TAO/docs/tutorials/Quoter/AMI/index.html

   1 <!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML//EN">
   2 <html>
   3   <head>
   4     <title>Asynchronous Method Invocation - CORBA for impatient clients</title>
   5     <!--  -->
   6   </head>
   7
   8   <BODY text = "#000000"
   9     link="#000fff"
  10     vlink="#ff0f0f"
  11     bgcolor="#ffffff">
  12
  13     <h3>Asynchronous Method Invocation - CORBA for impatient clients</h3>
  14
  15     <P>Our <A HREF="../Simple/Client/index.html">simple client</A>
  16       illustrated how to use the traditional CORBA synchronous method
  17       invocation to query a number of stock prices.
  18       Suppose that we had to query hundreds of stock prices, for
  19       example, during the initialization of a complex market analysis
  20       tool.
  21       In that case sending the requests in sequence is going to yield
  22       poor performance; we are not taking advantage of the natural
  23       parallelism in distributed systems, since we are waiting for the
  24       first response to come back before sending the next query.
  25       Traditionally this problem has been attacked using either
  26       <CODE>oneway</CODE> calls or multiple threads.  Both approaches can
  27       work, but they have some disadvantages:
  28       multi-threading programming can be hard and error-prone,
  29       oneways are unreliable and require callback interfaces to return
  30       the stock value.
  31       Recently the OMG approved the CORBA Messaging specification
  32       that extends the basic invocation model to include asynchronous
  33       calls.
  34       Unlike the old deferred synchronous model, the new model uses
  35       the IDL compiler and the SII to achieve type safety and improved
  36       performance, but the application does not block waiting for a
  37       response. Instead, it gives the ORB a reference to a reply
  38       handler that will receive the response asynchronously.
  39       The specification also defines a polling interface that we will
  40       not discuss as TAO does not implement it.
  41     </P>
  42
  43     <P>For this problem we will extend the IDL interface to include a
  44       new operation:
  45     </P>
  46 <PRE>
  47   interface Single_Query_Stock : Stock {
  48     double get_price_and_names (out string symbol,
  49                                out string full_name);
  50   };
  51 </PRE>
  52     <P>This will simplify some of the examples.
  53     </P>
  54
  55     <P>The first step is to generate the stubs and skeletons with
  56       support for the callback AMI.  We do this with the <CODE>-GC</CODE>
  57       flag:
  58     </P>
  59 <PRE>
  60 $ $TAO_ROOT/TAO_IDL/tao_idl -GC Quoter.idl
  61 </PRE>
  62     <P>You may want to take a brief look at the generated client side
  63       interface.
  64       The IDL compiler adds new methods to the <CODE>Quoter::Stock</CODE>
  65       interface.  In particular pay attention to this operation:
  66     </P>
  67 <PRE>
  68   virtual void sendc_get_price_and_names (
  69       AMI_Single_Query_StockHandler_ptr ami_handler
  70     );
  71 </PRE>
  72     <P>This is the operation used to send a request asynchronously.  The
  73       response is received in the handler object. This is a regular
  74       CORBA object with the following IDL interface:
  75     </P>
  76 <PRE>
  77 interface AMI_Single_Query_StockHandler {
  78   void get_price_and_names (in double ami_return_val,
  79                            in string symbol,
  80                            in string full_name);
  81 };
  82 </PRE>
  83     <P>You don't have to write this IDL interface.  The IDL compiler
  84       automatically generates the so-called <EM>implied IDL</EM>
  85       constructs from your original IDL.
  86       Notice how the arguments are generated.  The first argument is
  87       simply the return value, then the output arguments show up, but
  88       as <EM>input</EM> only since the handler has to receive the
  89       reply.
  90     </P>
  91
  92     <H3>Implementing the reply handler</H3>
  93
  94     <P>We will have to implement a servant for this new IDL interface
  95       so we can receive the reply,
  96       exactly as we do for servers:
  97     </P>
  98 <PRE>
  99 class Single_Query_Stock_Handler_i : public POA_Quoter::AMI_Single_Query_StockHandler
 100 {
 101 public:
 102   Single_Query_Stock_Handler_i (int *response_count)
 103     : response_count_ (response_count) {}
 104
 105   void get_price_and_names (CORBA::Double ami_return_val,
 106                            const char *symbol,
 107                            const char *full_name)
 108   {
 109     std::cout << "The price of one stock in \""
 110               << full_name << "\" (" << symbol << ") is "
 111               << ami_return_val << std::endl;
 112     *this->response_count_++;
 113   }
 114
 115 private:
 116   int *response_count_;
 117 };
 118 </PRE>
 119     <P>The <CODE>response_count_</CODE> field will be used to
 120       terminate the client when all the responses are received.
 121     </P>
 122
 123     <H3>Sending asynchronous method invocations</H3>
 124
 125     <P>The handler servant is activated as any other CORBA object:
 126     </P>
 127 <PRE>
 128     int response_count = 0;
 129     Single_Query_Stock_Handler_i handler_i (&response_count);
 130     Quoter::AMI_Single_Query_StockHandler_var handler =
 131       handler_i._this ();
 132 </PRE>
 133     <P>and now we change the loop to send all the requests at once:
 134     </P>
 135 <PRE>
 136     int request_count = 0;
 137     for (int i = 2; i != argc; ++i) {
 138       try {
 139         // Get the stock object
 140         Quoter::Stock_var tmp =
 141           factory->get_stock (argv[i]);
 142         Quoter::Single_Query_Stock_var stock =
 143           Quoter::Single_Query_Stock::_narrow (tmp.in ());
 144
 145         stock->sendc_get_price_and_names (handler.in ());
 146         request_count++;
 147       }
 148 </PRE>
 149     <P>after the loop we wait until all the responses have arrived:
 150     </P>
 151 <PRE>
 152     while (response_count < request_count
 153            && orb->work_pending ()) {
 154       orb->perform_work ();
 155     }
 156 </PRE>
 157
 158     <H3>Exercise 1</H3>
 159
 160     <P>Complete the <CODE>client.cpp</CODE> file.  Does this client play
 161       the server role too?  If not, what is the role with respect to
 162       the handler servant?  If you think it is a server too, what
 163       should you do about the POA?
 164     </P>
 165     <P>You can use the following files to complete your implementation:
 166       the <A HREF="Quoter.idl">Quoter.idl</A>,
 167       <A HREF="Handler_i.h">Handler_i.h</A>,
 168       <A HREF="Handler_i.cpp">Handler_i.cpp</A>.
 169       Remember that the simple client main program
 170       (located
 171       <A HREF="../Simple/Client/client.cpp">here</A>)
 172       is a good start.
 173     </P>
 174
 175     <H4>Solution</H4>
 176     <P>Look at
 177       <A HREF="client.cpp">client.cpp</A> file.  It should
 178       not be much different from yours.
 179     </P>
 180
 181     <H3>Testing</H3>
 182
 183     <P>A simple server is provided, based on the
 184       <A HREF="../Simple/Server/index.html">simple server</A>
 185       from the introduction.
 186       As before, you need the following files:
 187       <A HREF="Stock_i.h">Stock_i.h</A>,
 188       <A HREF="Stock_i.cpp">Stock_i.cpp</A>,
 189       <A HREF="Stock_Factory_i.h">Stock_Factory_i.h</A>
 190       <A HREF="Stock_Factory_i.cpp">Stock_Factory_i.cpp</A>
 191       and <A HREF="server.cpp">server.cpp</A>.
 192     </P>
 193
 194     <H2>Configuration</H2>
 195
 196     <P>So far we have used the default configuration in TAO,
 197       but AMI works better with some fine tuning.  For example,
 198       by default TAO uses a separate connection for each outstanding
 199       request.  With SMI this is a very good strategy, as separate
 200       threads can send concurrent requests without any shared
 201       resources,
 202       but this approach does not scale well with AMI, as it would
 203       create too many connections.
 204       The solution is to change the strategy to share connections.
 205       All we need to do is create a
 206       <A HREF="svc.conf">svc.conf</A> file with the following
 207       contents:
 208     </P>
 209 <PRE>
 210 static Client_Strategy_Factory "-ORBTransportMuxStrategy MUXED"
 211 </PRE>
 212     <P>There are many other configuration options, all of them
 213       documented in
 214       <A HREF="../../../Options.html">Options.html</A>,
 215       in
 216       <A HREF="../../../configurations.html">configurations.html</A>,
 217       and in the Developer's Guide available from
 218       <A HREF="http://www.theaceorb.com/">OCI</A>.
 219     </P>
 220
 221     <hr>
 222     <address><a href="mailto:coryan@cs.wustl.edu">Carlos O'Ryan</a></address>
 223   </body>
 224 </html>