200 likes | 303 Views
效能管理 ( Performance Management). 陳彥錚 (Yen-Cheng Chen) yencheng@mcu.edu.tw http://www.im.tj.mcu.edu.tw/~ycchen/. 效能管理 (PM) 之定義. Performance Management (PM) 量測與評估網路介質、硬體、軟體之效能,以促進有效的網路通訊。. 效能管理實施步驟. 1. 搜集目前網路設備與鏈結之使用效能相關資訊。 2. 分析所搜集之效能相關資訊。 3. 設定使用率或相關效能參數之臨界值 (Threshold) 。
E N D
效能管理( Performance Management) 陳彥錚 (Yen-Cheng Chen) yencheng@mcu.edu.tw http://www.im.tj.mcu.edu.tw/~ycchen/
效能管理 (PM)之定義 • Performance Management (PM) • 量測與評估網路介質、硬體、軟體之效能,以促進有效的網路通訊。
效能管理實施步驟 1. 搜集目前網路設備與鏈結之使用效能相關資訊。 2. 分析所搜集之效能相關資訊。 3. 設定使用率或相關效能參數之臨界值(Threshold) 。 4. 進行網路模擬。
搜集效能相關資訊 • 網路伺服器: • processor load, disk access rate, network interface card utilization • 橋接器/路由器: • packet forwarding rate, processor load, percentage of dropped frames on each interface, number of packets being held in a queue.
Max (total bits sent , total bits received) bandwidth bandwidth total bits sent + total bits received = util% util% = 網路鏈結使用率(Link Utilization) • 一般計算方式(e.g, Ethernet, Token Ring, FDDI) • Full-Duplex Serial Link (e.g. 專線64K, ..., T1, T3)
利用SNMP讀取流量資訊 • 一般網路設備會隨時記錄其網路介面卡從開機至目前為止已流入及已流出之位元組個數,即SNMP MIB II 中之 ifInOctets 及 ifOutOctets。 • 使用SNMP,每隔一段時間週期性地讀取對應某一網路介面卡之ifInOctets及ifOutOctets值。 • 此次所讀得之值減去上次所讀得之值,即為此段時間內之流量。 • 使用率 = 一週期內之流量 / (頻寬週期)
Example • 假設某一效能管理應用程式針對此網路設備之一T1專線介面(1.544Mbps)進行流量監測, 10:00 AM: ifInOctets = 1,500,000 ifOutOctets = 1,200,000 10:05 AM: ifInOctets = 2,500,000 ifOutOctets = 7, 200,000
使用率計算 • 流量計算 • In: 2,500,000 - 1,500,000 = 1,000,000 Bytes • Out: 7,200,000 - 1,200,000 = 6,000,000 Bytes • 流量 = Max (1,000,000,6,000,000) Bytes = 6, 000,000 Bytes = 48,000,000 bits • Util% = 48,000,000/(1,544,000 60 5) 100% = 10.36%
服務層次之效能量測 • Service Level Measurement • Total Response Time • Rejection Rate • Availability
Service Level Measurement • Total Response Time • the amount of time it takes a datum to enter the network and be processed and for a response to leave the network. • From the viewpoint of applications • Round Trip Time (R.T.T) is measured from the viewpoint of transport protocol.
Service Level Measurement • Rejection Rate • the percentage of time the network cannot transfer information because of the lack of resources and performance. • Availability • the percentage of time the network is accessible for use and operational. • Usually measured as MTBF (Mean Time Between Failure)
效能資訊分析 • 圖形化表現效能資訊 • Historical plots: weekly, monthly, quarterly, yearly • Real-time graphical analysis • Trend Prediction
What to be Analyzed/Graphed? • Device Information • memory usage, processor utilization, disk access rate, number of sessions. • Link Information • utilization, error rate, error percentage
效能臨界值之設定 • Set thresholds on a variety of items affecting network performance • When the thresholds are crossed, events are reported. • In general, the values of thresholds are determined according to past experience.
Thresholds • Threshold Priority • In general, priority: low, medium, high • Multiple threshold values for the same item • Thresholds for multiple items • Use rearm mechanism to avoid frequent threshold events
Rearm util% Threshold Rearm 1 2 3 4 5 6 7 time
效能預測(Prediction) • Use Regression to predict future trend. • Apply Statistics Theory • Should consider possible factors that affect the prediction. • Network Simulation
Prediction util% Threshold value . . Predicted utilization increase . . . . . . . . . . . . . . Computed actual utilization time