標題: RMON臨界值調整模式於乙太網路管理之研究
The RMON Threshold Configuration Model on Ethernet Management
作者: 江米鳳
Mi-Feng Chiang
羅濟群
Chi-Chun Lo
管理學院資訊管理學程
關鍵字: 網路管理;警訊;臨界值;MIB;RMON;Network Management;Alarm;Threshold;MIB;RMON
公開日期: 2005
摘要: RMON是IETF (Internet Engineering Task Force)所訂立的標準,網路設備廠商皆遵循此規格來進行網路錯誤管理,是普遍的作法之一。在RMON中定義了兩個重要的參數「Rising臨界值」以及「Falling臨界值」,利用這兩個臨界值的設定來鎖定網路中錯誤封包過多的通訊埠,由於其臨界值檢查的時間都不長,因此在網路發生異常時,能夠在第一時間便掌握到錯誤的來源,同時利用網路癱瘓前的短暫時間發出警訊去通知網路管理者,有助於網路錯誤的診斷與分析,加上如果長期去觀察各個錯誤的趨勢,更可以在網路產生不利的影響前去修正可能發生的錯誤。因此,若是不當的設定這兩個參數,將會造成錯誤的誤判以及誤發警訊,導致未能在異常發生時,即時去通知網路管理者。 因此,本研究採用計量管制圖中的平均數與標準差管制圖技術來建立RMON臨界值調整模式;首先蒐集在不同封包長度下單一介面的最大傳輸量,以作為不同封包種類其臨界值設定上限訂定的依據,同時採取異常管制圖判斷的法則來作為臨界值調整的依據。臨界值調整模式是依據平均流量以及六個標準差(six sigma)來作為管制界限範圍的界定,自動去修正上管制流量界限以及下管制流量界限。經由模擬實驗結果得知,本研究所提出的臨界值調整模式確實能夠反應網路使用狀況,即時回報網路的異常情況,來加強網路錯誤管理的效能。因此,期望以此研究提供給網路設備開發廠商在開發產品時或是網路管理者利用RMON進行網路管理時的一個參考。
RMON is a standard set up by IETF (Internet Engineering Task Force) that network equipment manufaturers could follow the spec to build the network device. It’s one of popular fault management means and defines two important parameters of “Rising Threshold” and “Falling Threshold”. It causes miscarriage judgment and false alarm message for management when network is normal condition actually if the two parameters are not applicable, RMON sets threshold to monitor if packets are abnormal or not for ports, due to the monitor interval time is so short that it could alarm the fault to network management immediately when it’s unstable but not yet getting paralysis. It’s helpful for anaylysis and diagnosis by tracking trend and the management could take action to stop the fault before damage happen. Thus, Our research tries to build up a dynamic RMON threshold adjustment model through variable control chart methodology. We collect maximun throughput in different packet length that could be baseline of maximun threshold referal, meanwhile use control chart judgment methodology-Six Sigma to dynamicly adjust “Rising Threshold” and “Falling Threshold”. The testing result shows our model could response realitic network status and reacts abnormality appropriately, which will enhance the management performace. We expect our research could provide a methodology when network equipment manufacturers develop products and network management set up thresholds.
URI: http://140.113.39.130/cdrfb3/record/nctu/#GT009364522
http://hdl.handle.net/11536/80008
顯示於類別:畢業論文