TAO/docs/releasenotes/orbcore.html

   1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
   2 <html>
   3 <head>
   4 <title>ORB Core Status</title>
   5 <!--  -->
   6 </head>
   7 <body text="#000000" bgcolor="#FFFFFF">
   8
   9 <!-- *********************************** -->
  10 <a name="toc"><h2>Table of Contents</h2></a>
  11 <ul>
  12   <li><a href="#current">Current Status</a>
  13   <li><a href="#issues">Known Issues</a>
  14   <li><a href="#ongoing">Ongoing work</a>
  15   <li><a href="#future">Future Work</a>
  16   <li><a href="#completed">Completed Work</a>
  17 </ul>
  18 <!-- *********************************** -->
  19
  20 <hr>
  21 <center>
  22 <h1> <a name="orb"></a>ORB &amp; ORB Core</h1>
  23
  24 Point of contact: <a href="mailto: bala@cs.wustl.edu">Balchandran Natarajan</a>
  25
  26 </center>
  27
  28 <!-- ============================================================ -->
  29 <a name="current"><h3>Current status:</h3></a>
  30
  31 <ul>
  32
  33   <li><p>We are compliant with OMG's CORBA 2.6 specification. </p>
  34   </li>
  35
  36 </ul>
  37
  38 <!-- --------------------- -->
  39 <a href="#toc">Back to TOC</a>
  40 <!-- --------------------- -->
  41
  42 <!-- ============================================================ -->
  43 <a name="issues"><h3>Known issues:</h3></a>
  44
  45 <ul>
  46   <li><p> TAO's support for wstring as per GIOP 1.2 rules are not
  47       perfect. There have been reported interoperability
  48       problems. <p></li>
  49
  50   <li><p> The multi-profile parsing code in TAO hasnt been tested well
  51       and we have been able to identify some issues with it. The
  52       details are documented in our  <A
  53       HREF="http://bugzilla.dre.vanderbilt.edu/index.cgi">bug tracking
  54       system</A> under id's  <A
  55       HREF="http://bugzilla.dre.vanderbilt.edu/show_bug.cgi?id=1237">1237</A>, <A
  56       HREF="http://bugzilla.dre.vanderbilt.edu/show_bug.cgi?id=1238">
  57       1238</A>and<A
  58       HREF="http://bugzilla.dre.vanderbilt.edu/show_bug.cgi?id=1239">1239</A> <p></li>
  59
  60   <li><p> Applications using the distributed callback paradigm could
  61       see crashes if they use multiple nested upcalls. The nested
  62       upcalls sets of a stack growth and crashes when it runs out of
  63       stack space. This happens when an upcall as a part of the
  64       request leads to another request. The client thread waiting for
  65       the reply in the reactor could in turn process yet another
  66       request to build up the stack. The unbounded growth on the stack
  67       leads to crash failures. <p>
  68   </li>
  69
  70
  71 </ul>
  72 <!-- --------------------- -->
  73 <a href="#toc">Back to TOC</a>
  74 <!-- --------------------- -->
  75
  76 <!-- ============================================================ -->
  77 <a name="ongoing"><h3>Ongoing Work:</h3></a>
  78
  79 <ul>
  80   <li><p>Support new concurrency models, e.g., Thread-per-Request,
  81       etc..</p>
  82   </li>
  83   <li><p>Improvement to connection cache on the acceptor side so that
  84       it will purge old connections when there are no more slots or
  85       sockets are available. The connection cache does good job of
  86       purging things on the connector side</p>
  87   </li>
  88
  89   <li><p> Prevent unbounded stack growth that leads to a failure in
  90       situations explained in <a href="#issues">known issues</a>.</p>
  91   </li>
  92
  93 </ul>
  94 <!-- --------------------- -->
  95 <a href="#toc">Back to TOC</a>
  96 <!-- --------------------- -->
  97
  98 <!-- ============================================================ -->
  99 <a name="future"><h3>Future work:</h3></a>
 100
 101 <ul>
 102   <li>Performance bottlenecks:</li>
 103       <ol>
 104         <li><p> Need to identify performance bottlenecks. Looks like
 105         TAO has slowed down a bit over the past two betas. Need to
 106         identify the bottlenecks before improving the performance.</p>
 107         </li>
 108
 109         <li><p> Its a known fact that thread-per-connection is faster
 110         than the reactive model, even for single threaded cases. Need
 111         to identify portions of the reactor that slows down things and
 112         fix them or create a fast path in the ORB just using plain
 113         vanilla select for the server.</p>
 114         </li>
 115
 116         <li><p> In the output data path, the CDR creates the message
 117         block and all other associted things on TSS. If we could
 118         change the CDR to first create them on the stack and then move
 119         the data to TSS we could achieve better performance for small
 120         messages.
 121       </ol>
 122
 123
 124   <li>Features:</li>
 125       <ol>
 126         <li><p>Implement DP-CORBA specification</p></li>
 127
 128         <li><p>Implement TII specification.</p></li>
 129       </ol>
 130
 131 </ul>
 132 <!-- --------------------- -->
 133 <a href="#toc">Back to TOC</a>
 134 <!-- --------------------- -->
 135
 136 <!-- ============================================================ -->
 137 <a name="completed"><h3>Completed Work:</h3></a>
 138
 139 <!-- ************************************************************ -->
 140 <!-- Please make sure you append new items at the end of the list -->
 141 <!-- ************************************************************ -->
 142 <ul>
 143     <li><p> Implemented a connection timeout policy in the ORB. This
 144         is useful for applications to set a timeout value associated
 145         with connection establishment. The name of the policy is
 146         <code>TAO::CONNECTION_TIMEOUT_POLICY_TYPE</code> and is
 147         proprietary to TAO. It comes as a prt of the TAO_Messaging
 148         library. Please see
 149         <code>$TAO_ROOT/tests/Connection_Timeout</code> for an example
 150         of how to use this feature.<p>
 151     </li>
 152
 153     <li><p> Implemented non-blocking connects with three different
 154         strategies for connection establishment. The details of the
 155         strategies and the related documentation can be found <A
 156         href="../Options.html">here</a>.</p>
 157     </li>
 158
 159     <li><p> Deprecated the option -ORBConnectionCachingStrategy. The
 160     option -ORBConnectionPurgingStrategy fills in the role of the
 161     deprecated strategy.</p>
 162     </li>
 163
 164     <li><p> Deprecated the option -ORBResources.</p></li>
 165
 166     <li><p> The connectors have been moved to the lanes.</p> </li>
 167
 168     <li><p> Fixed stack recursion problems that occurs, when a thread
 169         flushing a reply by entering the reactor starts processing
 170         incoming requests.</p>
 171     </li>
 172
 173     <li><p> Fixed connection handling and connection closing problems,
 174         when a write call to a remote host fails. </p>
 175     </li>
 176
 177     <li> <p> Moved the allocators used for allocating memory for the
 178         incoming data path to the lanes. Only allocators on the
 179         incoming data path are moved into the lanes. Allocators on the
 180         outgoing data path are still on a per-orb basis. Since the
 181         outgoing data path uses memory from the TSS, moving them to
 182         the lanes is not necessary.
 183         </p>
 184     </li>
 185     <li><p> Added an option, -ORBSingleReadOptimization that allows
 186         users to turn on or turn off the single read
 187         optimization. This is useful for RTCORBA server threads to
 188         avoid priority inversions. Buffering messages and sending
 189         notifications to the reactor, that does not know or honour
 190         priority could lead to priority inversions.
 191         </p>
 192     </li>
 193
 194     <li><p> Several changes were made to utilize the new memory management mechanisms
 195     of Event Handlers in TAO.  Previously, both the connection handlers and
 196     the transports had their own reference counting mechanism and
 197     independent lifetimes.  The reference counting was inadequate
 198     since it was not originated at the Reactor and the independent
 199     lifetimes unnecessarily completed protocol specific transport
 200     code.  <p>
 201
 202     The changes couple the lifetime of the connection handler with
 203     that of the transport and utilizes the reference count in the
 204     base ACE_Event_Handler class.  This allowed us to remove several
 205     hacks throughout TAO code which were designed to prevent (or
 206     delay) crashes.  Several extra checks and unnecessary locks were
 207     also removed. <p>
 208
 209     All of the following protocols were updated: DIOP, IIOP,
 210           SCIOP, SHMIOP, SSLIOP, UIOP, UIPMC. </p>
 211     </li>
 212
 213     <li><p> Support for the ORB methods create_value_tc, create_component_tc,
 214             create_event_tc, create_home_tc and create_recursive_tc added. </p>
 215     </li>
 216
 217 </ul>
 218 <!-- --------------------- -->
 219 <a href="#toc">Back to TOC</a>
 220 <!-- --------------------- -->
 221
 222 </BODY>
 223 </HTML>