You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@hbase.apache.org by Manthosh Kumar T <ma...@gmail.com> on 2014/04/03 14:59:43 UTC
Geographically distant client
Hi All,
I have around 20-30 geographically distant clients that need to
write data to a centralized HBase server. I have dedicated VPN for the
communication and hence bandwidth won't be a issue. is it a good idea to
make the clients directly send data to the centralized server?. Or a
geographically distributed Hadoop cluster is more efficient for this
scenario? Has anybody come across such use case?. Need suggestions on the
viability and efficiency of the setup to follow
--
Cheers,
Manthosh Kumar. T
Re: Geographically distant client
Posted by Manthosh Kumar T <ma...@gmail.com>.
Hi Jean,
Thanks. I might sound a bit lame. Can you just elaborate on the
gateway part?. What is the best practice?
On 3 April 2014 19:30, Jean-Marc Spaggiari <je...@spaggiari.org> wrote:
> Efficient? Probably not ;) But it's like if you are connecting your client
> app to a webserver local to the cluster and then the webserver connects to
> the cluster.
>
> I don't like the idea of having the cluster accessible from the outsite and
> usuall prefer to have kind of a gateway, but that's your call.
>
> You efficiency will mainly depend on the RPC calls you are doing. If you
> send or retreive big bunch of data at a time should not be that bad. But if
> you get cells one by one and send edits one by one, might not be very good.
>
> JM
>
>
> 2014-04-03 9:57 GMT-04:00 Manthosh Kumar T <ma...@gmail.com>:
>
> > Is that a good idea even if I don't have a VPN??. Will it be efficient
> in a
> > fairly good connection?
> >
> >
> > On 3 April 2014 19:25, Jean-Marc Spaggiari <je...@spaggiari.org>
> > wrote:
> >
> > > I will say, remote client connecting to a cluster is fine. But a
> cluster
> > > spread over multiple physical sites is not at all a good idea.
> > >
> > >
> > > 2014-04-03 9:28 GMT-04:00 Manthosh Kumar T <ma...@gmail.com>:
> > >
> > > > Pardon me if I miss anything, like any network issues
> > > >
> > > >
> > > > On 3 April 2014 18:57, Manthosh Kumar T <ma...@gmail.com> wrote:
> > > >
> > > > > I mean directly interacting with the remote zookeeper. Say I'm able
> > to
> > > > > access the zk server and hbase server externally.
> > > > >
> > > > >
> > > > > On 3 April 2014 18:54, Ted Yu <yu...@gmail.com> wrote:
> > > > >
> > > > >> Regions hosted by the server may be moved to other servers.
> > > > >>
> > > > >> Can you clarify what you meant by directly writing to the server ?
> > > > >>
> > > > >> Thanks
> > > > >>
> > > > >> On Apr 3, 2014, at 5:59 AM, Manthosh Kumar T <ma...@gmail.com>
> > > > wrote:
> > > > >>
> > > > >> > Hi All,
> > > > >> > I have around 20-30 geographically distant clients that
> > need
> > > > to
> > > > >> > write data to a centralized HBase server. I have dedicated VPN
> for
> > > the
> > > > >> > communication and hence bandwidth won't be a issue. is it a good
> > > idea
> > > > to
> > > > >> > make the clients directly send data to the centralized server?.
> > Or a
> > > > >> > geographically distributed Hadoop cluster is more efficient for
> > this
> > > > >> > scenario? Has anybody come across such use case?. Need
> suggestions
> > > on
> > > > >> the
> > > > >> > viability and efficiency of the setup to follow
> > > > >> >
> > > > >> > --
> > > > >> > Cheers,
> > > > >> > Manthosh Kumar. T
> > > > >>
> > > > >
> > > > >
> > > > >
> > > > > --
> > > > > Cheers,
> > > > > Manthosh Kumar. T
> > > > >
> > > > >
> > > >
> > > >
> > > > --
> > > > Cheers,
> > > > Manthosh Kumar. T
> > > >
> > >
> >
> >
> >
> > --
> > Cheers,
> > Manthosh Kumar. T
> >
>
--
Cheers,
Manthosh Kumar. T
Re: Geographically distant client
Posted by Jean-Marc Spaggiari <je...@spaggiari.org>.
Efficient? Probably not ;) But it's like if you are connecting your client
app to a webserver local to the cluster and then the webserver connects to
the cluster.
I don't like the idea of having the cluster accessible from the outsite and
usuall prefer to have kind of a gateway, but that's your call.
You efficiency will mainly depend on the RPC calls you are doing. If you
send or retreive big bunch of data at a time should not be that bad. But if
you get cells one by one and send edits one by one, might not be very good.
JM
2014-04-03 9:57 GMT-04:00 Manthosh Kumar T <ma...@gmail.com>:
> Is that a good idea even if I don't have a VPN??. Will it be efficient in a
> fairly good connection?
>
>
> On 3 April 2014 19:25, Jean-Marc Spaggiari <je...@spaggiari.org>
> wrote:
>
> > I will say, remote client connecting to a cluster is fine. But a cluster
> > spread over multiple physical sites is not at all a good idea.
> >
> >
> > 2014-04-03 9:28 GMT-04:00 Manthosh Kumar T <ma...@gmail.com>:
> >
> > > Pardon me if I miss anything, like any network issues
> > >
> > >
> > > On 3 April 2014 18:57, Manthosh Kumar T <ma...@gmail.com> wrote:
> > >
> > > > I mean directly interacting with the remote zookeeper. Say I'm able
> to
> > > > access the zk server and hbase server externally.
> > > >
> > > >
> > > > On 3 April 2014 18:54, Ted Yu <yu...@gmail.com> wrote:
> > > >
> > > >> Regions hosted by the server may be moved to other servers.
> > > >>
> > > >> Can you clarify what you meant by directly writing to the server ?
> > > >>
> > > >> Thanks
> > > >>
> > > >> On Apr 3, 2014, at 5:59 AM, Manthosh Kumar T <ma...@gmail.com>
> > > wrote:
> > > >>
> > > >> > Hi All,
> > > >> > I have around 20-30 geographically distant clients that
> need
> > > to
> > > >> > write data to a centralized HBase server. I have dedicated VPN for
> > the
> > > >> > communication and hence bandwidth won't be a issue. is it a good
> > idea
> > > to
> > > >> > make the clients directly send data to the centralized server?.
> Or a
> > > >> > geographically distributed Hadoop cluster is more efficient for
> this
> > > >> > scenario? Has anybody come across such use case?. Need suggestions
> > on
> > > >> the
> > > >> > viability and efficiency of the setup to follow
> > > >> >
> > > >> > --
> > > >> > Cheers,
> > > >> > Manthosh Kumar. T
> > > >>
> > > >
> > > >
> > > >
> > > > --
> > > > Cheers,
> > > > Manthosh Kumar. T
> > > >
> > > >
> > >
> > >
> > > --
> > > Cheers,
> > > Manthosh Kumar. T
> > >
> >
>
>
>
> --
> Cheers,
> Manthosh Kumar. T
>
Re: Geographically distant client
Posted by Manthosh Kumar T <ma...@gmail.com>.
Is that a good idea even if I don't have a VPN??. Will it be efficient in a
fairly good connection?
On 3 April 2014 19:25, Jean-Marc Spaggiari <je...@spaggiari.org> wrote:
> I will say, remote client connecting to a cluster is fine. But a cluster
> spread over multiple physical sites is not at all a good idea.
>
>
> 2014-04-03 9:28 GMT-04:00 Manthosh Kumar T <ma...@gmail.com>:
>
> > Pardon me if I miss anything, like any network issues
> >
> >
> > On 3 April 2014 18:57, Manthosh Kumar T <ma...@gmail.com> wrote:
> >
> > > I mean directly interacting with the remote zookeeper. Say I'm able to
> > > access the zk server and hbase server externally.
> > >
> > >
> > > On 3 April 2014 18:54, Ted Yu <yu...@gmail.com> wrote:
> > >
> > >> Regions hosted by the server may be moved to other servers.
> > >>
> > >> Can you clarify what you meant by directly writing to the server ?
> > >>
> > >> Thanks
> > >>
> > >> On Apr 3, 2014, at 5:59 AM, Manthosh Kumar T <ma...@gmail.com>
> > wrote:
> > >>
> > >> > Hi All,
> > >> > I have around 20-30 geographically distant clients that need
> > to
> > >> > write data to a centralized HBase server. I have dedicated VPN for
> the
> > >> > communication and hence bandwidth won't be a issue. is it a good
> idea
> > to
> > >> > make the clients directly send data to the centralized server?. Or a
> > >> > geographically distributed Hadoop cluster is more efficient for this
> > >> > scenario? Has anybody come across such use case?. Need suggestions
> on
> > >> the
> > >> > viability and efficiency of the setup to follow
> > >> >
> > >> > --
> > >> > Cheers,
> > >> > Manthosh Kumar. T
> > >>
> > >
> > >
> > >
> > > --
> > > Cheers,
> > > Manthosh Kumar. T
> > >
> > >
> >
> >
> > --
> > Cheers,
> > Manthosh Kumar. T
> >
>
--
Cheers,
Manthosh Kumar. T
Re: Geographically distant client
Posted by Jean-Marc Spaggiari <je...@spaggiari.org>.
I will say, remote client connecting to a cluster is fine. But a cluster
spread over multiple physical sites is not at all a good idea.
2014-04-03 9:28 GMT-04:00 Manthosh Kumar T <ma...@gmail.com>:
> Pardon me if I miss anything, like any network issues
>
>
> On 3 April 2014 18:57, Manthosh Kumar T <ma...@gmail.com> wrote:
>
> > I mean directly interacting with the remote zookeeper. Say I'm able to
> > access the zk server and hbase server externally.
> >
> >
> > On 3 April 2014 18:54, Ted Yu <yu...@gmail.com> wrote:
> >
> >> Regions hosted by the server may be moved to other servers.
> >>
> >> Can you clarify what you meant by directly writing to the server ?
> >>
> >> Thanks
> >>
> >> On Apr 3, 2014, at 5:59 AM, Manthosh Kumar T <ma...@gmail.com>
> wrote:
> >>
> >> > Hi All,
> >> > I have around 20-30 geographically distant clients that need
> to
> >> > write data to a centralized HBase server. I have dedicated VPN for the
> >> > communication and hence bandwidth won't be a issue. is it a good idea
> to
> >> > make the clients directly send data to the centralized server?. Or a
> >> > geographically distributed Hadoop cluster is more efficient for this
> >> > scenario? Has anybody come across such use case?. Need suggestions on
> >> the
> >> > viability and efficiency of the setup to follow
> >> >
> >> > --
> >> > Cheers,
> >> > Manthosh Kumar. T
> >>
> >
> >
> >
> > --
> > Cheers,
> > Manthosh Kumar. T
> >
> >
>
>
> --
> Cheers,
> Manthosh Kumar. T
>
Re: Geographically distant client
Posted by Manthosh Kumar T <ma...@gmail.com>.
Pardon me if I miss anything, like any network issues
On 3 April 2014 18:57, Manthosh Kumar T <ma...@gmail.com> wrote:
> I mean directly interacting with the remote zookeeper. Say I'm able to
> access the zk server and hbase server externally.
>
>
> On 3 April 2014 18:54, Ted Yu <yu...@gmail.com> wrote:
>
>> Regions hosted by the server may be moved to other servers.
>>
>> Can you clarify what you meant by directly writing to the server ?
>>
>> Thanks
>>
>> On Apr 3, 2014, at 5:59 AM, Manthosh Kumar T <ma...@gmail.com> wrote:
>>
>> > Hi All,
>> > I have around 20-30 geographically distant clients that need to
>> > write data to a centralized HBase server. I have dedicated VPN for the
>> > communication and hence bandwidth won't be a issue. is it a good idea to
>> > make the clients directly send data to the centralized server?. Or a
>> > geographically distributed Hadoop cluster is more efficient for this
>> > scenario? Has anybody come across such use case?. Need suggestions on
>> the
>> > viability and efficiency of the setup to follow
>> >
>> > --
>> > Cheers,
>> > Manthosh Kumar. T
>>
>
>
>
> --
> Cheers,
> Manthosh Kumar. T
>
>
--
Cheers,
Manthosh Kumar. T
Re: Geographically distant client
Posted by Manthosh Kumar T <ma...@gmail.com>.
I mean directly interacting with the remote zookeeper. Say I'm able to
access the zk server and hbase server externally.
On 3 April 2014 18:54, Ted Yu <yu...@gmail.com> wrote:
> Regions hosted by the server may be moved to other servers.
>
> Can you clarify what you meant by directly writing to the server ?
>
> Thanks
>
> On Apr 3, 2014, at 5:59 AM, Manthosh Kumar T <ma...@gmail.com> wrote:
>
> > Hi All,
> > I have around 20-30 geographically distant clients that need to
> > write data to a centralized HBase server. I have dedicated VPN for the
> > communication and hence bandwidth won't be a issue. is it a good idea to
> > make the clients directly send data to the centralized server?. Or a
> > geographically distributed Hadoop cluster is more efficient for this
> > scenario? Has anybody come across such use case?. Need suggestions on the
> > viability and efficiency of the setup to follow
> >
> > --
> > Cheers,
> > Manthosh Kumar. T
>
--
Cheers,
Manthosh Kumar. T
Re: Geographically distant client
Posted by Ted Yu <yu...@gmail.com>.
Regions hosted by the server may be moved to other servers.
Can you clarify what you meant by directly writing to the server ?
Thanks
On Apr 3, 2014, at 5:59 AM, Manthosh Kumar T <ma...@gmail.com> wrote:
> Hi All,
> I have around 20-30 geographically distant clients that need to
> write data to a centralized HBase server. I have dedicated VPN for the
> communication and hence bandwidth won't be a issue. is it a good idea to
> make the clients directly send data to the centralized server?. Or a
> geographically distributed Hadoop cluster is more efficient for this
> scenario? Has anybody come across such use case?. Need suggestions on the
> viability and efficiency of the setup to follow
>
> --
> Cheers,
> Manthosh Kumar. T