Home » Eclipse Projects » Commercial » [commercial, off-topic?] Need external review. Data models, meta-data and database integration for r
| [commercial, off-topic?] Need external review. Data models, meta-data and database integration for r [message #17888] |
Thu, 12 March 2009 13:21  |
Eclipse User |
|
|
|
Originally posted by: eclipse.efef.dk
Dear EMF News Group,
Pre amble
This is an attempt to see if anyone (rather someone, not anyone can do
it) on this NG would be interested in a piece of consultant work. If
not, skip the rest, or 'flame' me to post a more appropriate place, and
sorry for wasting your time.
If your curiosity is piqued, read on to see if it sounds like something
within your area of special expertise. If it is, I would very much like
to hear from you.
In brief. In my paid day job, I am responsible for a new reporting
system, architecture, project etc. The project is of strategic
importance to the company, so the C* and the board would like an
external review of the core datamodels; I see it as their insurance and
my opportunity to get expert advisory by the very best :) The exact
scope is not yet determined.
Summary of the architecture
We have created a meta-model (call it our DomainModel) which is an EMF
model. The DomainModel specifies two levels, the domain models, and
instances of domain models. An example of domain level data is that we
have actions which can take certain parameters; we have browsers with
certain properties. At the instance level we have listed the valid
actions, and we have a list of browsers that we populate with our
knowledge of them (supports css2, has bug X, etc.).
The domain model is attempted kept simple, yet relatively powerful. Logs
will use UUID or similar references to data in the DomainModel and add
values for parameters (user input, e.g.) where necessary.
In other words, the data stored in the DomainModel is an ontology for
the data we will accept logged; it is also a container for domain
specific knowledge, that can be used in reports. From the data in the
DomainModel, we can 1) generate API's to be used by logees, 2) validate
logs, 3) to easily extend the domain for logs, but at the same time
keeping a rather strict control with the logs we allow (in order to
fight the garbage in, garbage out nuisance).
The DomainModel is Teneo-enabled (performance is not crucial, since I
expect to to be slowly varying and of a size that allows for caching all
data needed when processing logs).
The DomainModel is inspired by RDF, OWL as well as EMF, and could
probably very well have been expressed in Ecore itself.
The logs, after validation is stored two places. The logs will be stored
in a raw-logs database (possibly in schema described by the DomainModel,
but separate from the domain data itself - this is still TBD by
prototype experimentation). A subset of the logs, determined by
reporting requirements, will also be stored in a denormalised model
(simplest case, one table with very very many columns and lots of
NULLs). The dataware house will be extensible by configuration, and I
imagine generated HQL or SQL to take care of that part. In the dataware
house and in the raw logs, the UUID references will be replaced with
simpler foreign keys in the database to increase performance.
Post amble
That's the very short story. If interested, you can choose to post to
the news group, or send me a private email, I will watch both.
I have no specific data on the payment amount and model yet, so if you
have done this before, I'd welcome your proposal in your response.
We are already in contact with a local consultancy company, that I do
not know personally, but I personally do know that a lot of the EMF NG
hangarounds definitely would have the skill set we are looking for in a
reviewer.
Cheers,
/\/ikolaj
P.S. If Ed reaches out for this task, I, personally would not invest
much time in competing for it ;)
P.P.S. If the commercial impact of this post is offending, but the
problem is interesting, I could consider to to find out if I could post
the details and get the 'review' for free by the comments in the NG.
P.P.P.S Sorry for cross posting to e.t.EMF, but I myself did not know,
or subscribe to commercial/employment before looking for the "right
place to post"; and definitely myself could have been interested in this
post.
|
|
|
| Re: [commercial, off-topic?] Need external review. Data models, meta-data and database integration f [message #17905 is a reply to message #17888] |
Thu, 12 March 2009 15:02  |
 Ed Merks
Messages: 33140
Registered: July 2009 |
Senior Member |
|
|
/\/ikolaj,
Comments below.
berntsen wrote:
> Dear EMF News Group,
>
> Pre amble
>
> This is an attempt to see if anyone (rather someone, not anyone can do
> it) on this NG would be interested in a piece of consultant work. If
> not, skip the rest, or 'flame' me to post a more appropriate place,
> and sorry for wasting your time.
I'm an independent consulting now and I work in partnership with itemis,
so I'm definitely interested in the post, though I'm not sure it's so
appropriate on the EMF newsgroup; I'll keep reading and form an opinion. :-)
>
> If your curiosity is piqued, read on to see if it sounds like
> something within your area of special expertise. If it is, I would
> very much like to hear from you.
>
> In brief. In my paid day job, I am responsible for a new reporting
> system, architecture, project etc. The project is of strategic
> importance to the company, so the C* and the board would like an
> external review of the core datamodels; I see it as their insurance
> and my opportunity to get expert advisory by the very best :) The
> exact scope is not yet determined.
Still sounds interesting. After all, when making a long term decision,
it's certainly a good idea to be sure the designs are sound and worth
the investment...
>
> Summary of the architecture
>
> We have created a meta-model (call it our DomainModel) which is an EMF
> model. The DomainModel specifies two levels, the domain models, and
> instances of domain models. An example of domain level data is that we
> have actions which can take certain parameters; we have browsers with
> certain properties. At the instance level we have listed the valid
> actions, and we have a list of browsers that we populate with our
> knowledge of them (supports css2, has bug X, etc.).
Like a DSL...
>
> The domain model is attempted kept simple, yet relatively powerful.
Like Ecore. :-P
> Logs will use UUID or similar references to data in the DomainModel
> and add values for parameters (user input, e.g.) where necessary.
>
> In other words, the data stored in the DomainModel is an ontology for
> the data we will accept logged; it is also a container for domain
> specific knowledge, that can be used in reports. From the data in the
> DomainModel, we can 1) generate API's to be used by logees, 2)
> validate logs, 3) to easily extend the domain for logs, but at the
> same time keeping a rather strict control with the logs we allow (in
> order to fight the garbage in, garbage out nuisance).
Sounds cool.
>
> The DomainModel is Teneo-enabled (performance is not crucial, since I
> expect to to be slowly varying and of a size that allows for caching
> all data needed when processing logs).
>
> The DomainModel is inspired by RDF, OWL as well as EMF, and could
> probably very well have been expressed in Ecore itself.
:-P
>
> The logs, after validation is stored two places. The logs will be
> stored in a raw-logs database (possibly in schema described by the
> DomainModel, but separate from the domain data itself - this is still
> TBD by prototype experimentation). A subset of the logs, determined by
> reporting requirements, will also be stored in a denormalised model
> (simplest case, one table with very very many columns and lots of
> NULLs). The dataware house will be extensible by configuration, and I
> imagine generated HQL or SQL to take care of that part. In the
> dataware house and in the raw logs, the UUID references will be
> replaced with simpler foreign keys in the database to increase
> performance.
>
> Post amble
>
> That's the very short story. If interested, you can choose to post to
> the news group, or send me a private email, I will watch both.
It sounds interesting.
>
> I have no specific data on the payment amount and model yet, so if you
> have done this before, I'd welcome your proposal in your response.
Hehehehe. Not in pubic. Send a note to ed.merks at gmail dot com
>
> We are already in contact with a local consultancy company, that I do
> not know personally, but I personally do know that a lot of the EMF NG
> hangarounds definitely would have the skill set we are looking for in
> a reviewer.
By now I think I've seen it all: the good, the bad, and the ugly.
>
> Cheers,
> /\/ikolaj
>
>
> P.S. If Ed reaches out for this task, I, personally would not invest
> much time in competing for it ;)
It's nice to know I'm such stiff competition. :-P
>
> P.P.S. If the commercial impact of this post is offending, but the
> problem is interesting, I could consider to to find out if I could
> post the details and get the 'review' for free by the comments in the NG.
It doesn't strike me as unreasonable to solicit paid help on a newsgroup
where usually you get free help.
>
> P.P.P.S Sorry for cross posting to e.t.EMF, but I myself did not
> know, or subscribe to commercial/employment before looking for the
> "right place to post"; and definitely myself could have been
> interested in this post.
I didn't know about the commercial one; I do monitor the employment one.
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
|
| Re: [commercial, off-topic?] Need external review. Data models, meta-data and database integration f [message #566856 is a reply to message #17888] |
Thu, 12 March 2009 15:02 |
Ed Merks
Messages: 33140
Registered: July 2009 |
Senior Member |
|
|
/\/ikolaj,
Comments below.
berntsen wrote:
> Dear EMF News Group,
>
> Pre amble
>
> This is an attempt to see if anyone (rather someone, not anyone can do
> it) on this NG would be interested in a piece of consultant work. If
> not, skip the rest, or 'flame' me to post a more appropriate place,
> and sorry for wasting your time.
I'm an independent consulting now and I work in partnership with itemis,
so I'm definitely interested in the post, though I'm not sure it's so
appropriate on the EMF newsgroup; I'll keep reading and form an opinion. :-)
>
> If your curiosity is piqued, read on to see if it sounds like
> something within your area of special expertise. If it is, I would
> very much like to hear from you.
>
> In brief. In my paid day job, I am responsible for a new reporting
> system, architecture, project etc. The project is of strategic
> importance to the company, so the C* and the board would like an
> external review of the core datamodels; I see it as their insurance
> and my opportunity to get expert advisory by the very best :) The
> exact scope is not yet determined.
Still sounds interesting. After all, when making a long term decision,
it's certainly a good idea to be sure the designs are sound and worth
the investment...
>
> Summary of the architecture
>
> We have created a meta-model (call it our DomainModel) which is an EMF
> model. The DomainModel specifies two levels, the domain models, and
> instances of domain models. An example of domain level data is that we
> have actions which can take certain parameters; we have browsers with
> certain properties. At the instance level we have listed the valid
> actions, and we have a list of browsers that we populate with our
> knowledge of them (supports css2, has bug X, etc.).
Like a DSL...
>
> The domain model is attempted kept simple, yet relatively powerful.
Like Ecore. :-P
> Logs will use UUID or similar references to data in the DomainModel
> and add values for parameters (user input, e.g.) where necessary.
>
> In other words, the data stored in the DomainModel is an ontology for
> the data we will accept logged; it is also a container for domain
> specific knowledge, that can be used in reports. From the data in the
> DomainModel, we can 1) generate API's to be used by logees, 2)
> validate logs, 3) to easily extend the domain for logs, but at the
> same time keeping a rather strict control with the logs we allow (in
> order to fight the garbage in, garbage out nuisance).
Sounds cool.
>
> The DomainModel is Teneo-enabled (performance is not crucial, since I
> expect to to be slowly varying and of a size that allows for caching
> all data needed when processing logs).
>
> The DomainModel is inspired by RDF, OWL as well as EMF, and could
> probably very well have been expressed in Ecore itself.
:-P
>
> The logs, after validation is stored two places. The logs will be
> stored in a raw-logs database (possibly in schema described by the
> DomainModel, but separate from the domain data itself - this is still
> TBD by prototype experimentation). A subset of the logs, determined by
> reporting requirements, will also be stored in a denormalised model
> (simplest case, one table with very very many columns and lots of
> NULLs). The dataware house will be extensible by configuration, and I
> imagine generated HQL or SQL to take care of that part. In the
> dataware house and in the raw logs, the UUID references will be
> replaced with simpler foreign keys in the database to increase
> performance.
>
> Post amble
>
> That's the very short story. If interested, you can choose to post to
> the news group, or send me a private email, I will watch both.
It sounds interesting.
>
> I have no specific data on the payment amount and model yet, so if you
> have done this before, I'd welcome your proposal in your response.
Hehehehe. Not in pubic. Send a note to ed.merks at gmail dot com
>
> We are already in contact with a local consultancy company, that I do
> not know personally, but I personally do know that a lot of the EMF NG
> hangarounds definitely would have the skill set we are looking for in
> a reviewer.
By now I think I've seen it all: the good, the bad, and the ugly.
>
> Cheers,
> /\/ikolaj
>
>
> P.S. If Ed reaches out for this task, I, personally would not invest
> much time in competing for it ;)
It's nice to know I'm such stiff competition. :-P
>
> P.P.S. If the commercial impact of this post is offending, but the
> problem is interesting, I could consider to to find out if I could
> post the details and get the 'review' for free by the comments in the NG.
It doesn't strike me as unreasonable to solicit paid help on a newsgroup
where usually you get free help.
>
> P.P.P.S Sorry for cross posting to e.t.EMF, but I myself did not
> know, or subscribe to commercial/employment before looking for the
> "right place to post"; and definitely myself could have been
> interested in this post.
I didn't know about the commercial one; I do monitor the employment one.
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
|
Goto Forum:
Current Time: Thu Apr 25 08:21:17 GMT 2024
Powered by FUDForum. Page generated in 0.03270 seconds |
] 
Search
Help
Register
Login
Home
