Re: Load balancing ignite get requests

2018-09-11 Thread ezhuravlev
Hi,

Well, it depends on a lot of things - if you have small amount of the data,
which can easily fit in memory on each node, then, you can use Replicated
cache.

On the other hand, if you have quite big dataset, you may consider using
Partitioned cache and executing affinity runs.

Evgenii



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/


Re: Load balancing ignite get requests

2018-08-31 Thread Anirudha Jadhav
Thanks, Dmitry,

we seem to have maxed out the performance we can get out of a single ignite
node on a DL380 / whats the best option to scale out and make all replicas
also serve data over get requests?

-Ani

On Fri, Aug 31, 2018 at 12:34 PM dkarachentsev 
wrote:

> Hi,
>
> get() operation from client always go to the primary node. If you run
> compute task on other nodes, where each will do get() request for that key,
> it will read local value. REPLICATED has many other optimizations, for
> example for SQL queries.
>
> Thanks!
> -Dmitry
>
>
>
> --
> Sent from: http://apache-ignite-users.70518.x6.nabble.com/
>


-- 
Anirudha P. Jadhav


Re: Load balancing ignite get requests

2018-08-31 Thread luqmanahmad
Anirudha, also see [1] for built-in load balancing which could be useful

[1]  AdaptiveLoadProbe

  



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/


Re: Load balancing ignite get requests

2018-08-31 Thread dkarachentsev
Hi,

get() operation from client always go to the primary node. If you run
compute task on other nodes, where each will do get() request for that key,
it will read local value. REPLICATED has many other optimizations, for
example for SQL queries.

Thanks!
-Dmitry



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/


Load balancing ignite get requests

2018-08-29 Thread Anirudha Jadhav
Ignite 2.5 - fully replicated 8 node cluster

When running load tests with 1 unique key, only one of the servers gets all
the requests.

Is there something to be enabled in configuration for balancing load
between all machines.


Thanks
-- 
Anirudha P. Jadhav