{"type":"doc","content":[{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"設計理念","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"空間換時間","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"多級緩存，靜態化","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 客戶端頁面緩存（http header中包含Expires/Cache of Control，last modified(304，server不返回body，客戶端可以繼續用cache，減少流量)，ETag）","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"strong","attrs":{}}],"text":"反向代理緩存","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"strong","attrs":{}}],"text":"應用端的緩存(memcache)","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"strong","attrs":{}}],"text":"內存數據庫","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"strong","attrs":{}}],"text":"Buffer、cache機制（數據庫，中間件等）","attrs":{}}]}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"索引機制","attrs":{}}]},{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":" 哈希、B（B+）樹、倒排、bitmap","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"哈希索引：適合綜合數組的尋址和鏈表的插入特性，可以實現數據的快速存取。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"B（B+）樹索引：適合於查詢爲主導的場景，避免多次的IO，提高查詢的效率。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"倒排索引：實現單詞到文檔映射關係的最佳實現方式和最有效的索引結構，廣泛用在搜索領域。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"Bitmap：是一種非常簡潔快速的","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}}],"text":"數據結構","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"，他能同時使存儲空間和速度最優化（而不必空間換時間），適合於海量數據的的計算場景。","attrs":{}}]}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"並行與分佈式計算","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"任務切分、分而治之(MR)","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"大規模的數據中，數據存在一定局部性的特徵，利用局部性的原理將海量數據計算的問題分而治之","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":"MR模型是無共享的架構，數據集分佈至各個節點。處理時，每個節點就近讀取本地存儲的數據處理(map)，將處理後的數據進行合併(","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"combine","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":")、排序(","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"shuffle and sort","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":")後再分發(至reduce節點)，避免了大量數據的傳輸，提高了處理效率。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"多進程、多線程並行執行(MPP)","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"並行計算（Parallel Computing）是指同時使用多種計算資源解決計算問題的過程，是提高計算機系統計算速度和處理能力的一種有效手段。","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":"它的基本思想是用多個處理器/進程/線程來協同求解同一問題，即將被求解的問題分解成若干個部分，各部分均由一個獨立的處理機來並行計算。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":"和MR的區別在於，它是基於問題分解的，而不是基於數據分解。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"多維度的可用","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"負載均衡、容災、備份","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"隨着平臺併發量的增大，需要擴容節點進行集羣，利用負載均衡設備進行請求的分發；","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"負載均衡","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"設備通常在提供","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"負載均衡","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"的同時，也提供","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"失效檢測","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"功能；同時爲了提高可用性，需要有容災備份，以防止節點宕機失效帶來的不可用問題；備份有在線的和離線備份，可以根據失效性要求的不同，進行選擇不同備份策略。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"讀寫分離","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"讀寫分離是對數據庫來講的，隨着系統併發量的增大，提高數據訪問可用性的一個重要手段就是寫數據和讀數據進行分離；","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"當然在讀寫分離的同時，需要關注數據的一致性問題；對於一致性的問題，在分佈式的系統CAP定量中，更多的關注於可用性。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"依賴關係","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"平臺中各個模塊之間的關係儘量是低耦合的，可以通過相關的消息組件進行交互，能異步則異步，分清楚數據流轉的主流程和副流程，主副是異步的，比如記錄日誌可以是異步操作的，增加整個系統的可用性。當然在異步處理中，爲了確保數據得到接收或者處理，往往需要確認機制(confirm、ack)。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 但是有些場景中，雖然請求已經得到處理，但是因其他原因(比如網絡不穩定)，確認消息沒有返回，那麼這種情況下需要進行請求的重發，對請求的處理設計因重發因素需要考慮冪等性。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"監控","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":" 監控也是提高整個平臺可用性的一個重要手段，多平臺進行多個維度的監控；模塊在運行時候是透明的，以達到運行期白盒化。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"伸縮","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"拆分","attrs":{}}]},{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":" 拆分包括對業務的拆分和對數據庫的拆分。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"系統的資源總是有限的，一段比較長的業務執行如果是一竿子執行的方式，在大量併發的操作下，這種阻塞的方式，無法有效的及時釋放資源給其他進程執行，這樣系統的吞吐量不高。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 需要把業務進行邏輯的分段，採用異步非阻塞的方式，提高系統的吞吐量。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 隨着數據量和併發量的增加，讀寫分離不能滿足系統併發性能的要求，需要對數據進行切分，包括對數據進行分庫和分表。這種分庫分表的方式，需要增加對數據的路由邏輯支持。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"無狀態","attrs":{}}]},{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 對於系統的伸縮性而言，模塊最好是無狀態的，通過增加節點就可以提高整個的吞吐量。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"優化資源利用","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"系統容量有限","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 系統的容量是有限的，承受的併發量也是有限的，在架構設計時，一定需要考慮流量的控制，防止因意外攻擊或者瞬時併發量的衝擊導致系統崩潰。在設計時增加流控的措施，可考慮對請求進行排隊，超出預期的範圍，可以進行告警或者丟棄。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"原子操作與併發控制","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"對於共享資源的訪問，爲了防止衝突，需要進行併發的控制，同時有些交易需要有事務性來保證交易的一致性，所以在交易系統的設計時，需考慮原子操作和併發控制。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 保證併發控制一些常用高性能手段有，樂觀鎖、Latch、mutex、寫時複製、CAS等；多版本的併發控制MVCC通常是保證一致性的重要手段，這個在數據庫的設計中經常會用到。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"邏輯的不同，採取不一樣策略","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":" 平臺中業務邏輯存在不同的類型，有計算複雜型的，有消耗IO型的，同時就同一種類型而言，不同的業務邏輯消耗的資源數量也是不一樣的，這就需要針對不同的邏輯採取不同的策略。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":" 針對IO型的，可以採取基於事件驅動的異步非阻塞的方式，","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"單線程方式可以減少線程的切換引起的開銷，或者在多線程的情況下采取自旋spin的方式","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}}],"text":"，","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"減少對線程的切換","attrs":{}}]},{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 對於計算型的，充分利用多線程進行操作。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 同一類型的調用方式，不同的業務進行合適的資源分配，設置不同的計算節點數量或者線程數量，對業務進行分流，優先執行優先級別高的業務。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"容錯隔離","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 系統的有些業務模塊在出現錯誤時，爲了減少併發下對正常請求的處理的影響，有時候需要考慮對這些異常狀態的請求進行單獨渠道的處理，","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"甚至暫時自動禁止這些異常的業務模塊","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 有些請求的失敗可能是偶然的暫時的失敗(比如網絡不穩定)，需要進行請求重試的考慮。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"資源釋放","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 系統的資源是有限的，在使用資源時，一定要在最後釋放資源，無論是請求走的是正常路徑還是異常的路徑，以便於資源的及時回收，供其他請求使用。在設計通信的架構時，往往需要考慮超時的控制。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}}]}
🌏【架構師指南】帶你全面認識實現"三高"架構設計方案
{"type":"doc","content":[{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"設計理念","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"空間換時間","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"多級緩存，靜態化","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 客戶端頁面緩存（http header中包含Expires/Cache of Control，last modified(304，server不返回body，客戶端可以繼續用cache，減少流量)，ETag）","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"strong","attrs":{}}],"text":"反向代理緩存","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"strong","attrs":{}}],"text":"應用端的緩存(memcache)","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"strong","attrs":{}}],"text":"內存數據庫","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"strong","attrs":{}}],"text":"Buffer、cache機制（數據庫，中間件等）","attrs":{}}]}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"索引機制","attrs":{}}]},{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":" 哈希、B（B+）樹、倒排、bitmap","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"哈希索引：適合綜合數組的尋址和鏈表的插入特性，可以實現數據的快速存取。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"B（B+）樹索引：適合於查詢爲主導的場景，避免多次的IO，提高查詢的效率。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"倒排索引：實現單詞到文檔映射關係的最佳實現方式和最有效的索引結構，廣泛用在搜索領域。","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"Bitmap：是一種非常簡潔快速的","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}}],"text":"數據結構","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":9}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"，他能同時使存儲空間和速度最優化（而不必空間換時間），適合於海量數據的的計算場景。","attrs":{}}]}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"並行與分佈式計算","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"任務切分、分而治之(MR)","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"大規模的數據中，數據存在一定局部性的特徵，利用局部性的原理將海量數據計算的問題分而治之","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":"MR模型是無共享的架構，數據集分佈至各個節點。處理時，每個節點就近讀取本地存儲的數據處理(map)，將處理後的數據進行合併(","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"combine","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":")、排序(","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"shuffle and sort","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":")後再分發(至reduce節點)，避免了大量數據的傳輸，提高了處理效率。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"多進程、多線程並行執行(MPP)","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"並行計算（Parallel Computing）是指同時使用多種計算資源解決計算問題的過程，是提高計算機系統計算速度和處理能力的一種有效手段。","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":"它的基本思想是用多個處理器/進程/線程來協同求解同一問題，即將被求解的問題分解成若干個部分，各部分均由一個獨立的處理機來並行計算。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":"和MR的區別在於，它是基於問題分解的，而不是基於數據分解。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"多維度的可用","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"負載均衡、容災、備份","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"隨着平臺併發量的增大，需要擴容節點進行集羣，利用負載均衡設備進行請求的分發；","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"負載均衡","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"設備通常在提供","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"負載均衡","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"的同時，也提供","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"失效檢測","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"功能；同時爲了提高可用性，需要有容災備份，以防止節點宕機失效帶來的不可用問題；備份有在線的和離線備份，可以根據失效性要求的不同，進行選擇不同備份策略。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"讀寫分離","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"讀寫分離是對數據庫來講的，隨着系統併發量的增大，提高數據訪問可用性的一個重要手段就是寫數據和讀數據進行分離；","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"當然在讀寫分離的同時，需要關注數據的一致性問題；對於一致性的問題，在分佈式的系統CAP定量中，更多的關注於可用性。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"依賴關係","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"平臺中各個模塊之間的關係儘量是低耦合的，可以通過相關的消息組件進行交互，能異步則異步，分清楚數據流轉的主流程和副流程，主副是異步的，比如記錄日誌可以是異步操作的，增加整個系統的可用性。當然在異步處理中，爲了確保數據得到接收或者處理，往往需要確認機制(confirm、ack)。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 但是有些場景中，雖然請求已經得到處理，但是因其他原因(比如網絡不穩定)，確認消息沒有返回，那麼這種情況下需要進行請求的重發，對請求的處理設計因重發因素需要考慮冪等性。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"監控","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":" 監控也是提高整個平臺可用性的一個重要手段，多平臺進行多個維度的監控；模塊在運行時候是透明的，以達到運行期白盒化。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"伸縮","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"拆分","attrs":{}}]},{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":" 拆分包括對業務的拆分和對數據庫的拆分。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"系統的資源總是有限的，一段比較長的業務執行如果是一竿子執行的方式，在大量併發的操作下，這種阻塞的方式，無法有效的及時釋放資源給其他進程執行，這樣系統的吞吐量不高。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 需要把業務進行邏輯的分段，採用異步非阻塞的方式，提高系統的吞吐量。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 隨着數據量和併發量的增加，讀寫分離不能滿足系統併發性能的要求，需要對數據進行切分，包括對數據進行分庫和分表。這種分庫分表的方式，需要增加對數據的路由邏輯支持。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"無狀態","attrs":{}}]},{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 對於系統的伸縮性而言，模塊最好是無狀態的，通過增加節點就可以提高整個的吞吐量。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":4},"content":[{"type":"text","text":"優化資源利用","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"系統容量有限","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 系統的容量是有限的，承受的併發量也是有限的，在架構設計時，一定需要考慮流量的控制，防止因意外攻擊或者瞬時併發量的衝擊導致系統崩潰。在設計時增加流控的措施，可考慮對請求進行排隊，超出預期的範圍，可以進行告警或者丟棄。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"原子操作與併發控制","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"對於共享資源的訪問，爲了防止衝突，需要進行併發的控制，同時有些交易需要有事務性來保證交易的一致性，所以在交易系統的設計時，需考慮原子操作和併發控制。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 保證併發控制一些常用高性能手段有，樂觀鎖、Latch、mutex、寫時複製、CAS等；多版本的併發控制MVCC通常是保證一致性的重要手段，這個在數據庫的設計中經常會用到。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"邏輯的不同，採取不一樣策略","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":" 平臺中業務邏輯存在不同的類型，有計算複雜型的，有消耗IO型的，同時就同一種類型而言，不同的業務邏輯消耗的資源數量也是不一樣的，這就需要針對不同的邏輯採取不同的策略。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"strong","attrs":{}}],"text":" 針對IO型的，可以採取基於事件驅動的異步非阻塞的方式，","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"單線程方式可以減少線程的切換引起的開銷，或者在多線程的情況下采取自旋spin的方式","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}}],"text":"，","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"減少對線程的切換","attrs":{}}]},{"type":"paragraph","attrs":{"indent":1,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 對於計算型的，充分利用多線程進行操作。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 同一類型的調用方式，不同的業務進行合適的資源分配，設置不同的計算節點數量或者線程數量，對業務進行分流，優先執行優先級別高的業務。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"容錯隔離","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 系統的有些業務模塊在出現錯誤時，爲了減少併發下對正常請求的處理的影響，有時候需要考慮對這些異常狀態的請求進行單獨渠道的處理，","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}},{"type":"color","attrs":{"color":"#F5222D","name":"red"}},{"type":"strong","attrs":{}}],"text":"甚至暫時自動禁止這些異常的業務模塊","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":"。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 有些請求的失敗可能是偶然的暫時的失敗(比如網絡不穩定)，需要進行請求重試的考慮。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":5},"content":[{"type":"text","text":"資源釋放","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}},{"type":"text","marks":[{"type":"size","attrs":{"size":10}}],"text":" 系統的資源是有限的，在使用資源時，一定要在最後釋放資源，無論是請求走的是正常路徑還是異常的路徑，以便於資源的及時回收，供其他請求使用。在設計通信的架構時，往往需要考慮超時的控制。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}}]}
發表評論
所有評論
還沒有人評論，想成為第一個評論的人麼? 請在上方評論欄輸入並且點擊發布.
相關文章