Disconnected / Distributed Architecture

I'm working with a customer that wants to implement a disconnected / distributed architecture. They have groups of inspectors that travel around the world, including 3rd world countries with no possibility of cellular data access. They have a pool of about 70 laptops that they use. Each person in the group of inspectors will have their own laptop and use it to do an on site evaluation. Currently they then export these as files and the existing software brings these files together in a report. The problem is it doesn't really work and the company that built it went out of business.
Whenever I hear someone wanting to do disconnected mobile I strongly advise against it then run the other way. It's a very challenging problem to sync either the data or the app, let alone both. Someone else proposed APEX, and as much as I love APEX it just doesn't feel like the right technology. If skill-set weren't an issue, I'd probably build it in Adobe Air with a local SQL-Lite database that syncs up to the Oracle "mother-ship". At least Air has the concept of automatically updating the application whenever it connects built into it's core architecture.
I have a few thoughts on syncing such as:
- Roll your own data sync with PL/SQL over DB links. Should be less than 20 tables.
- Advanced Replication
- GoldenGate (if price is not an issue). Could possible sync the APEX_ tables to sync up the app too
- Write a VB app that checks for application updates when they are connected, then downloads the APEX app and installs it
So, my questions are:
- Has anyone here done something similar in APEX? If so, can you discuss the details including data and application updates?
- Has anyone done something similar with another technology stack?
Tyler Muth
http://tylermuth.wordpress.com
"Applied Oracle Security: Developing Secure Database and Middleware Environments": http://sn.im/aos.book

Hi Tyler,
I share your first thougts - run the other way. ;)
I haven't done something like that with APEX, but I guess using DB-Links is one option, but not the ideal one. Treating an application like data is possible, though the deployment could get a little tricky, if you want this to happen automaticly.
- Of course, you could store the application archive as BLOB or even CLOB column in a table along with other meta data (e.g. build number, timestamp, etc.).
- You could build an application that acts as update manager that stores information (build numbers, timestamps etc.) for locally installed applications and data, and looks up possible updates - for both data and applications.
- I would try to do the data transfer with web services rather than using data base links. APEX-integration is getting better with every new version and it sounds more handy to me, especially error handling etc.
Leaves you with the point how to deploy your application. Since the archive itself is SQL, you could download it using a select statement and simply execute them. There exists a nice API, but I don't know if you can use it within APEX easily. As for images and other items: Since you probably use a scenario with EPG as HTTP-Server for APEX, it should not be too hard to retrieve BLOB-data and load it into XDB.
It get's a lot easier, if the inspectors would do the deployment part. Since this is not too complicated in APEX, it might be just the update manager that offers them the appropriate downloads.
I'd be intrested to know how you actually realized this architecture, if you ever get it running before it get's you to run. ;)
-Udo

Similar Messages

BPEL 10.1.2 hub-and-spoke or distributed architecture?

Hi,
I'm currently wrestling with the following question:BPEL 10.1.2 hub-and-spoke or distributed architecture?
Hi,
I'm currently wrestling with the following question:
An ESB as per definition of e.g. Forrester should be capable of supporting a distributed bus architecture. From my understanding this distributed bus architecture is achieved by installing some sort of ESB component(s) on all machines that are participating in this infrastructure, together forming a ‘bus’.
As I understand the BPEL 10.1.2 product basically offers two categories of functionality: orchestration and integration. Does this integration part offer ESB alike functionality and more specific allows for a distributed bus architecture? As far as I can see the BPEL 10.1.2 offers limited ESB alike functionality and only supports a hub-and-spoke architecture.
Other threads in this forum talk about using BPEL 10.1.2 together with InterConnect in order to foresee in ESB functionality. What does InterConnect add to the BPEL 10.1.2 integration functionality?
As of SOA suite 10.1.3 these products have been split up into a BPEL product and an ESB product. Is the ESB product in SOA suite 10.1.3 a combination of the integration from BPEL 10.1.2 and InterConnect? Is this new ESB product able to support a distributed architecture?
I’m very much in favor of a distributed architecture compared to hub-and-spoke, as hub-and-spoke requires a very solid and redundant system that is going to handling all message traffic and other functions. When moving towards a SOA giving an ESB a back-bone role, I’m not very keen on introducing a single system that should actually make up this ESB. Distributed would mean all machines are taking care of some basic functions resulting in a fully functional ESB, even when one or more machines are down.
Am I making sense with this? I would like to know how others are looking at these topics.
Regards,
Gershon Janssen

We are struggling with this issue too. The "all-pervasive" vision of the ESB visionaries, in my opinion, means that every node in my enterprise architecture should have access to the bus and I should be able to orchestrate anything that is running on any node in my architecture. We were told to think of an ESB as the equivalent of the hardware bus in computers. From that perspective I thought that:
(1) BPEL and ESB functions would add a marginal increment to the licensing cost of an app server - I should be able to afford an "all-pervasive" architecture.
(2) BPEL and ESB functions would add a "marginal increment" to the memory and resource foot-print. Again, I should be able to afford an "all-pervasive" architecture.
Are these two satisfied by Oracle's products ? Some vendors don't seem to support the above two. In that case there is a disconnect between the marketing and technology departments of the vendors. What am I missing ?
If BPEL engines and ESBs are priced very high, based on economics we will end up with a hub-and-spoke model.
Thanks

Distributed architecture using JAVA RMI

Hi!
I want to implement a distributed architecture, in which clients invoke any of several servers, depending on their availability. The clients are not browsers, so web-server (e.g. Tomcat) clustering is not relevant here. It's a cluster of servers, each capable of providing the same functionality, but clients should be able to "find" the less loaded one and invoke it. Now, with EJB's I believe this is provided out-of-the-box, but what about using RMI? How do two processes find each other via RMI, and is it possible to put a load-balancer in the middle?
Is it possible to setup a load-balancer IP as RMI host endpoint?
Thanks a lot
A.

They find one another using the rmiregistry tool from the JDK bin, which servers register with and clients use as a lookup. AFAIK this doesn't provide provision for anything resembling load balancing. Java RMI is a fairly basic facility.
You should probably take a look at a framework like osgi.

Design of first distributed architecture

Okay, so I'm trying to design and implement my first distributed project, and I want to make sure my general design is okay before I start. Like I say, I've never used J2EE before, and I'm still learning about it. This is very general.
Problem: Design a login system complete with the following aspects:
- A basic login window/applet that takes a username and password and verifies them. If they are valid (as determined by specific business logic), return a cookie (assume they are allowed for now) and allow the user access to the requested resources on the server. (Note: I think I'm using the term business logic correctly, but tell me if I'm not.)
- A basic registration window/applet that takes a user's information. Depending on the business logic, the user will be registered to the database for a certain access level or will become a �pending user,� which will be reviewed by an administrator before being allowed to access the server�s resources.
- An administrator�s window/applet that is basically a view of all that goes on in the server. This includes
* current users
* pending users
* audits
* files on the server
* any directory and access protocols
* server settings
Proposal: I plan to follow basic J2EE architecture patterns: The browser will allow access to all of the windows. The windows will call on servlets, which will call on EJBs to access the appropriate tables, all in the same database. I will make a servlet for each table category: UserServlet (for accessing the user list, maybe the list of pending users), AuditServlet, SettingsServlet, LdapServlet, etc.
Should I make a separate servlet and EJB for each category (as listed above)? I still have to read up on EJBs before I implement this.
So that is my basic skeleton, but I have no idea as to its quality. What do you think?
theAmerican

_dnoyeB said:
An EJB per table seems odd. You should consider some
form of ORM. Perhaps EJB3 will work for you. As I
understand it, and its beena while since i did j2ee,
EJB represents a thing, not a databaase table. i.e.
an EJB represents a person, not the person table.Okay, that makes sense. (The idea of one EJB/table didn't seem exactly right to me.) By ORM, you mean something like Hibernate or EJB3? (Could Spring be involved here? I know it isn't solely concerned with DBs.) I appreciate the advice.
Saish said:
I would also consider implementing an out-of-the-box controller
ala Struts or Spring rather than dozens of Servlets. I was thinking about using Spring. Would you recommend doing that before I even do my first distributed app? Also, which do you prefer: Struts or Spring? Thanks for the input.
Hoos said:
If you are using J2EE I would suggest you read up on the security services
provided by your J2EE container. It should provide you with instructions on
setting up realms etc and also provide components to access security data
in order to perform authentication and authorisation etc.
Unless you have need for extended security you will probably find that the
containers services means you wont need to write your own components e.g.
LoginEJB that checks usernames and passwords etc.Awesome. Thanks for the advice. I'll look into it. (This is just kindof an excercise to get my feet wet in distributed programming.) I may consider doing another project instead.
theAmerican

Installation of Hyperion Planning in a distributed Architecture

Hi All,
I am new to the world of Hyperion..
I am trying to install Hyperion Palnning 11.1.1.1.0 in a distributed environment.. and facing difficulty deciding where each component should go..
Here's a description of the Architecure In which i want to get Hyperion Planning up and running:
One Windows server box -- EPMA should be installed on windows as i have read through the installation guides available ..
Two AIX boxes -- have decided to put the database and Essbase on one and Planning on the other.
I am using Oracle 11g as the daatbase. And Oracle App server for deployment.
According to my understanding .. I need the following compenents to be installed for Planning...
Shared Services
Essbase --client , server , administrative services, provider services(for Smart office )
Planning
EPMA
WorkSpace
Can somebody plaese advice on how i should go about installing the components?
An advice would be appreciated.
Thanks in advance!!!

Hello John,
Thanks for replying..
Now under Essbase in the config utility I have the following options
Essbase --Essbase Custom Configuration
Essbase Administration Services -- Configure Database, Deploy to Application server
Habnet Server -- Deploy to Application Server
Provider Services - Deploy to Application Server
Smart Search - Deploy to Appplication Server
Studio Configure Database
So among these Shall i select only Essbase for installing the Essbase db..?
Also can you please explain how does shared service work?
Say i deployed Planning to the app server on the same machine and EPMA to the app server on the windows machine
then how do they work together..?
WIll i need to modify some properties file..?
Thanks,
Maya

Use of flat files in distributed architecture probs.

Hi Guiess,
I am working in a module in distributed environment.
Its client is swing application.I have a one doubt and the doubt is -
I am making some flow chart sort of thing at client side and saving this work in
a flat file (java property file/xml file) at the server side.
Could n e 1 suggest me that saving flat file at the server side would work,
coz this file would share by multiple clients.And how locking facility j2e gives whilesomebody is modifing one file.
Thnax in advance -
Vikas Kumar

Hi,
side and saving this work in
a flat file (java property file/xml file) at the
server side.
Could n e 1 suggest me that saving flat file at the
server side would work,
coz this file would share by multiple clients.And how
locking facility j2e gives whilesomebody is modifing
one file.Does "server side" mean EJBs? Is so, from the EJB 2.1 specs document, section 25.1.2 :
"An enterprise bean must not use the java.io package to attempt to access files and directories in the file system.
The file system APIs are not well-suited for business components to access data. Business components should use a resource manager API, such as JDBC, to store data."
Just make sure you use a resource manager to access files.
Kexkey

SAP MII 15.0:HANA as a back-end ! Distributed Architecture

Dear MII Consultants & Architects,
I have a query here regarding SAP MII 15.0 when it is deployed as de-centralized.
Lets say there are 3 Plants.Each plant needs a local MII system.
How a local MII system will interact with HANA if HANA is a central entity?
Will all the local MII installations which have their respective SQL/Oracle do a SLT replication with HANA?
Regards,
Muddassar

There's not just a single way to interact between MII & HANA; there are many options that are available so that we can support the underlying business scenario in the best way. For options and architectures that support them have a look here:
SAP MII, HANA/ESP and Lumira (Technical)
I also have a technical architectures deck that I can share with you if you like just email me directly for it.
Please let me know what you think of my post about MII & HANA,
Sam

Distributed Architecture

I have made plenty of Web Components using Servlets and JSP and deployed few EJB applications.
I have been trying figure out for a long time how these applications work in real life scenario.
Q1. What is the architecture of Server farms ?
Q2. How these components are deployed on Servers. Do they deploy multiple copies of application on each server or what type of logic they use so that they are able to work efficiently, reliably ?
Q3. What is the hardware configuration is used in Server farms ?
Q4. Do Database Systems have extra memory to work with ? What happens if one Database Server is not enough ? If more than one Database Server is used, how they ensure they have same data.
Q5. Can you give a case scenario or real life example.
Thank you for reading my questions.

gtjdjtnd wrote:
I have made plenty of Web Components using Servlets and JSP and deployed few EJB applications.
I have been trying figure out for a long time how these applications work in real life scenario.
Q1. What is the architecture of Server farms ?Buildings with air conditioning. Usually rectangular, maybe square. Some rooms in them to house servers.
Q2. How these components are deployed on Servers. Do they deploy multiple copies of application on each server or what type of logic they use so that they are able to work efficiently, reliably ? Maven POMs coupled with GNU make files.
Q3. What is the hardware configuration is used in Server farms ?A processor, some memory, an I/O Bus, some storage.
Q4. Do Database Systems have extra memory to work with ? What happens if one Database Server is not enough ? If more than one Database Server is used, how they ensure they have same data.No. No extra memory is needed. IF an extra DB server is required, it means a poor design. The architect should focus on doing more with less so that CEOs can keep their private jets and multi-million dollar bonuses.
Q5. Can you give a case scenario or real life example.No. I work on Object-Oriented PL/1.
Thank you for reading my questions.You are welcome/

Disconnected Distributed Computing in Java

Hi, the organization I work for is heavily into JSP/Servlet applications. However, there are a large number of computers which have very slow network links, so they want to find some technology in which we can develop applications which are disconnected, then do a mass synchronization with a Servlet every hour or evening etc.
Is there a Java technology which:
1. application source is centralized and downloadable to client
2. allow for disconnected use of the application
3. synchronize with the server when the application becomes connected with the server
thanks
Don

Jini, is for application to application communication over hetrogenous networks and object protocols. It does not manage distribution of application or synchronization of application data for limited connection applications.
Don

Distributed HR/PY and Check Clearing

Hi experts -
Changing our landscape from a single integrated system to a distributed HR/PY and FICO environment (I.e. 2 different instances). In the distributed architecture I understand the Payroll Check (tcode RFFOUS_C) print program still runs in the HR/PY instance and the check register is also updated in the HR/PY instance. However, the GL posting that include open item managed bank clearing account(s) for payroll checks is updated only in the FICO instance (via ALE) by the Posting to accounting process.
Can anyone tell me how and where the EBS should be run to clear the payroll checks and the GL payroll check clearing open item account? Again, the payroll checks will be in the check register in the HR/PY system but the GL payroll check clearing open item account will be in FICO instance.
Anyone out there have experience with this scenario?
Thanks!

Have you worked specifically with a distributed HR/PY and FICO architecture? Or, are you answering this question based on your understanding of the payment process from FI? If you have not worked hands-on with the distributed HR/PY and FICO architecture I do not think you understand the issue.
In the distributed HR/PY architecture the payroll checks are created in the HR/PY instance. The creation of the GL entry for the payroll check bank clearing account is created by an entirely separate process - the PY GL Posting run and the distributed model posts the GL to the FICO instance. Therefore, the check register is in the HR/PY instance and GL payroll check bank clearing account in in FICO.

Distributed FICO and HR/PY and EBS Questions

Hi experts -
Changing our landscape from a single integrated system to a distributed HR/PY and FICO environment (I.e. 2 different instances). In the distributed architecture I understand the Payroll Check (tcode RFFOUS_C) print program still runs in the HR/PY instance and the check register is also updated in the HR/PY instance. However, the GL posting that include open item managed bank clearing account(s) for payroll checks is updated only in the FICO instance (via ALE) by the Posting to accounting process.
Can anyone tell me how and where the EBS should be run to clear the payroll checks and the GL payroll check clearing open item account? Again, the payroll checks will be in the check register in the HR/PY system but the GL payroll check clearing open item account will be in FICO instance.
Anyone out there have experience with this scenario?
Thanks!

Have you worked specifically with a distributed HR/PY and FICO architecture? Or, are you answering this question based on your understanding of the payment process from FI? If you have not worked hands-on with the distributed HR/PY and FICO architecture I do not think you understand the issue.
In the distributed HR/PY architecture the payroll checks are created in the HR/PY instance. The creation of the GL entry for the payroll check bank clearing account is created by an entirely separate process - the PY GL Posting run and the distributed model posts the GL to the FICO instance. Therefore, the check register is in the HR/PY instance and GL payroll check bank clearing account in in FICO.

WebCenter Content Architecture

Hello everyone,
I just need an advice on how to solve a particular client's document management requirement. Need help on a good (if not best) architecture for their case.
In a nutshell, they have satellite/offsite branches that can receive documents from their customers. They want to get those documents digitized and kept in a doc. mgmt. solution. I've looked into some implementation examples from this site: http://docs.oracle.com/cd/E10316_01/cs/cs_doc_10/implementation/wwhelp/wwhimpl/js/html/wwhelp.htm
but there's no WCC configuration where a content contributor is outside the client's intranet (or extranet). Is it ok to just have one Content Server in the head office/data center, and the branches can connect to WCC (client-server mode) from a different geographical location? Or should there be a branch content server where scanning is available?
Hope you can throw in some ideas.
Thanks,
Jason

Thank you for the explanation! I've commented inline on some of the items.
Well, I don't know. This very much depends on what requirements
> get those documents digitized and kept in a doc. mgmt. solution
and
> a content contributor is outside the client's intranet
really mean.
The process basically is that customers go to either the head office or satellite branches to file an application. In order for the application to be processed, supporting documents must be presented. Currently, the client is photocopying these documents and filing away manually in cabinets. If an application needs to be reviewed again later, they need to sift through their storage.
Since they can accept applications from different locations, scanning and indexing (or in WCC, uploading/checking-in) of documents will be outside the data center where WCC will reside.
a) It is certainly true that WebCenter Content is a centralized, not distributed system.
OK, noted!
b) There are few exceptions (that might be irrelevant to your use case)
Desktop Integration Suite supports also offline mode (for reading, but even updates of documents)
Some news from PM suggest that the new 11.1.1.8 release will support mobile solutions, where content can be ingested even in offline mode (where a mobile device has no signal)
The mobile support will definitely be great as some branches have limited physical space. Although, I might put this in the backburner for now.
However, those scenarios are usually only for exceptions - you cannot expect that a remote site would work with a single DIS client and/or mobile device.c) ODDC is also a centralized solution, but you could have one ODDC server per branch, if necessary, supplemented by ODC. However, ODC/ODDC can be used only for content ingestion (scanning/importing documents, initial metadata). Once a document is committed to a content repository, it is (usually) no longer available. Besides, there is no support for scenarios like searching, revisions, content retrieval, etc.d) Even though, WCC is a centralized system, it does not mean that you could not try to implement a distributed scenario using content migration/replication (Archiver utility). There are, however, two main reasons against it:
Costs - CPU/NUP license model is very effective for a centralized solution, because you can benefit from synergy effects of clustering. I remember a project where we had 80+ branches, altogether with 4K+ users that could run on something like 4 CPUs (eq. to 200 NUPs), if centralized
Manageability - everyone in OCS (Oracle Consulting Services) strongly discouraged us from any distributed design. Imagine a star architecture, where you have one central node containing everything, and a number of smaller branch systems containing items "belonging" to the branch. Unless the product supports a distributed locking (check-out) you can easily end up with inconsistent data. Again, there can be exceptions - in our case, we had just one revision, data were synchronized overnight and it could never happen that data was updated by anyone, but the branch that "owned" them.
In our case, the issue was the internet connectivity, and fortunately, we were finally able to convince the customer to solve the root cause, rather than go with an overkill architecture.
I think the architecture can survive with just one ODDC server where branches (not that many) can connect and send scanned documents to. There are no hard requirements on revisions yet. I also need an application to actually encode the customer's application, and is thinking of a custom-made ADF application, deployed on the WLS where WCC will run on. Or can I just create pages on WCC and expose that?
IMHO,
> a content contributor is outside the client's intranet (or extranet)
is not an argument for a distributed architecture.
OK, understood.
If your concern is security, you may install web server to DMZ. One remaining challenge might be managing user identities, but even for that there are solutions available.
OK, I'll look into this.
Thanks for all the inputs!

EPM System Installation in Distributed Environment...

Hi all,
I am planning to install EPM 11.1.1.1 in a distributed environment. I have one XP machine and two Windows 2003 machines. Can anybody please tell me which components of EPM I put in which machine? Where should I put 11g database and Oracle Application Server 10g?? A response will be appreciated.
Thanks.
Edited by: Sap B on May 12, 2009 1:06 AM

Hi,
Have a read of this post which is very similiar :- Installation of Hyperion Planning in a distributed Architecture
Once you have read it and if there are any areas you don't understand then ask.
Cheers
John
http://john-goodwin.blogspot.com/

DTO, Business Objects and architectural layers design

Hi everybody,
What would prevent one from using Business objects (properties + getters/setters and bevaviour related to a business entity) thru all tiers of a J2EE Web application ?
Example : DAO layer populates BO properties, Business layer uses BO behaviour , web layer stores BO in some scope (request,session,...) for view presentation of the BO properties.
Compared to a DTO/VO (only properties, no behaviour) solution :
- If the application is distributed the BO solution adds serialization overhead (possibl performance bottlenecks) if BO's are large objects.
- If there is requirements for clustering , storing large BO's in session adds session replication overhead.
- BO solution breaks layer isolation , but practically speaking what could be the consequences of it ?
- Some BO properties need formatting for text deiting by the views and cannot be displayed as is.
- BO solution is a more OO approach than using only data objects
What other arguments can you think of ?

"Session" data applies to the session. Thus when I
order a book and it is added to my shopping cart the
session data might be used to store that information.
But that data in of itself is not a "business
s object". It is, by definition, session data.Ok conceptually speaking it is session data, but from
a technical point of view you need an Object instance
of some sort to hold that data.No actually you don't.
There are a number of ways to maintain session data. With tradeoffs between each.
What I was talking about (which does not mean I think
It is a good idea...) is using the BO properties for
holding that data and storing the BO instance(s) in
the session.
And I still don't see that this is a "Business Object". No more so than a security key is a "Business Object".
The best term is session data and you might have different groupings (data not behavior) within that.
Exactly what behavior do you envision?Any computational behaviour that uses the BO
properties and is needed to implement a use case.
This behaviour is used to calculate data that is not
persisted by the DAO as is but rather computed from
m other persisted data.
Give me an exact example.
For example it would be quite reasonable to have a DTO that represents a "rectangle" and that DTO would have a height and width and the DTO would have a method that return the area (multiplying the two data points.)
Conversely my previous example of the zipcode, the DTO should NOT have a method to validate that the zipcode exists although it could validate that it has the correct form (5/9 digits in the US.)
The problem is coupling. If the BO needs the
database layer (see the zipcode example above) then
the GUI layer is going to need the database layer is
well. Even though the GUI will never use the
database layer.I understand that this coupling applies to
distributed architectures where you package your
applications in different jars/wars/ears. In this
case If BO's ared used thru all layers, The GUI
(web) layer needs the data access layer jars even if
it will never use it.
This clearly adds coupling.
But in the case of a collocated application (a single
war) it does not change anything.
No. The point is not how the application is architected.
The point is that the GUI layer is then dependent on the database layer internals.
Consider another example, the DTO is produces a formatted xml string using a XML library. Now you need to create the database layer for the system. The database layer does not need the XML at all. And yet just to compile the database layer you are going to have to provide the XML library. This represents coupling that serves no purpose. And although in very small projects it might be convienent in larger projects coupling like that will cause problems and confusion.
And? Are you suggesting that the BO should do the
formatting?No I mean that DTO/VO's could have either text
formatting/editing methods or property types (e.g
String formatted property for a date) that facilitate
the work of the presentation layer. And most of the time that is going to be incorrect.
"Formatting" for a web page is likely to be different than for a printed report. And for a backend system it might need XML which is still a different format. Adding all of that to the DTO is not the way to do it. (In the same way that the database and gui code should not be intermixed.)
I would not put
any similar formatting behaviour in the BO. That's
why I consider this issue while comparing BO and
DTO's
- BO solution is a more OO approach than using only
data objects
No it isn't.
An object that prints itself is not more OO than one
that is printed.
I thought one of the purpose of OO programming was to
group behaviour and dataCertainly, but it doesn't mandate that one form of grouping is better than another. After if that was the sole concern then why not have just one object for the entire application - that is one way to "group" it.

Parallelising App server architecture

Hi:
I m fairly new to the J2EE technologies. ( thought I ve been reading
about them for the last couple of months or so. )
I m trying to understand how an organisation can scale a J2EE
application, when its load increases. Essentially expanding
breadthwise ( more parallelism ).
Basically the most distributed architecture is this
1. Weserver runs in machine A, it forwards the requests from the
client web browser to Servlet engines running on Machine B. Machine B
delegates the requests to EJB containers running in Machine C. From
here it can go to machine D where database engine is running. Is it
possible to expand this any further.
2. Also in teh above scenario, for web servers ( running in A ) to be
able to forward the requests to Servlet engines ( running in B ), a
small plugin from teh Servlet engine vendor needs to be installed. (
Pl correct me if there is a standards beased/portable request
forwarding mechanism/protocol from the webserver to the Servlet engine
) ( I heard a protocol term with respect to tomcat - can t remember it
now - but it was specific to tomcat IIRC )
3. Once in Servlet the only was I can send a request to EJB running on
a container on machine C, is by using the ejb home object through a
JNDI lookup. Or in other words, there is no automatic mechanism like
teh web server-servlet engine bridge, to forward the requests from a
servlet to its EJB.
4. If 3. above is true, the only way to load balance between multiple
EJB containers running in multiple machines ( C1, C2, C3... ) is by
hosting different home object references on the JNDI tree, and to
fetch the references in some round robin fashion.
5. By using round robin DNS or web server switches ( CISCO or other
hardware that connects to different machines for teh same IP address
in a time shared fashion ), in general it is possible to load share
between multiple web servers, even if client requests belong to
sessions, that might require all the requests belonging to teh same
session go to the same web server.
6. Also though there might be vendor specific mechanisms, how it is
acheived, it is in general for a given web server to load balance
between multiple Servlet engines ( B1, B2, B3 ... )
My ultimate goal is to see if it is possible to have an M:N:O type
configuration ( M web servers, N Servlet engines, O Ejb container
hosting machines )
Any answers even parts of it will be very helpful.
Thanks
--sony

Please don't cross-post.
4. If 3. above is true, the only way to load balance between multiple
EJB containers running in multiple machines ( C1, C2, C3... ) is by
hosting different home object references on the JNDI tree, and to
fetch the references in some round robin fashion.No you don't have to load balance yourself.
When clustering is configured on an EJB, the home object you get from one
server, and it knows what are the other servers. It will load balance the
requests for you. It's called "cluster-aware stubs". I call it "smart
stubs".
===
My personal opinion regarding clustering; don't try to make a complicated
configuration. You need load-balancing? Use up-front (Web) load balancing
and then stick to a single server in the cluster, containing everything.
This will make your everything easier: configuration, deployment, upgrades,
troubleshooting, etc. Access your EJBs locally for more speed.
==
Emmanuel

Disconnected / Distributed Architecture

Similar Messages

Maybe you are looking for