[
https://issues.apache.org/jira/browse/OMID-131?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16817230#comment-16817230
]
Lars Hofhansl commented on OMID-131:
------------------------------------
For the record I do not like that part of Thephra's design, because it is very
difficult to know when we no longer need to to send the list of uncommitted
transaction timestamps - it's hard to guarantee that all their data has been
collected. So that list could grow.
But that's not needed with Omid as long there still at least one shadow column
per row, right?
Why would the scan cost be increased? Because you'd need to find the one shadow
column? Or because the shadow columns are no longer adjacent to the data
columns? Or because the timestamp logic is more complex?
It seems that:
* Tephra's has a larger per transaction cost (both read and write), but no per
row or column cost.
* Omid's has no (or negligible) per transaction cost, and a small per column
cost (also both read and write).
> Enhance row level conlict detection to reduce client memory footprint.
> ----------------------------------------------------------------------
>
> Key: OMID-131
> URL: https://issues.apache.org/jira/browse/OMID-131
> Project: Apache Omid
> Issue Type: Improvement
> Affects Versions: 1.0.1
> Reporter: Yonatan Gottesman
> Assignee: Yonatan Gottesman
> Priority: Major
> Time Spent: 20m
> Remaining Estimate: 0h
>
> Currently even if omid is working in row level conflict detection, all the
> cells are kept by the client until after commit for writing shadow cells or
> removing them after abort.
> Instead the client should store only the rows or only the family.
> If we choose to store only the row, the client memory footprint will be
> smaller, but if a region server wants to query a shadow cell for a cell it
> might have to query a different region.
> On the other hand client can store the famelies of a writeset. This will have
> a bigger footprint if we have lots of families but if a region queries a
> shadow cell it will be on the same region.
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
