You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@accumulo.apache.org by Diego Woitasen <di...@woitasen.com.ar> on 2014/04/23 14:24:27 UTC
Memory tables?
Hi,
I was looking at the documentation but I couldn't find the answer. Is
there a way to configure tables to be served from memory?
I've been thinking about different options, not sure what's available:
1- Memory tables clustered? So, if a node dies we don't lost the data.
2- Backed to disk, but always served from memory.
3- Is it possible to configure it table by table?
Because for some tables we want high throughtput, so if we can set a
flag for a table to be served from memory would be great.
Regards,
Diego
--
Diego Woitasen
- Linux and Open Source solutions architect
- DevOps Engineer, Infrastructure developer
http://www.woitasen.com.ar
Re: Memory tables?
Posted by Josh Elser <jo...@gmail.com>.
I made https://issues.apache.org/jira/browse/ACCUMULO-1015 quite a while
back, but I don't know of anyone looking at implementing it yet.
On 4/23/14, 8:29 AM, Eric Newton wrote:
> The only option available today is to increase the size of the data
> block cache and enable it for your table. Assuming your table is
> accessed frequently, it will be served almost always from the cache.
> This is done for the metadata table, for example.
>
> Of course, it will be backed by HDFS and replication.
>
> -Eric
>
>
>
> On Wed, Apr 23, 2014 at 8:24 AM, Diego Woitasen <diego@woitasen.com.ar
> <ma...@woitasen.com.ar>> wrote:
>
> Hi,
> I was looking at the documentation but I couldn't find the answer. Is
> there a way to configure tables to be served from memory?
>
> I've been thinking about different options, not sure what's available:
>
> 1- Memory tables clustered? So, if a node dies we don't lost the data.
> 2- Backed to disk, but always served from memory.
> 3- Is it possible to configure it table by table?
>
> Because for some tables we want high throughtput, so if we can set a
> flag for a table to be served from memory would be great.
>
> Regards,
> Diego
>
> --
> Diego Woitasen
> - Linux and Open Source solutions architect
> - DevOps Engineer, Infrastructure developer
> http://www.woitasen.com.ar
>
>
Re: Memory tables?
Posted by Eric Newton <er...@gmail.com>.
The only option available today is to increase the size of the data block
cache and enable it for your table. Assuming your table is accessed
frequently, it will be served almost always from the cache. This is done
for the metadata table, for example.
Of course, it will be backed by HDFS and replication.
-Eric
On Wed, Apr 23, 2014 at 8:24 AM, Diego Woitasen <di...@woitasen.com.ar>wrote:
> Hi,
> I was looking at the documentation but I couldn't find the answer. Is
> there a way to configure tables to be served from memory?
>
> I've been thinking about different options, not sure what's available:
>
> 1- Memory tables clustered? So, if a node dies we don't lost the data.
> 2- Backed to disk, but always served from memory.
> 3- Is it possible to configure it table by table?
>
> Because for some tables we want high throughtput, so if we can set a
> flag for a table to be served from memory would be great.
>
> Regards,
> Diego
>
> --
> Diego Woitasen
> - Linux and Open Source solutions architect
> - DevOps Engineer, Infrastructure developer
> http://www.woitasen.com.ar
>
Re: Memory tables?
Posted by David Medinets <da...@gmail.com>.
Are you seeing a performance issue that you need to resolve?
On Wed, Apr 23, 2014 at 8:24 AM, Diego Woitasen <di...@woitasen.com.ar>wrote:
> Hi,
> I was looking at the documentation but I couldn't find the answer. Is
> there a way to configure tables to be served from memory?
>
> I've been thinking about different options, not sure what's available:
>
> 1- Memory tables clustered? So, if a node dies we don't lost the data.
> 2- Backed to disk, but always served from memory.
> 3- Is it possible to configure it table by table?
>
> Because for some tables we want high throughtput, so if we can set a
> flag for a table to be served from memory would be great.
>
> Regards,
> Diego
>
> --
> Diego Woitasen
> - Linux and Open Source solutions architect
> - DevOps Engineer, Infrastructure developer
> http://www.woitasen.com.ar
>