EJB BioNet Integration

1
EJB BioNet Integration

• Implementing EJB via agent-based BioNet
  architecture

2
EJB 101

• Based on factory design pattern (separates
  implementation and interface of object).

List objList ListFactory.newList() objList.add(
1, new Integer(1)) int nJustStored
objList.get(1)
User code
public List newList() return (List)(new
ListAsArray())
Factory class
public interface List public void add(int
index, Object toAdd) public Object get(int
index)
Interface
public class ListAsArrray implements List . . .

Implementation
3
EJB 101

• In addition to the factory design pattern
  facilitating object distribution, EJB also
  specifies several different types of remote
  objects (beans).
• There are two major categories session beans
  and entity beans.

4
Session Beans

• Session beans are designed to perform some sort
  of business task.
• Examples a credit card charging bean, a
  order-placing bean, or a interest-calculating
  bean.
• Two kinds of session beans stateless and
  stateful session beans.

5
Stateless Session Beans

• Do not retain state information between method
  calls.
• Marking a session bean stateless allows EJB
  container manage bean to meet demand for
  effectively.
• Although user code thinks it has the same bean
  instance, the instance may be shared, serialized,
  or even changed between method calls.

6
Entity Beans

• Support object persistence. Entity beans do not
  disappear during container startup/shutdowns/crash
  es.
• Home interface supports finding entity beans by
  some attribute.
• Can be thought of as a collection of objects that
  never disappear, often implemented with a SQL
  database by the EJB container.

7
EJB Problems

• Although EJB containers often implement
  load-balancing techniques, EJB has limited
  internet-scope because of the number of network
  hopes needed to do one remote method invocation,
  even to a server farm.
• In addition, EJB containers usually must be
  extremely robust, scalable, and reliable but,
  this means that objects have to be isolated on
  systems far from user systems.

8
BioNet Offers Solutions

• Since the container layer can take action
  transparently to the user code, specialized
  BioNet based techniques can be applied
  transparently.
• Main goal To improve performance of remote
  method calls without sacrificing robustness or
  scalability.

9
Migration Services

• An EJB object is often called. Since each call
  results in network trips, migrating a frequently
  called object to the callers system provides
  large performance gains.
• The BioNet architecture provides adaptive
  techniques for deciding when to migrate an
  agent, or in this case, possibly a remote EJB
  object.
• BioNet also provides facilities for performing
  the object/agent migration.

10
Reproduction Services

• If more than one network host needs to access a
  given EJB object, possibly the reproduction
  facilities of BioNet can be employed to allow a
  bean to reproduce and move to both systems.
• Again, the BioNet adaptive ability can help tune
  the EJB object reproduction to correctly scale to
  user demand.

11
Relationship Services

• However, reproduction of EJB objects is a
  double-edged sword.
• Provides better performance and results in less
  network bandwidth consumption.
• Also provides robustness if one EJB object
  container dies, the other reproduced object still
  exists.
• But, the potential for locking and coherency
  problems now exists.

12
Relationship Services - Locking

• By having the reproduced EJB objects maintain
  relationships with one another, and by creating a
  lock entity when an EJB object is initially
  reproduced, these problems can be handled.
• Since lock entity is unique to each EJB object
  instance, and is a BioNet entity itself, locality
  is maintained and optimal locking ability can be
  achieved.
• Loss of lock entity is easily handled since all
  EJB objects stay in relationship.

13
Relationship Services Robust

• By having co-reproduced EJB objects all stay in
  relationship with one another, a minimum
  population for a given EJB object instance can be
  maintained.
• Minimum population will reside on different
  systems, meaning that a system failure will not
  result in loss of data.

14
Migration Services - Robust

• Migration services also provide robustness
  because entity beans no longer reside on a single
  group of close systems.
• Rather, data is distributed across large numbers
  of systems, and over time, will migrate to the
  most optimal distribution for the average network
  load.

15
High-level Implementation

• Every BioNet platform can serve as a EJB
  container.
• Every EJB object that is created has an
  associated BioNet entity that is responsible for
  the EJB object.
• The EJB object stub probably will communicate
  with this BioNet entity to allow the entity make
  decisions.
• The BioNet entity will most likely communicate
  with the EJB object skeleton to allow the EJB
  object to migrate, etc.

16
Summary

• BioNet and EJB have very similar goals.
• BioNet concepts easily apply to an implementation
  of EJB container middleware.
• Goal implemented a distributed object technology
  effectively using an distributed entity-based
  solution.