Enhancing the performance of Readonly transactions using speculation

1
Performance Enhancement of Read-only
Transactions Using Speculative Locking Protocol
T. Ragunathan and Dr. P. K. Reddy
Example

• Data currency of ROTs improved
• significantly over SI.

• Problems in Processing Read-only
• Transactions (ROTs)
• Correctness ( Serializability
  criteria)
• Data Currency ( Transactions receive old
• values)
• Performance ( Throughput)
• The performance of 2PL degrades
• with data contention.
• Example

• Simulation Parameters and Values.

• Requires manageable extra
• processing resources ( 1.4 times
• of memory is required).

• Simulation Results
• Throughput performance is
• Improved significantly over SI
• and 2PL.

• Snapshot Isolation(SI)-based
• protocols violate serializability
• criteria and provide less data
• currency to ROTs.
• Example

• Conclusions
• Proposed speculation based approach
• improves the performance of ROTs
• significantly over other protocols.
• It does not violate serializability
  criteria.
• It processes ROTs without data currency
• problems.
• The experimental results show that it
  requires
• manageable ( 1.4t imes) extra processing
  
• resources.

• Proposed Speculation based approach
• (SLR)
• ROTs are processed with speculation.
• Whenever a UT produces after-image, ROTs
• access both before- and after-images and
  start
• speculative executions.

Center for Data Engineering International
Institute of Information Technology,
Hyderabad http//cde.iiit.ac.in
2
Asynchronous Speculative Locking Protocol for
improving the performance of Read-only
Transactions - T. Ragunathan and Dr. P. K.
Reddy
Example
Example

• The performance of 2PL degrades
• with data contention.

• Snapshot Isolation(SI)-based
• protocols violate serializability
• criteria and provide less data
• currency to ROTs.

• Synchronous speculation based
• approach (SLR)
• Improves performance and data currency.
• ROTs wait for lock conversion ( EW-lock
  in to
• SPW-lock)

• Requires manageable extra
• processing resources ( 1.5 times
• of memory is required).

• Simulation Parameters and Values.

• Proposed Asynchronous
• speculative locking protocol
• (ASLR)
• ROTs do not wait for lock conversion
  and can start
• executions with the available
  before-images.
• However ROTs perform speculative
  executions
• whenever UTs produce after-images.
• When an ROT commits, it commits with the
• appropriate speculative execution.
• Performance is better than SLR.

• Conclusions
• Proposed asynchronous speculation based
  approach
• improves the performance of ROTs
  significantly over
• other protocols.
• It does not violate serializability
  criteria.
• The experimental results show that it
  requires
• manageable ( 1.5 times) extra
  processing
• resources.

• Simulation Results

• Throughput performance is
• Improved significantly over SI,
• 2PL and SLR.

Center for Data Engineering International
Institute of Information Technology,
Hyderabad http//cde.iiit.ac.in