You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@impala.apache.org by Nasron Cheong <na...@upsight.com> on 2017/09/07 20:43:12 UTC
Cloudera jdbc driver vs apache hive jdbc driver performance difference
Hi,
I'm seeing a static several second performance difference between the hive
and cloudera jdbc drivers against our Impala installation.
Are there settings that should be tuned on either the driver or the server
that would account for this?
On a related question, is it possible to get PROFILE results from the
cloudera jdbc driver?
Thanks.
Re: Cloudera jdbc driver vs apache hive jdbc driver performance difference
Posted by Nasron Cheong <na...@upsight.com>.
Tried that, no difference. I think for now we'll stick to the hive jdbc
driver.
- Nasron
On Fri, Sep 8, 2017 at 12:09 PM, Alexander Behm <al...@cloudera.com>
wrote:
> Try setting useNativeQuery=1 in the Cloudera JDBC driver, i.e. connect
> with something like this:
>
> jdbc:impala://impaladhost:21050/mydatabase;UseNativeQuery=1;
>
> Not sure if it's possible to get the query profile through JDBC, maybe
> someone else can chime in.
>
> On Fri, Sep 8, 2017 at 7:30 AM, Nasron Cheong <na...@upsight.com>
> wrote:
>
>> Hi Jeszy,
>>
>> The Hive driver appears faster - but it seems to be a static cost during
>> connection, but I need to confirm that.
>>
>> Additionally, is there a way to easily tell from a jdbc query execution
>> which impala server was used, and/or what the query id is?
>>
>> We do not have CM.
>>
>> Thanks.
>>
>> - Nasron
>>
>> On Thu, Sep 7, 2017 at 4:48 PM, Jeszy <je...@gmail.com> wrote:
>>
>>> Hello,
>>>
>>> It's not possible unfortunately to get the profile directly through JDBC.
>>> Can you please clarify which driver is faster? I expect Cloudera's
>>> Impala (and not hive, that's untested) JDBC driver to be faster than
>>> the Apache Hive driver. Can you attach profiles for both runs? Even
>>> though you can't get them via JDBC, they are available through the web
>>> UI / CM (if you have that).
>>>
>>> Thanks!
>>>
>>> On 7 September 2017 at 22:43, Nasron Cheong <na...@upsight.com>
>>> wrote:
>>> > Hi,
>>> >
>>> > I'm seeing a static several second performance difference between the
>>> hive
>>> > and cloudera jdbc drivers against our Impala installation.
>>> >
>>> > Are there settings that should be tuned on either the driver or the
>>> server
>>> > that would account for this?
>>> >
>>> > On a related question, is it possible to get PROFILE results from the
>>> > cloudera jdbc driver?
>>> >
>>> > Thanks.
>>>
>>
>>
>
Re: Cloudera jdbc driver vs apache hive jdbc driver performance difference
Posted by Alexander Behm <al...@cloudera.com>.
Try setting useNativeQuery=1 in the Cloudera JDBC driver, i.e. connect with
something like this:
jdbc:impala://impaladhost:21050/mydatabase;UseNativeQuery=1;
Not sure if it's possible to get the query profile through JDBC, maybe
someone else can chime in.
On Fri, Sep 8, 2017 at 7:30 AM, Nasron Cheong <na...@upsight.com>
wrote:
> Hi Jeszy,
>
> The Hive driver appears faster - but it seems to be a static cost during
> connection, but I need to confirm that.
>
> Additionally, is there a way to easily tell from a jdbc query execution
> which impala server was used, and/or what the query id is?
>
> We do not have CM.
>
> Thanks.
>
> - Nasron
>
> On Thu, Sep 7, 2017 at 4:48 PM, Jeszy <je...@gmail.com> wrote:
>
>> Hello,
>>
>> It's not possible unfortunately to get the profile directly through JDBC.
>> Can you please clarify which driver is faster? I expect Cloudera's
>> Impala (and not hive, that's untested) JDBC driver to be faster than
>> the Apache Hive driver. Can you attach profiles for both runs? Even
>> though you can't get them via JDBC, they are available through the web
>> UI / CM (if you have that).
>>
>> Thanks!
>>
>> On 7 September 2017 at 22:43, Nasron Cheong <na...@upsight.com>
>> wrote:
>> > Hi,
>> >
>> > I'm seeing a static several second performance difference between the
>> hive
>> > and cloudera jdbc drivers against our Impala installation.
>> >
>> > Are there settings that should be tuned on either the driver or the
>> server
>> > that would account for this?
>> >
>> > On a related question, is it possible to get PROFILE results from the
>> > cloudera jdbc driver?
>> >
>> > Thanks.
>>
>
>
Re: Cloudera jdbc driver vs apache hive jdbc driver performance difference
Posted by Nasron Cheong <na...@upsight.com>.
Hi Jeszy,
The Hive driver appears faster - but it seems to be a static cost during
connection, but I need to confirm that.
Additionally, is there a way to easily tell from a jdbc query execution
which impala server was used, and/or what the query id is?
We do not have CM.
Thanks.
- Nasron
On Thu, Sep 7, 2017 at 4:48 PM, Jeszy <je...@gmail.com> wrote:
> Hello,
>
> It's not possible unfortunately to get the profile directly through JDBC.
> Can you please clarify which driver is faster? I expect Cloudera's
> Impala (and not hive, that's untested) JDBC driver to be faster than
> the Apache Hive driver. Can you attach profiles for both runs? Even
> though you can't get them via JDBC, they are available through the web
> UI / CM (if you have that).
>
> Thanks!
>
> On 7 September 2017 at 22:43, Nasron Cheong <na...@upsight.com>
> wrote:
> > Hi,
> >
> > I'm seeing a static several second performance difference between the
> hive
> > and cloudera jdbc drivers against our Impala installation.
> >
> > Are there settings that should be tuned on either the driver or the
> server
> > that would account for this?
> >
> > On a related question, is it possible to get PROFILE results from the
> > cloudera jdbc driver?
> >
> > Thanks.
>
Re: Cloudera jdbc driver vs apache hive jdbc driver performance difference
Posted by Jeszy <je...@gmail.com>.
Hello,
It's not possible unfortunately to get the profile directly through JDBC.
Can you please clarify which driver is faster? I expect Cloudera's
Impala (and not hive, that's untested) JDBC driver to be faster than
the Apache Hive driver. Can you attach profiles for both runs? Even
though you can't get them via JDBC, they are available through the web
UI / CM (if you have that).
Thanks!
On 7 September 2017 at 22:43, Nasron Cheong <na...@upsight.com> wrote:
> Hi,
>
> I'm seeing a static several second performance difference between the hive
> and cloudera jdbc drivers against our Impala installation.
>
> Are there settings that should be tuned on either the driver or the server
> that would account for this?
>
> On a related question, is it possible to get PROFILE results from the
> cloudera jdbc driver?
>
> Thanks.