[jira] [Updated] (IGNITE-17267) Preparing a checkpoint for a DeltaFilePageStore

2022-06-29 Thread Kirill Tkalenko (Jira)
Title: Message Title


 
 
 
 

 
 
 

 
   
 Kirill Tkalenko updated an issue  
 

  
 
 
 
 

 
 
  
 
 
 
 

 
 Ignite /  IGNITE-17267  
 
 
  Preparing a checkpoint for a DeltaFilePageStore   
 

  
 
 
 
 

 
Change By: 
 Kirill Tkalenko  
 

  
 
 
 
 

 
 To implement the *DeltaFilePageStore*, we need to prepare a checkpoint for their creation.What should be done:# Sort dirty pages (*FullPageId*) by *groupId*, *partitionId*, *pageIdx*;# In several threads, write dirty pages in batches for each partition and know the number of these pages.What can we look at:* *org.apache.ignite.internal.pagememory.persistence.checkpoint.CheckpointWorkflow#beginCheckpoint** *org.apache.ignite.internal.pagememory.persistence.checkpoint.CheckpointWorkflow#splitAndSortCheckpointPagesIfNeeded** *org.apache.ignite.internal.pagememory.persistence.checkpoint.Checkpointer#writePages*Some implementation notes:* We can replace *CheckpointWorkflow#parallelSortThreshold* with a constant at 40k, according to the  article from 1  [paper|https://www . researchgate.net/publication/331742843_Threshold_Analysis_and_Comparison_of_Sequential_and_Parallel_Divide_and_Conquer_Sorting_Algorithms].  
 

  
 
 
 
 

 
 
 

 
 
 Add Comment  
 

  
 

  
 
 
 
  
 

  
 
 
 
 

 
 This message was sent by Atlassian Jira (v8.20.10#820010-sha1:ace47f9)  
 
 

 
   
 

  
 

[jira] [Updated] (IGNITE-17267) Preparing a checkpoint for a DeltaFilePageStore

2022-06-29 Thread Kirill Tkalenko (Jira)
Title: Message Title


 
 
 
 

 
 
 

 
   
 Kirill Tkalenko updated an issue  
 

  
 
 
 
 

 
 
  
 
 
 
 

 
 Ignite /  IGNITE-17267  
 
 
  Preparing a checkpoint for a DeltaFilePageStore   
 

  
 
 
 
 

 
Change By: 
 Kirill Tkalenko  
 

  
 
 
 
 

 
 To implement the *DeltaFilePageStore*, we need to prepare a checkpoint for their creation.What should be done:# Sort dirty pages (*FullPageId*) by *groupId*, *partitionId*, *pageIdx*;# In several threads, write dirty pages in batches for each partition and know the number of these pages.What can we look at:* *org.apache.ignite.internal.pagememory.persistence.checkpoint.CheckpointWorkflow#beginCheckpoint** *org.apache.ignite.internal.pagememory.persistence.checkpoint.CheckpointWorkflow#splitAndSortCheckpointPagesIfNeeded** *org.apache.ignite.internal.pagememory.persistence.checkpoint.Checkpointer#writePages* Some implementation notes:* We can replace *CheckpointWorkflow#parallelSortThreshold* with a constant at 40k, according to the article from 1.  
 

  
 
 
 
 

 
 
 

 
 
 Add Comment  
 

  
 

  
 
 
 
  
 

  
 
 
 
 

 
 This message was sent by Atlassian Jira (v8.20.10#820010-sha1:ace47f9)