incubator-cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Adrian Cole <adrian.f.c...@gmail.com>
Subject Re: Hibernate
Date Thu, 09 Aug 2012 15:58:44 GMT
Having worked with the cloudbridge, and experimented ripping out soap, s3,
duplicated or unreferenced code, the module becomes quite small, apart from
generated classes.  There is hardly any state, and certainly not highly
relational state.

What I'm getting at is that I see no reason to make small things that have
a fighting chance to be succinct at some point, to be bogged down with
inheriting machinery from cloudstack core.

The irony is that most of the bridge is modular or decoupled, even if not
packaged well, and at the same time we discuss architecture to reduce
coupling in the monster code, we are increasing coupling in this!

Cloud bridge persistence, at least ec2, can largely be solved by a Map.
Moving to any heavier persistence layer by default seems overkill to me.  I
think any heavy integration should be a plugin, (and in fact cloudstack API
integration should be! *).

Long way of saying let's not get too concerned about making cloudbridge
follow patterns we have doubts about anyway, and instead use that effort to
clean it up.

My 2p
-A

* aside, but EC2Engine should be an interface, etc. Right now, unit testing
is pretty non existent.
On Aug 9, 2012 8:14 AM, "David Nalley" <david@gnsa.us> wrote:

> Rajesh, is this done?
>
> --David
>
> On Mon, Jul 16, 2012 at 11:49 PM, Rajesh Battala
> <rajesh.battala@citrix.com> wrote:
> > Yes Ewan, I would be doing this work.
> >
> > Thanks
> > Rajesh Battala
> >
> >> -----Original Message-----
> >> From: Ewan Mellor [mailto:Ewan.Mellor@eu.citrix.com]
> >> Sent: Tuesday, July 17, 2012 9:03 AM
> >> To: cloudstack-dev@incubator.apache.org
> >> Subject: RE: Hibernate
> >>
> >> OK, good.  So it's a small job.  So is Rajesh doing that?
> >>
> >> Ewan.
> >>
> >> > -----Original Message-----
> >> > From: Alex Huang [mailto:Alex.Huang@citrix.com]
> >> > Sent: 16 July 2012 17:03
> >> > To: cloudstack-dev@incubator.apache.org
> >> > Subject: RE: Hibernate
> >> >
> >> > My understanding is it's one table and is already wrapped by a DAO.
> >> > The DAO object uses hibernate underneath.  Don't think it's a big job.
> >> >
> >> > --Alex
> >> >
> >> > > -----Original Message-----
> >> > > From: Ewan Mellor [mailto:Ewan.Mellor@eu.citrix.com]
> >> > > Sent: Monday, July 16, 2012 4:43 PM
> >> > > To: cloudstack-dev@incubator.apache.org
> >> > > Subject: RE: Hibernate
> >> > >
> >> > > If I read this infinitely long thread correctly, the conclusion is
> >> > that
> >> > > CloudStack/awsapi needs to be rewritten to use the CloudStack DAO
> >> > instead
> >> > > of Hibernate.  Is that correct?
> >> > >
> >> > > If so, who is doing this work, and when?  That sounds like a big
> job.
> >> > >
> >> > > Thanks,
> >> > >
> >> > > Ewan.
> >> > >
> >> > > > -----Original Message-----
> >> > > > From: Alex Huang [mailto:Alex.Huang@citrix.com]
> >> > > > Sent: Monday, July 02, 2012 11:31 PM
> >> > > > To: cloudstack-dev@incubator.apache.org
> >> > > > Subject: RE: Hibernate
> >> > > >
> >> > > > Rajesh,
> >> > > >
> >> > > > You should use the existing CloudStack DAO framework.  If
> >> > cloudstack
> >> > > > will move to another ORM framework, it will have to be done in
> >> > > > such
> >> > a
> >> > > > fashion that everything above the DAO layer doesn't have to change
> >> > and
> >> > > > the majority of the current DAO code do not have to change as
> well.
> >> > > > Which would mean your code should migrate over along with the
rest
> >> > of
> >> > > > the changes.
> >> > > >
> >> > > > --Alex
> >> > > >
> >> > > > > -----Original Message-----
> >> > > > > From: Rajesh Battala [mailto:rajesh.battala@citrix.com]
> >> > > > > Sent: Monday, July 02, 2012 10:59 PM
> >> > > > > To: cloudstack-dev@incubator.apache.org
> >> > > > > Subject: RE: Hibernate
> >> > > > >
> >> > > > >
> >> > > > >
> >> > > > > > -----Original Message-----
> >> > > > > > From: Kelven Yang [mailto:kelven.yang@citrix.com]
> >> > > > > > Sent: Saturday, June 30, 2012 3:31 AM
> >> > > > > > To: cloudstack-dev@incubator.apache.org
> >> > > > > > Subject: RE: Hibernate
> >> > > > > >
> >> > > > > >
> >> > > > > >
> >> > > > > > > -----Original Message-----
> >> > > > > > > From: David Nalley [mailto:david@gnsa.us]
> >> > > > > > > Sent: Friday, June 29, 2012 10:37 AM
> >> > > > > > > To: cloudstack-dev@incubator.apache.org
> >> > > > > > > Subject: Re: Hibernate
> >> > > > > > >
> >> > > > > > > On Fri, Jun 29, 2012 at 12:13 PM, Rajesh Battala
> >> > > > > > > <rajesh.battala@citrix.com> wrote:
> >> > > > > > > >
> >> > > > > > > >
> >> > > > > > > >> -----Original Message-----
> >> > > > > > > >> From: David Nalley [mailto:david@gnsa.us]
> >> > > > > > > >> Sent: Friday, June 29, 2012 8:40 PM
> >> > > > > > > >> To: cloudstack-dev@incubator.apache.org
> >> > > > > > > >> Subject: Re: Hibernate
> >> > > > > > > >>
> >> > > > > > > >> On Fri, Jun 29, 2012 at 9:27 AM, Rajesh
Battala
> >> > > > > > > >> <rajesh.battala@citrix.com>
> >> > > > > > > >> wrote:
> >> > > > > > > >> >
> >> > > > > > > >> > Shall I proceed to replace Hibernate
Framework with
> >> > Spring
> >> > > > JDBC.
> >> > > > > > > >> >
> >> > > > > > > >>
> >> > > > > > > >> This is not a design decision to take
lightly, or rush -
> >> > as
> >> > > > we've
> >> > > > > > > >> talked about replacing the custom ORM
that CloudStack
> >> > > > > > > >> uses
> >> > as
> >> > > > > > > >> well
> >> > > > > > > >> - we'd want to make a decision on something
that we are
> >> > > > > > > >> comfortable with
> >> > > > > > > in that role as well.
> >> > > > > > > >>
> >> > > > > > > > [Rajesh Battala]
> >> > > > > > > > So we are planning to implement some ORM
framework for
> >> > > > CloudStack
> >> > > > > > also?
> >> > > > > > > > If so, then we need to evaluate the available
ORM
> >> > frameworks
> >> > > > for
> >> > > > > > > > java
> >> > > > > > > which are based on Apache License?
> >> > > > > > > >
> >> > > > > > >
> >> > > > > > >
> >> > > > > > > I don't want to speak for Alex, but I have heard
him mention
> >> > > > > > > such
> >> > > > a
> >> > > > > > > change number of times, and have seen it in his
slides as
> >> > well.
> >> > > > > > >
> >> > > > > > > But the larger sense is that if we are going to
change
> >> > something
> >> > > > in
> >> > > > > > > AWSAPI for communicating to the database, it should
be
> >> > something
> >> > > > > > > that we are comfortable with adopting in the long
term for
> >> > the
> >> > > > rest
> >> > > > > > > of CloudStack, otherwise we'll end up with two
different
> >> > > > > > > ways for connecting to the database, which is
fail.
> >> > > > > >
> >> > > > > > For Rajesh's concern in the short term (to get rid
of
> >> > Hibernate),
> >> > > > we
> >> > > > > > basically have following options,
> >> > > > > > 1) Make the project depend on utils.jar in CloudStack
and use
> >> > > > existing
> >> > > > > > CloudStack persistent layer.
> >> > > > > > 2) Continue to make CloudBridge project independent
> >> > > > > >     a) use direct JDBC
> >> > > > > >     b) use Spring JDBC
> >> > > > > >     c) use ORM framework on Spring (i.e. OpenJPA etc)
> >> > > > > >
> >> > > > > > I vote on 2-b for a short term resolution. For a couple
of
> >> > > > > > reasons,
> >> > > > > >
> >> > > > > > The major reason is for adopting Spring, EC2/S3 API
in the
> >> > future
> >> > > > may
> >> > > > > > both need to support SOAP based web service API and
REST API,
> >> > > > Spring
> >> > > > > > comes with support of both, in addition, it provides
> >> > > > > > dependency injection, AOP and pluggable ORM framework,
> >> > > > > > distributed transaction
> >> > > > > management, etc.
> >> > > > > >
> >> > > > > > CloudBridge project has a very simple data model (it
has only
> >> > one
> >> > > > > > table for EC2), original ORM (hibernate) usage is just
for
> >> > > > convenience
> >> > > > > > (as Alex pointed that it tried to avoid direct SQL
> >> > manipulation),
> >> > > > on
> >> > > > > > the other hand, giving the DAO abstraction, a limited
amount
> >> > > > > > of
> >> > > > JDBC
> >> > > > > > should not matter too much either. With Spring and
already
> >> > > > > > existing DAO abstraction, we can leave bigger ORM decision
for
> >> > the
> >> > > future.
> >> > > > > >
> >> > > > > > In the long term, ORM choice to us seems to be secondary
> >> > > > > > consideration, giving the fact that our current custom
ORM
> >> > > > > > implementation serves its purpose well enough in CloudStack.
> >> > But I
> >> > > > do
> >> > > > > > see the needs of more flexible dependency injection
or
> >> > transaction
> >> > > > > > management framework in order to incorporate better
> >> > > > > > integration
> >> > > > with
> >> > > > > > external transactional systems like message queue servers.
I
> >> > also
> >> > > > see
> >> > > > > > the needs of standardizing AOP patterns used in current
> >> > CloudStack
> >> > > > > > code base, Spring provides some good answers for that.
So
> >> > > > > > there
> >> > is
> >> > > > a
> >> > > > > > possibility that it will be good for CloudStack to
adopt it
> >> > like
> >> > > > many
> >> > > > > > other Java communities do, so that we can focus on
our own
> >> > > > > > business logic and leave these external concerns to
people who
> >> > are
> >> > > > > > doing it
> >> > > > for a
> >> > > > > living.
> >> > > > > >
> >> > > > > > Kelven
> >> > > > > >
> >> > > > > >
> >> > > > > > >
> >> > > > > > > --David
> >> > > > > [Rajesh Battala]
> >> > > > >
> >> > > > > Please let me know conclusion of the thread.
> >> > > > > I didn't see any reply to Kelven suggestions.
> >> > > > >
> >> > > > > Thanks
> >> > > > > Rajesh Battala
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message