1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
|
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="GENERATOR" CONTENT="Mozilla/4.5b2 [en] (WinNT; I) [Netscape]">
<TITLE>ORB Core Status</TITLE>
<!-- $Id$ -->
</HEAD>
<BODY TEXT="#000000" BGCOLOR="#FFFFFF">
<H3>
<A NAME="orb"></A>ORB & ORB Core</H3>
Point of contact: <A HREF="mailto:nanbor@cs.wustl.edu">Nanbor Wang</A>
<P>Last Update: $Date$
<P><B>Current status:</B>
<UL>
<LI>
Provides dynamically linked wholesale assignment of ORB strategies for
communication, concurrency, demultiplexing, scheduling, and dispatching
using the ACE Service Configurator framework.</LI>
<LI>
Supports several concurrency models out of the box: Reactive, Thread-per-Connection,
and ORB-per-thread. The ORB-per-thread design is optimized for real-time
I/O subsystems. (See: <A HREF="../configurations.html">Configuration Guide</A>.)</LI>
<LI>
Supports <TT>resolve_initial_references("<CORBA Service>")</TT> using
ACE support for multicast.</LI>
</UL>
<B>Known issues:</B>
<UL>
<LI>
Some CORBA objects need to have access to the ORB. We tend to use
<CODE>TAO_ORB_Core_instance()->orb()</CODE> to fetch the right ORB the
object is using. This is most often true when implementing the
shutdown method which allow other process to shutdown a remote
ORB. Althought this may be the "right" thing to do in TAO, it is not
CORBA compliant and hurts the portability of object implementation. A
common way of getting around this problem is to cache the ORB pointer
in the object when it is created and simply use the cached ORB when
needed. However, remember that there could be multiple ORBs in a
process and a POA may be shared by several ORBs. That means, we may
not handling the object with the same ORB it was created. Collocation
optimization further complicated the problem. <P> At the moment, using
<CODE>TAO_ORB_Core_instance()->orb()</CODE> is the only sane way to
get arond the problem. Though there may be a better way. <P>
<LI> Object references for objects dynamically created within a server
may not have proper host. (See also <A HREF="#hostfix">recent
fix</A>.) When an object is dynamically created and its object
reference returned, it simply has to "guess" at the host information
that is returned in the IIOP Profile. Currently, it guesses using the
host name, which is a reasonable guess in most cases. However, if
there are multiple interfaces on a machine <I>and</I> they have
different host names, it is possible that the desired effect will not
be realized, i.e., the connection to that new object may use a
different interface from the original object. Currently the only way
to change this is to use the <A
HREF="../Options.html#-ORBhost"><B>-ORBhost</B> option</A> to manually
specify that dynamically-created objects bind to a specific interface
so that their IIOP Profile has the desired hostname. A more desirable
fix is being investigated.<P> </UL>
<B>Recently Completed Work:</B>
<UL>
<LI>
The ORB Core is ready to throw exceptions received from the wire as a native
C++ exception. Similarly, on the server side, native C++ exceptions are
catched and sent over the wire to the client.</LI>
<LI>
Added new option <TT>-ORBgioplite</TT> when this option is enabled the
ORB removes a few fields of the GIOP messages that are normally not used,
such as the version numbers, the magic 'GIOP' string, the service context,
the Principal object, etc. This option give us a slight performance improvement
but it is important for extremely low-latency systems (such as avionics),
that fortunately have tightly controlled environments so this kind of optimization
is safe to use.</LI>
<LI>
The buffers for outgoing CDR streams are allocated from TSS memory pools,
reducing locking on the critical path.</LI>
<LI>
Several optimizations on the critical path have been implemented, carefully
removing excesive locking, memory allocations and data copying. In some
cases special demarshaling and marshaling functions where written for the
request headers.</LI>
<LI>
Adding a new option into TAO_Resource_Factory to control the internal lock
used by ORB's reacotr. With "<TT>-ORBreactorlock null</TT>" added into
<TT>svc.conf</TT>, you can eliminate the token inside the Select_Reactor
inside the ORB even on multi-threaded platforms. This eliminate some overheads
caused by synchronizing access to the Select_Reactor. Usually, if you have
TSS ORB and are using reactive concurrency startegy, the ORB are not accessed
by multiple threads, then, you can safely turn the lock off.
<LI>
Strategized the connection management scheme so that we don't always have
to pay the price of a look up in the connection cache. Basically, we "hold"
a service handler for a string of requests.</LI>
<LI>
There is no need to call ORB::open if the port is 0. It will automagically
be called when the first stub is created.
<LI> The ORB now handles nested upcalls correctly. See <A
HREF="../leader_follower.html">this </A>for details on the design of
the solution.
<LI> Making the collocation table configurable. We can either use a
global collocation table to share objects among ORBs, or use the
per-ORB collocation table.
<LI>
Implement the inheritance mechanism in Default_Server_Connection_Handler
to access the parent's POA. This is necessary when we are using thread-per-connection
concurrency model.
<LI> It is often desireable for a server to start up and not have to
specify the TCP port number on which the ORB will listen, i.e., let
the OS decide. While this works properly (by providing an argument
<TT>-ORBport 0</TT>), the actual port number used by the server won't
be available after the server starts listening. TAO now tries to bind
the address when <TT>-ORBport 0</TT> is issued so it can get the
actual address to publish the IOR for clients' use.
<LI> Added optimization for collocated objects. This optimization
enables TAO to bypass communication layers and talks to collocated
objects directly.
<LI>
Broke the tight coupling between <TT>CORBA_Object</TT> and <TT>IIOP_Object</TT>
so that multiple <TT>CORBA_Object</TT>s can share a single <TT>IIOP_Object</TT>.
This has a big win in <TT>_narrow</TT> operations.
<LI>
Eliminated substantial memory leaks in servers.
<LI>
Added documentation for <B><A HREF="../Options.html#-ORBpoa">-ORBpoa</A></B>,
which allows individual tweaking of the POA's locality (global vs. thread-specific)
independent of the locality of other ORB resources. Also, <B><A HREF="../Options.html#-ORBpoalock">-ORBpoalock</A></B>
and <B><A HREF="../Options.html#-ORBpoamgrlock">-ORBpoamgrlock</A></B>
control locking in the POA and POA manager, respectively, according to
application needs.
<LI>
Began the process of changing the internals of TAO to use the "underbar"
namespace mapping. This should be completed soon.
<LI>
The ORBs manipulation of object keys and object ids is now consistent.
Canonically, everything is managed as an opaque sequence of octets. Conversion
to/from strings is permitted, with non-printables turning into a trigraph
<B>\hl</B> where <B>h</B> is the high nibble digit in hex, and <B>l</B>
is the low nibble in hex.</LI>
<LI>
<TT>CORBA_ORB::run()</TT> should be able to be called multiply in the same
execution path now.</LI>
<LI>
Fixed the <B>LOCATION_FORWARD</B> capabilities to work as described in
the CORBA 2.1 and the POA specification.</LI>
<LI>
Fixed problem in the IDL Cubit example which caused the server to dump
core when dealing with <TT>_out</TT> sequence types <I>ONLY</I> when compiled
with G++. It seems that G++ decided to interpret the expression <TT>output
== 0</TT> (where <TT>output</TT> is of type <TT>vector_out</TT>) differently
from Sun's C++ compiler.</LI>
<LI>
Fixed a slight problem in <A HREF="#hostfix">other recent work</A> where
a servant listening on <TT>INADDR_ANY</TT> reported an IOR with address
<TT>0.0.0.0</TT>. It now reports the default hostname instead.</LI>
<LI>
Working version of TAO on VxWorks with VME transport.</LI>
<LI>
TAO now reacts gracefully in the face of a non-existent <TT>svc.conf</TT>
file or simply missing entries in the file. It logs a warning message and
uses default values versions of items which it would normally obtain from
the Service Repository.</LI>
<LI>
<A NAME="hostfix"></A>Fixed problem where object references for objects
dynamically created within a server do not have proper host, at least if
the desired host was specified using <B><A HREF="../Options.html#-ORBhost">-ORBhost</A></B>
on the command line.</LI>
<LI>
Added support for <B><A HREF="../Options.html#-ORBsndsock">-ORBsndsock</A></B>
and <B><A HREF="../Options.html#-ORBrcvsock">-ORBrcvsock</A></B> ORB options.
See <A HREF="../Options.html">options documentation</A> for details.</LI>
<LI>
Added support for pre-established connections using the <B><A HREF="../Options.html#-ORBpreconnect">-ORBpreconnect</A></B>
option. This allows connections to be esablished before any CORBA requests
are made. See <A HREF="../Options.html">options documentation</A> for details.</LI>
<LI>
Eliminated deprecated <B>-OAxxx</B> options parsed by the Object Adapter
<TT>init</TT> method but no longer used by part of the ORB.</LI>
<LI>
Made argument parsing within the ORB and POA consume the arguments it parses.</LI>
<LI>
Placed an optimization in IIOP::Profile that should speed up invocations
on the client side rather substantially. This was done by caching an instance
of <TT>ACE_INET_Addr</TT> in the private portion of <TT>Profile</TT> because
the cost of constructing one for every invocation (in <TT>TAO_GIOP_Invocation::start</TT>)
was simply enormous--the construction was something like 100 cycles on
Solaris. This is all because deep down this ends up calling <TT>gethostbyname()</TT>,
which can be very costly. Of course, this is platform-dependent, so the
improvements may not be as great on a different platform.</LI>
<LI>
Got rid of need to have compile in TSS ORB Core (<TT>-DTAO_HAS_TSS_ORBCORE</TT>).
This is accomplished by having a Resource Factory (<TT>TAO_Resource_Factory</TT>)
singleton which the ORB Core consults in order to get pointers to resources.
The Resource Factory can be configured to return pointers from thread-specific
storage, or process-global storage. By doing this, the ORB Core is shielded
from ever having to know whether or not its resources are global or not.</LI>
<LI>
Made all command-line like options look like the standard-professed options,
i.e., <B>-ORBxxx</B> or <B>-OAxxxx</B>.</LI>
<LI>
Moved ORB option parsing into <TT>TAO_ORB_Core::init()</TT> rather than
in <TT>ORB_init()</TT>.</LI>
<LI>
Define <TT>_FAR</TT> for all cases as an interim fix for LynxOS.</LI>
<LI>
Fixed TAO so that the default port of 5001 is defined in <TT>orb_core.h</TT>
config header somewhere rather than in <TT>orbobj.cpp</TT>, and use <B>MAXHOSTNAMELEN</B>
ACE
constant for the hostname buffer.</LI>
<LI>
Eliminated need for <TT>CORBA::Object</TT> to keep a pointer to its orb
sitting around anywhere.</LI>
<LI>
Tested <TT>CORBA_ORB::shutdown()</TT> as a manner in which a thread of
control can request that <TT>CORBA_ORB::run()</TT> should return and no
longer service requests. It only honors a <I>wait_for_completion</I> argument
of <B>FALSE</B> right now because we haven't fully explored the issues
which surround doing all the completion.</LI>
<LI>
Created a brief section in <TT><A HREF="../../TAO-INSTALL.html">$TAO_ROOT/TAO-INSTALL.html</A></TT>
that <I>briefly</I> documents installation on Unix platforms. There is
also a brief section which documents issues that apply across ALL platforms,
such as the fact that TAO and its application will likely not work well
if your ACE is linked through Orbix, VisiBroker, or another ORB. <A HREF="../../ChangeLog">ChangeLog</A>
discovery attribution for this should go to <I>Steve Wohlever <wohlever@mitre.org></I>.</LI>
<LI>
Implemented <TT>CORBA_ORB::shutdown()</TT> as a manner in which a thread
of control can request that <TT>CORBA_ORB::run()</TT> should return and
no longer service requests.</LI>
<LI>
Validated movement of Strategy Connector into ORB Core.</LI>
<LI>
Tested and commited <TT>Connector::connect_n</TT> changes to ACE.</LI>
<LI>
ACE & TAO compiling cleanly using Sun's CC on Solaris and G++ on Linux.
Also, the ACE tests run properly.</LI>
<LI>
Use <TT>truss</TT> to verify for <A HREF="mailto:PAUL.G.WEHLAGE@cdev.com">Paul
Wehlage</A> that TAO really uses <TT>select</TT> vs. <TT>recv</TT> for
<B>-R</B> vs. <B>-T</B>.</LI>
<LI>
Renamed <TT>boa.*</TT> to <TT>poa.*</TT>.</LI>
<LI>
Renamed following files: Orb_Core.* ->orb_core.* & orbobj.* -> corba_orb.*.</LI>
<LI>
The lock used in <TT>giop.cpp:626</TT> originally appeared in the IIOP-1.4
code base. Its purpose was to protect access to the <TT>fwd_profile</TT>
data member. The original features are available, but they are currently
not used because none of our concurrency models present a thread-unsafe
condition. The <A HREF="../../ChangeLog">ChangeLog</A> presents more details
on the subject.</LI>
<LI>
Fixed problems seen by <A HREF="mailto:brian.r.mendel@boeing.com">Brian
Mendel</A> of Boeing where the Cubit server application was crashing in
<TT>TAO_OA_Connection_Handler::handle_message</TT> when it was trying to
call <TT>CORBA::POA::handle_request</TT> through a NULL <TT>root_poa</TT>
pointer. The NULL pointer resulted from the inadvertent use of the Thread-per-Connection
concurrency model, which is currently incompatible with the ORB-per-thread
model. It was not caught because the change is normally done in configuration
files parsed by the Service Configurator, and VxWorks requires special
code in that area which was not tested until Boeing got hold of the code.</LI>
<LI>
Clean up build under g++. All 'function used before declared inline' warnings
have been eliminated. Also, all other warnings (except for one that is
nearly unavoidable) have been eliminated.</LI>
<LI>
Removed BOA and ROA names from the ORB.</LI>
<LI>
Changed <TT>CORBA_ORB::open()</TT> to return an error status (first cut
at this already there--needs to be checked out).</LI>
<LI>
Added README file to multiCubit test.</LI>
<LI>
Added support for the -ORBInitRef option. This allows run-time specification of
an ObjectId:IOR mapping. It is used by resolve_initial_references () and
overrides the orb-install-time defaults.</LI>
</UL>
<B>Ongoing Work:</B>
<UL>
<LI>
Verify ACE & TAO on Linux using g++, Solaris using CC, and Solaris
using g++.</LI>
<LI>
Use Purify to eliminate straggling problems.</LI>
<LI>
Clean up the internals and analyze for performance bottlenecks using Quantify.</LI>
<LI>
Resolve <TT>// @@</TT> comment blocks. These are questions or comments
placed in the code by a reviewer (typically Doug).</LI>
<LI>
Make more and better documentation on the internals of TAO.</LI>
</UL>
<B>Current Work:</B>
<UL>Verify that, in the ORB-per-thread concurrency model, it's possible
to have a factory in each thread that creates a client and returns an object
reference to that client within that thread. This is different from the
model that Sumedh originally had where the factory was trying to turn out
a client in a separate thread/ORB combination.
There is no application which attempts this at the moment, so both a
client and a server will need to be manufactured. Using existing Cubit-like
examples (that poor app is so weary) should reduce development time to
a couple of hours.
The real strategy/solution is to make <TT>CORBA::Object</TT> not keep
a pointer to its orb sitting around anywhere. It should only get it when
it needs it. This should not be difficult to change.</UL>
The interface Dynamic
Any (CORBA v.2.2) is being implemented. It will conform to and implement
all areas of the spec except for the following:
<UL>
<LI>
Interface DynFixed not implemented since TAO does not currently support
the IDL fixed data type.</LI>
<LI>
Long double data type not supported in accessor and mutator functions.</LI>
<LI>
Wstring (wide character string) data type not supported in accessor and
mutator functions.</LI>
<LI>
DynUnion attribute set_as_default in treated as read only.</LI>
</UL>
<B>Critical Work:</B>
<UL>
<LI>
Identify/fix the problems in multithreaded client with TSS resources. This
could be done using the leader-follower model (thread-pool.)</LI>
<LI>
Support new concurrency models, e.g., Thread-per-Request, thread pool,
etc..</LI>
</UL>
<B>Future work:</B>
<UL>
<LI>
Performance optimizations, such as:</LI>
<OL>
<LI>
Replace all uses of <TT>new</TT> and <TT>delete</TT> with allocators that
can be put into TSS. The main motivation for this is to eliminate lock
contention when multiple threads need to allocate/deallocate memory, however
it might also be possible to substitute an allocator better-tuned to an
application's specific usage.</LI>
</OL>
<LI>
Robustness improvements:</LI>
<OL>
<LI>
Improve connection cache so that it won't crash when there are no more
slots or sockets.</LI>
<LI>
Add an option to the ORB core, e.g., <B>-ORBnameservicetimeout</B>, to
allow specification of an upper bound on the itime-to-wait for a response
from the ORB's multicast "search" for a running name server.</LI>
</OL>
<LI>
New features:</LI>
<OL>
<LI>
Implement <TT>ORB::perform_work</TT> and <TT>ORB::work_pending</TT> so
that <TT>ORB::run</TT> <I>could</I> be implemented in terms of them.</LI>
<LI>
Improve the <TT>Service Configurator</TT> service entry faking on VxWorks.</LI>
<LI>
Integrate with realtime upcalls (RTUs).</LI>
</OL>
<LI>
Internal restructuring:</LI>
<OL>
<LI>
Create some sort of internal "request" object which carries enough context
to suport deferred method invocation such as one might find in the thread-pool
approach or wherever queuing might occur.</LI>
</OL>
<LI>
Documentation:</LI>
<OL>
<LI>
Update ORB Patterns paper to reflect the latest TAO has to offer. Specifically</LI>
<UL>
<LI>
sections on the Abstract Factory and Concurrency strategies need to be
updated, and the entire paper needs to be reviewed for accuracy w.r.t the
current TAO.</LI>
<LI>
Metrics must be re-computed.</LI>
<LI>
Tighten existing discussions.</LI>
<LI>
Add information on new patterns (such as TSS) which have been used since
initial work.</LI>
</UL>
<LI>
Make the section in <TT><A HREF="../../TAO-Install.html">$TAO_ROOT/TAO-Install.html</A></TT>
that documents installation on Unix platforms more complete.</LI>
</OL>
<LI>
Miscellany:</LI>
<OL>
<LI>
Assist in porting higher level services (e.g., <A HREF="ec.html">Event
Channel</A>) to TAO.</LI>
<LI>
Compile and verify ACE and TAO on Linux using the Kuck and Associates,
Inc. (KAI) compiler tools.</LI>
</OL>
<LI>
Potential Optimizations:</LI>
<OL>
<LI>
There is a certain amount of overhead introduced by the use of TSS Singletons.
TSS singletons are very convenient as the variable does not have to be
passed from function to function. Even though sort of access is generally
acceptable, the overhead of this approach maybe too high when used in code
which is executed frequently (e.g., during an upcall). This critical code
may have to be restructured not to use TSS Singletons because of the high
overhead associated with it.</LI>
</OL>
</UL>
</BODY>
</HTML>
|
