Allgemein §

What causes Deadlocks? §

rir:LockUnlockHow to manage Deadlocks? §

• Prevent:
  • Lock records at the beginning of a transaction
  • Use two-phase locking protocol
    • growing phase
    • shrinking phase
• Resolve:
  • Allow them to occur

rir:GitBranchAlternative to Locking: MultiVersioning §

• Logging of steps:

1. Read
2. Manipulate
3. Commit
   • Only works if manipulated file = same version as current file

rir:ArrowLeftRightSequential Consistency §

• = Interpreting parallell manipulations as sequential

rir:CloudBig Data §

• Volume
  • Large amounts
• Veriety
  • many different files
• Velovity
  • a lot of new data
• Veracity
• Value
  • worh

rir:PencilRulerSchema on Read Vs. Schema on Write §

• Schema on Write (traditional)
  • Design Schema before writing data
• Schema on Read (big data)
  • Design Schema after writing data

rir:DatabaseNot Only SQL (NOSQL) §

• Do not need to understand relations before making changes
• -> Better performance, because relations are optional
• -> Data that’s not defined in the DB can still be saved
• data storage/retrieval technologies natural for cloud environment
• not ACID compliant
• BASE
  • basically available
  • soft state
  • eventually consistent

rir:PriceTag2NOSQL Classifications §

• Key-value stores
  • simple key-value map
• Document stores
  • Values -> Documents
• Wide-Column stores
  • Rows and columns
• Graph

rir:HonorOfKingsHADOOP §

• = open source implementation framework of MapReduce
• How to analyze data if it is stored across multiple computers (cloud)?

rir:FileCloudHadoop Distributed File System (HDFS) §

• File system for data stored in cloud
• Data -> broken into Blocks -> stored in nodes -> stored in clusters
• Cluster:
  • consists of NameNode (master server) and DataNodes (slaves)
• Overall controll through YARN
• No updates just appending

rir:MapMapReduce Design Pattern §

• Requirement for HADOOP
• Enables parallelization of data storage across multiple servers

1. Map
   • Divide tasks so that multiple nodes can work on it
2. Reduce
   • integrate two results into one
   • repeat

Important: §

• Map-Reduce
• Schema read/write

Resources: §

• NOSQL https://www.youtube.com/watch?v=BgQFJ_UNIgw