Hello!

I doubt that you will have speed-up here compared to just using JDBC
(possibly with randomized endpoints list).

Regards,
-- 
Ilya Kasnacheev


ср, 27 февр. 2019 г. в 14:52, 李玉珏@163 <[email protected]>:

> The main consideration is that using JDBC interface, the existing code
> modification workload is small.
> 在 2019/2/27 下午5:31, Stephen Darlington 写道:
>
> If you’re already using Ignite-specific APIs (IgniteCallable), why not use
> the other Ignite-native APIs for reading/writing/processing data? That way
> you can use affinity functions for load balancing where it makes sense and
> Ignite’s normal load balancing processing for general compute tasks.
>
> Regards,
> Stephen
>
> On 27 Feb 2019, at 06:00, 李玉珏@163 <[email protected]> wrote:
>
> Hi,
> Since JDBC can't achieve multi-endpoint load balancing, we want to use
> affinityCall (...) mechanism to achieve load balancing, that is, to obtain
> and use JDBC Connection in IgniteCallable implementation.
> How to efficiently access and use JDBC Connection?
>
> -------- 转发的消息 --------
> 主题: Re: On Multiple Endpoints Mode of JDBC Driver
> 日期: Tue, 26 Feb 2019 14:53:17 -0800
> 发件人: Denis Magda <[email protected]> <[email protected]>
> 回复地址: [email protected]
> 收件人: dev <[email protected]> <[email protected]>
>
> Hello,
>
> You provide a list of IP addresses for the sake of high-availability - if
> one of the servers goes down then the client will reconnect to the next IP
> automatically. There is no any load balancing in place presently. But! In
> the next Ignite version, we're planning to roll out partition-awareness
> support - the client will send a request to the nodes who hold the data
> needed for the request.
>
> -
> Denis
>
>
> On Tue, Feb 26, 2019 at 2:48 PM 李玉珏 <[email protected]>
> <[email protected]> wrote:
>
> Hi,
>
> Does have load balancing function in Multiple Endpoints mode of JDBC
> driver?For example, "jdbc: ignite: thin://192.168.0.50:101,192.188.5.40:101, 
> 192.168.10.230:101"
> If not, will one node become the bottleneck of the whole system?
>
>
>
>
>
>

Reply via email to