wiki:jazz/NoSQL

Version 18 (modified by jazz, 14 years ago) (diff)

--

NoSQL

Histroy

Definition

Primer

Papers and Websites

趨勢分析 Trend Observation

  • 當然縱使 MongoDB 與 CouchDB 已經有很高的排行榜,跟 SQLite 比起來,還是差很多。 從 Google Trends 的搜尋趨勢可以看到 SQLite > Google Gears > CouchDB 的趨勢,象徵著 distributed database 的影響力尚未普及。 (2009-09-23)

Comparison

Distributed Not Distributed (responsibility on client)
* Amazon Dynamo
* Amazon S3
* Scalaris
* Voldemort
* CouchDb? (thru Lounge)
* Riak
* MongoDb? (in alpha)
* BigTable
* Cassandra
* HyperTable?
* HBase
* Redis
* Tokyo Tyrant
* MemcacheDb?
* Amazon SimpleDb?

Open Source Projects

  • gears-dblib - A simple abstraction on top of the Database object in Gears (2009-09-23)
  • orient - NoSQL document database light, portable and fast. Supports ACID Tx, Indexes, asynch queries, SQL layer, clustering, etc (2009-09-23)

NoSQL : HBase

  • 11:10–12:00 Upcoming improvements for HBase - Andrew Purtell (Trend Micro) (2010-04-24)
    • Big Data -> Medium Data 都需要
    • Cloud Computing - Scale Free
    • Disk Seek time remains nearly constant -> Index(B-Tree), Seek (RMDB) 慢!!
    • No distributed transactions, no complex locking, no waits or deadlocks
    • 不要用 Spreadsheet 的想法看待 HBase, 或許可以用 Tag 的想法去看待它。
    • HBase 跟 BigTable 都是 CP 架構(注重 Consistancy 與 Partition Tolerance,根據 CAP Theorem 因此無法確保 Avaibility,寧可服務中斷也要資料正確!!)
    • HDFS-200 (working append) 將在 HBase 0.20.5 加入支援資料持續遞增的功能。
    • ACID ? - atomicity, consistency, isolation, durability
    • 新功能:
      • 跨資料中心備份 - 透過 Log Ship HBASE-1295 : Multi data center replication
      • 安全性強化 - 支援 authentication, authorization,Yahoo! 寫了很多新的安全性支援,包括 Kerberos 認證、Data isolation at the HDFS layer、Secure RPC。因此必須新增角色來作存取控管(Access Control Role)
      • Coprocessor - 靈感來自於 BigTable 的新功能 Coprocessor,加入 RegionObservor (需要再花點時間看清楚用途!!)

NoSQL : Cassandra

  • 09:00–09:50 nosql cassandra - Gasol (Pixnet) (2010-04-24)
    • http://cassandra.apache.org/
    • 具備副本機制,優先存在記憶體中,後續寫入 commit log 中。採取完全平等的分散式架構,沒有 Hadoop NameNode 單點失效問題(Single Point of Failure)
    • CAP theorem - Consistency (一致性), Availibility (可用性), Partition Tolerance (容錯性)三個勢必要犧牲其中一個!!

NoSQL : MongoDB

NoSQL : CouchDB

NoSQL : Voldemort

  • Voldemort - a distributed key-value storage system (2009-09-18)
    • used at LinkedIn for certain high-scalability storage problems

NoSQL : Redis

Use Case

Attachments (27)