Distributed Transactions

Question 1

Shared Memory Properties

Answer 1

• Nodes share both RAM and disk
• Dozens to hundreds of processors
• Easy to use and program
• But very expensive to scale
  -> (Cheaper) RDMA may change the picture

Question 2

Shared Disk Properties

Answer 2

• All nodes access the same disks
  -> Found in the largest “single-box” (non-cluster) multiprocessors
• Traditionally: Hard to scale past a certain point
  -> Contention on storage bandwidth
  -> Existing deployments typically have <10 machines
• Scale to Cloud

Question 3

Shared Nothing

Answer 3

• Cluster of machines on high-speed network
• Each machine has its own memory and disk
• Most scalable architecture
• Most difficult to administer and tune

Question 4

Ways to implement Distributed Locking

Answer 4

• Centralized: One site does all locking
• Primary Copy: All locking for an object done at the primary copy site for this object (reading requires access to locking site as well as site where the object is stored)
• Fully distributed: Locking for a copy done at site where the copy is stored (locks at all sites while writing an object)

Question 5

Distributed Deadlock Detection

Answer 5

• Each site maintains a local waits-for graph
• A global deadlock might exist even with acyclic local graphs
• Three solutions:
  -> Centralized: send all local graphs to one site
  -> Hierarchical: organize sites into a hierarchy and send local graphs to parent in the hierarchy
  -> Timeout: abort transaction if it waits too long