[
https://issues.apache.org/jira/browse/HBASE-12853?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14280440#comment-14280440
]
Michael Segel commented on HBASE-12853:
----------------------------------------
First, lets get away from using the term salted.
Salts do have a specific meaning and its associated with cryptography. While
we're clearly not talking about cryptography, it implies that the prefix is
orthogonal to the data set and the number of salted values is bound by the
width of the prefix.
Using the term bucketing the table would be more appropriate because in this
example, you're assigning a prefix from a round robin approach.
I have to apologize, I don't play with HBase that much these days... my work is
client driven.
With respect to client/server it seems that the delineation between client and
server appears to be a bit different from what I would expect from other
databases. In HBase, the client creates a scan, and then has the hmaster will
manage the scan and return a pointer to the result set?
With respect to the client side code... you're missing the point. You want to
abstract the bucketing from the client. So that the same scan will run against
a bucketed table and an un-bucketed table. The only exposed difference is that
the metadata for the table will specify the number of buckets which defaults to
1 (no bucketing)
> distributed write pattern to replace ad hoc 'salting'
> -----------------------------------------------------
>
> Key: HBASE-12853
> URL: https://issues.apache.org/jira/browse/HBASE-12853
> Project: HBase
> Issue Type: New Feature
> Reporter: Michael Segel
> Priority: Minor
>
> In reviewing HBASE-11682 (Description of Hot Spotting), one of the issues is
> that while 'salting' alleviated regional hot spotting, it increased the
> complexity required to utilize the data.
> Through the use of coprocessors, it should be possible to offer a method
> which distributes the data on write across the cluster and then manages
> reading the data returning a sort ordered result set, abstracting the
> underlying process.
> On table creation, a flag is set to indicate that this is a parallel table.
> On insert in to the table, if the flag is set to true then a prefix is added
> to the key. e.g. <region server#>- or <region server #|| where the region
> server # is an integer between 1 and the number of region servers defined.
> On read (scan) for each region server defined, a separate scan is created
> adding the prefix. Since each scan will be in sort order, its possible to
> strip the prefix and return the lowest value key from each of the subsets.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
