隨著企業資訊環境朝向高速網路、虛擬化平台及混合式資訊架構發展，資料傳輸效能已成為影響資訊服務穩定性的重要因素。近年來，企業核心網路已普遍升級至10Gbps 或25Gbps，但既有伺服器、虛擬機器及儲存設備仍大量使用1Gbps或更低速率之網路介面，使企業環境逐漸形成非對稱頻寬架構。當高速發送端持續向低速接收端進行大量 TCP資料傳輸時，可能導致端點TCP接收緩衝區壅塞，造成資料傳輸效能下降。然而，企業資訊維運多以網路設備為主要分析對象，較少從端點層級探討資料傳輸異常形成原因。 本研究以企業資訊環境為背景，採用實驗研究法建置符合企業情境之實驗平台，模擬高速發送端經交換器降速至1Gbps，再由Guest OS虛擬網路介面限速100Mbps之非對稱頻寬環境，利用Windows Performance Monitor、Wireshark、VMware vCenter及 iPerf3等工具收集端點遠端監測資訊，分析CPU Ready Time、RTT、RTT Jitter、TCP Retransmission、Receive Window、Zero Window與有效網路傳輸量等效能指標之變化，以探討端點TCP緩衝區壅塞形成過程。研究結果顯示，端點TCP緩衝區壅塞並非瞬間發生，而是由Hypervisor資源競爭、端點處理延遲及TCP流量控制等因素逐步形成之連續過程。CPU Ready Time、RTT 與 RTT Jitter通常最先產生變化，其後依序出現TCP Retransmission、Receive Window縮小及Zero Window等現象，而有效網路傳輸量則於後期才明顯下降，顯示端點遠端監測資訊具有預警能力。 依據實驗結果，本研究提出四層端點遠端監測分析架構(Four-layer Endpoint Telemetry Analysis Framework)，整合作業系統層、TCP 協定層、Hypervisor層及資料傳輸層之監測資訊，並提出黃金救援時間(Golden Recovery Window)之操作性定義，作為資料傳輸效能尚未明顯惡化前之最佳介入時機。此外，本研究建立企業端點TCP緩衝區壅塞偵測流程、調校流程及維運標準作業程序（SOP），提供企業資訊部門建立一致化之資料傳輸分析與維運管理機制。 本研究之主要貢獻包括：建立四層端點遠端監測分析架構，提供跨層級資料傳輸分析方法；提出黃金救援時間之操作性定義，建立端點壅塞預警機制；建立企業端點TCP 緩衝區壅塞偵測、調校及維運標準作業程序，以提升故障定位效率及資訊服務品質。研究成果可作為企業資訊管理、資訊服務管理及資訊治理之實務參考，亦可作為後續端點遠端監測與資料傳輸效能管理研究之基礎。;With the widespread adoption of high-speed networking， virtualization platforms， and hybrid IT infrastructures， data transmission performance has become a critical factor affecting enterprise information services. Although enterprise backbone networks have been upgraded to 10GbE or 25GbE， many legacy servers， virtual machines， and storage systems continue to operate with 1 GbE or lower-speed network interfaces， resulting in asymmetric bandwidth environments. Continuous TCP traffic transmitted from high-speed senders to lower-speed receivers may gradually lead to endpoint TCP receive buffer congestion and transmission performance degradation. However， enterprise troubleshooting is generally focused on network devices rather than server endpoints. This study investigates the formation mechanism of endpoint TCP buffer congestion in asymmetric bandwidth environments from the perspective of enterprise information systems. An experimental platform representing a typical enterprise environment was established using Microsoft Windows Server， VMware ESXi， Windows Performance Monitor， Wireshark， VMware vCenter， and iPerf3. Long-duration TCP traffic was generated through an asymmetric bandwidth architecture in which traffic was limited from a 1 GbE physical connection to a 100 Mbps virtual network interface. Endpoint telemetry was collected from the operating system， TCP protocol， hypervisor， and data transmission layers. Key performance indicators， including CPU Ready Time， Round-Trip Time(RTT)， RTT Jitter， TCP Retransmission， Receive Window， Zero Window， and effective network throughput， were analyzed to investigate the congestion formation process. The experimental results indicate that endpoint TCP buffer congestion is a progressive process rather than an instantaneous event. CPU Ready Time， RTT， and RTT Jitter increase during the early stage， followed by TCP retransmissions， shrinking Receive Window， and Zero Window events， while effective network throughput decreases only in the later stage. These findings demonstrate that endpoint telemetry provides valuable early-warning capability for congestion detection. Based on the experimental results， this study proposes a Four-layer Endpoint Telemetry Analysis Framework， integrating telemetry information from the operating system， TCP protocol， hypervisor， and data transmission layers. Furthermore， an operational definition of the Golden Recovery Window is introduced to identify the optimal intervention period before transmission performance significantly deteriorates. An enterprise-oriented detection procedure， tuning procedure， and standard operating procedure (SOP) are also established to support standardized transmission performance analysis and operational management. The major contributions of this study are threefold. First， a Four-layer Endpoint Telemetry Analysis Framework is established to support cross-layer transmission analysis. Second， an operational definition of the Golden Recovery Window is proposed as an early-warning mechanism for endpoint congestion. Third， standardized procedures for endpoint TCP buffer congestion detection， tuning， and operational maintenance are developed to improve fault localization efficiency and enhance information service quality. The proposed framework provides practical references for Enterprise Information Management， IT Service Management (ITSM)， and IT Governance， and serves as a foundation for future research on endpoint telemetry and enterprise transmission performance management.
