳輸延遲時間的應用，增加網路頻寬或許就已足夠應付需求，但是對於那些具有互動特性 (interactive)、重視傳輸延遲時間的應用，像VoIP，除了增加網路頻寬外，All-IP網路必須提供服務品質保證才能獲得網路營運者的支持。本研究團隊設計一個管理架構，在此架構上提供完整的End-to-End QoS保證，以符合All-IP網路上各種不同服務需求。本文中另外提出以預先批購頻寬的方式進行核心網路資源規劃，根據需求預測，考量批購成本期望值，決定出適當之頻寬預購值。於執行時段提出數個允入控制資源不足解決方案，並且配合執行時段頻寬管理機制，掌握頻寬使用情形，以達到順利允入網路訊務之目的。最後於NS2平台以實驗模擬的方式，評估本文中所提出之預先批購頻寬與執行時段頻寬管理機制，從結果中我們發現本研究所提出之頻寬預購方法可以有效預防因為預測誤差所造成之資源不足現象，配合執行時段頻寬管理機制可根據執行時期之資源使用狀況，在資源缺乏時提前進行頻寬補充，以順利允入訊務，提升使用率。 / Because of great progress of communication and computer technology, aggressive deployment of broadband fiber optical network, advance of Internet technology, and the global standardization of IP technology, the telecommunication industry is moving toward a converged network, which uses a single global IP based packet-switching network to carry all types of network traffics. In these types of network traffics, different traffic types require corresponding service to ensure end- to-end quality. For carrying all types of network traffics on All-IP network, BBQ (Budget-Based QoS) research group propose a QoS system architecture to provide end-to-end QoS guarantee. In this thesis, basing on BBQ QoS system, we propose resource pre-planning mechanism to management core network. According to demand forecast from historical data and considering pre-planning cost, pre-planning mechanism will find optimal policy to reduce management cost. Besides resource pre- planning, we also propose several solutions to avoid resource shortage at run-time. Through intensive evaluation in network simulatior-2(ns2), we demonstrate that our resource pre-planning can minimize resource cost and cover some forecasting error. And run-time resource management can maintain reasonable in-hand resource at run-time to reduce the effect of resource shortage.

當有線網路已經發展成熟的同時，整個網路的發展趨勢是走向寬頻、行動化，以達到隨處可上網為目的。因此，無線網路技術成為現在炙手可熱的熱門話題，然而在目前的無線傳輸環境中，許多問題是在有線架構上看不到的，包含了無線平台的可移動性、電力、高封包遺失率及相較有線環境下的低頻寬，所以要在無線環境提供良好的服務品質，仍然有許多技術尚待克服。 這篇論文提出了一個整合差別服務在無線網路應用上的方法。在這個方法中，我們觀察無線環境的封包遺失情況，以及封包被轉送的次數，在擷取點上根據我們提出的滿意度計算公式，適當地提升封包的優先權值，使無線主機的封包在有線環境中，較不容易被丟棄。透過這個方法，可以提升無線主機上的使用者對網路品質的滿意度，並且對於整體的網路，可以節省下因為丟棄無線主機封包所浪費的資源。在模擬實驗中，我們測試了UDP及TCP資料流的不同組合，並且觀察比較單跳接網路和多跳接網路的差異，以找出在何種情況下，我們的方法會有最好的效果。 / When the wired network technologies are getting matured, the whole network trends are to develop "broadband" and "mobile" network infrastructure, and to enable everyone access the Internet everywhere. Therefore, the wireless network technology is a hot topic in current network research. However, there are many problems in the wireless network, including mobility of mobile hosts, battery power, high packet loss rate, and lower bandwidth that never appear in the wired environment. In order to provide good Quality of Service(QoS) in wireless environments, there are still many problems needed to be solved. This thesis proposed a method to integrate Differentiated Service(Diffserv) and wireless networks. To provide certain QoS for mobile applications, we define a "satisfaction" formula. By monitoring the packet loss in the wireless environment, the number of times of packet forwarding, we can base on the formula to raise the priority of wireless packets. By doing this, we could improve the network quality of satisfaction for mobile users, and save the resource of the entire networks, which is wasted by dropping wireless packets. In the simulation, we tested the combinations of UDP and TCP data flows, and compared their performance in both single-hop and multi-hop wireless networks. The results gave us implications how to utilize the Diffserv for mobile applications under any circumstances.

Thesis (Ph. D.)--University of Texas at Austin, 2000. / Vita. Includes bibliographical references (leaves 126-132). Available also in a digital version from Dissertation Abstracts.

The main aim of this work was to study the issue of the SNMP protocol and MIB database in order to focus on management and configuration of MIB database using the SNMP protocol. The Master's thesis in detail describes especially creating a communications model of the Manager - Agent type in an OPNET Modeler. In this modeling tool we created exchange of information from the side of the Manager to the side of the Agent via SNMP packets. The Master's thesis is mainly focused on the creation of the SNMP packet according to defined rules, then on reading of information from the MIB database on the side of the Agent and saves this information to the Manager’s MIB database. Exchange of informations is realized via the messages of type GetRequest, GetNext and GetResponse. These models are programmed in language C/C++ and then implemented into the environment of OPNET Modeler.

The issue of converged networks is to ensure the sufficient quality of services along the entire length of the communication transmission. This issue is closely connected to the real-time services, such as VoIP (Voice over Internet Protocol) and videoconferencing. These services require strict adherence to quality parameters, otherwise their function is not guaranteed. This problem particulary resolves subsystem IMS (IP Multimedia Subsystem), which concluded on the basis of user profiles can provide the required quality of service. Therefore the theoretical part deals with the description of the structure of the system and protocols designed to signal the network. Various mechanisms to support quality of services in access and backbone networks are also described. The following section explains the principle of provision of quality requirements of end-user networks. In the practical part is this theoretical knowledge used for designing and configuration of the network consisting of various technologies. The resulting model is then simulated in Opnet Modeler program, which is used for designing and testing of packet networks. Each simulation shows the effect of mapping quality requirements in the different access network on technologies, which are supported in the backbone. The outcome of this work is detailed network analysis and comparison of mechanisms for implementing quality of service. The conclusion summarises all simulation outcomes.

Nowadays, MPLS VPN has become a widely used solution on the issue of QoS guarantees against unexpected changes of network environments. This thesis investigates the system performance of the BGP-based MPLS VPN, which supports per class differentiated services. The results are compared with those without VPN. In this study, the target network is simulated through the OPNET simulator. Through adjusting the network parameters and creating different scenarios, such as network congestion and disconnection, we make statistical analyses based on the simulation results. It is observed that in addition to increasing the labels contained in each packet, MPLS VPN require PE routers be capable of supporting more protocols, such as searching for IP tables, and transferring various tables for the use of routing. Therefore, introducing VPN may increase processing load and overhead for data transmission. On the other hand, MPLS VPN may take longer convergence time in establishing entire routing messages than Non-VPN. However, when network disconnection occurs, the former has better throughput than the latter due to the shorter convergence time in the search of new routes. Also, if the networks become congested, the transmission delay of EF traffic in MPLS VPN is smaller since the alternative LSP for it is pre-established. For the same disconnected LSP route, MPLS VPN network also achieves better throughput, because the guaranteed EF traffic can be rerouted via different LSP routes. According to our simulation results, the background traffic has the largest throughput decrease, while the EF Traffic has the least.

Techn. University, Diss., 2003--Berlin.