diff --git a/ompi/communicator/comm_init.c b/ompi/communicator/comm_init.c
index 914f58a7119..d3f24115ad7 100644
--- a/ompi/communicator/comm_init.c
+++ b/ompi/communicator/comm_init.c
@@ -219,6 +219,7 @@ ompi_communicator_t *ompi_comm_allocate ( int local_size, int remote_size )
 
     /* create new communicator element */
     new_comm = OBJ_NEW(ompi_communicator_t);
+    Tau_start_class_allocation(new_comm->c_base.obj_class->cls_name, 0, 0);
     new_comm->c_local_group = ompi_group_allocate ( local_size );
     if ( 0 < remote_size ) {
         new_comm->c_remote_group = ompi_group_allocate (remote_size);
@@ -234,6 +235,7 @@ ompi_communicator_t *ompi_comm_allocate ( int local_size, int remote_size )
 
     /* fill in the inscribing hyper-cube dimensions */
     new_comm->c_cube_dim = opal_cube_dim(local_size);
+    Tau_stop_class_allocation(new_comm->c_base.obj_class->cls_name, 0);
 
     return new_comm;
 }
diff --git a/ompi/group/group_init.c b/ompi/group/group_init.c
index 2155d262470..c0c50fb09b6 100644
--- a/ompi/group/group_init.c
+++ b/ompi/group/group_init.c
@@ -57,7 +57,11 @@ ompi_predefined_group_t *ompi_mpi_group_null_addr = &ompi_mpi_group_null;
 ompi_group_t *ompi_group_allocate(int group_size)
 {
     /* local variables */
+    Tau_start_class_allocation("ompi_group_t", 0, 0);
     ompi_proc_t **procs = calloc (group_size, sizeof (ompi_proc_t *));
+    Tau_start_class_allocation("ompi_proc_t **", group_size * sizeof(ompi_proc_t *), 0);
+    Tau_stop_class_allocation("ompi_proc_t **", 1);
+    Tau_stop_class_allocation("ompi_group_t", 0);
     ompi_group_t *new_group;
 
     if (NULL == procs) {
@@ -81,6 +85,7 @@ ompi_group_t *ompi_group_allocate_plist_w_procs (ompi_proc_t **procs, int group_
 
     /* create new group group element */
     new_group = OBJ_NEW(ompi_group_t);
+    Tau_start_class_allocation(new_group->super.obj_class->cls_name, 0, 0);
 
     if (NULL == new_group) {
         return NULL;
@@ -105,7 +110,7 @@ ompi_group_t *ompi_group_allocate_plist_w_procs (ompi_proc_t **procs, int group_
     OMPI_GROUP_SET_DENSE(new_group);
 
     ompi_group_increment_proc_count (new_group);
-
+    Tau_stop_class_allocation(new_group->super.obj_class->cls_name, 0);
     return new_group;
 }
 
@@ -126,11 +131,14 @@ ompi_group_t *ompi_group_allocate_sporadic(int group_size)
         new_group = NULL;
         goto error_exit;
     }
+    Tau_start_class_allocation(new_group->super.obj_class->cls_name, 0, 0);
     /* allocate array of (grp_sporadic_list )'s */
     if (0 < group_size) {
         new_group->sparse_data.grp_sporadic.grp_sporadic_list =
             (struct ompi_group_sporadic_list_t *)malloc
             (sizeof(struct ompi_group_sporadic_list_t ) * group_size);
+        Tau_start_class_allocation("ompi_group_sporadic_list_t *", sizeof(struct ompi_group_sporadic_list_t ) * group_size, 0);
+        Tau_stop_class_allocation("ompi_group_sporadic_list_t *", 1);
 
         /* non-empty group */
         if ( NULL == new_group->sparse_data.grp_sporadic.grp_sporadic_list) {
@@ -151,6 +159,7 @@ ompi_group_t *ompi_group_allocate_sporadic(int group_size)
     OMPI_GROUP_SET_SPORADIC(new_group);
 
  error_exit:
+    Tau_stop_class_allocation(new_group->super.obj_class->cls_name, 0);
     return new_group;
 }
 
@@ -168,6 +177,7 @@ ompi_group_t *ompi_group_allocate_strided(void)
         new_group = NULL;
         goto error_exit;
     }
+    Tau_start_class_allocation(new_group->super.obj_class->cls_name, 0, 0);
     /* initialize our rank to MPI_UNDEFINED */
     new_group->grp_my_rank    = MPI_UNDEFINED;
     new_group->grp_proc_pointers     = NULL;
@@ -177,6 +187,7 @@ ompi_group_t *ompi_group_allocate_strided(void)
     new_group->sparse_data.grp_strided.grp_strided_last_element   = -1;
  error_exit:
     /* return */
+    Tau_stop_class_allocation(new_group->super.obj_class->cls_name, 0);
     return new_group;
 }
 ompi_group_t *ompi_group_allocate_bmap(int orig_group_size , int group_size)
@@ -196,8 +207,11 @@ ompi_group_t *ompi_group_allocate_bmap(int orig_group_size , int group_size)
         goto error_exit;
     }
     /* allocate the unsigned char list */
+    Tau_start_class_allocation(new_group->super.obj_class->cls_name, 0, 0);
     new_group->sparse_data.grp_bitmap.grp_bitmap_array = (unsigned char *)malloc
         (sizeof(unsigned char) * ompi_group_div_ceil(orig_group_size,BSIZE));
+    Tau_start_class_allocation("unsigned char *", sizeof(struct ompi_group_sporadic_list_t ) * group_size, 0);
+    Tau_stop_class_allocation("unsigned char *", 1);
 
     new_group->sparse_data.grp_bitmap.grp_bitmap_array_len =
         ompi_group_div_ceil(orig_group_size,BSIZE);
@@ -211,6 +225,7 @@ ompi_group_t *ompi_group_allocate_bmap(int orig_group_size , int group_size)
 
  error_exit:
     /* return */
+    Tau_stop_class_allocation(new_group->super.obj_class->cls_name, 0);
     return new_group;
 }
 
diff --git a/ompi/mca/pml/ob1/pml_ob1_comm.c b/ompi/mca/pml/ob1/pml_ob1_comm.c
index 510704849da..824cbfe5138 100644
--- a/ompi/mca/pml/ob1/pml_ob1_comm.c
+++ b/ompi/mca/pml/ob1/pml_ob1_comm.c
@@ -91,7 +91,11 @@ OBJ_CLASS_INSTANCE(
 int mca_pml_ob1_comm_init_size (mca_pml_ob1_comm_t* comm, size_t size)
 {
     /* send message sequence-number support - sender side */
+    Tau_start_class_allocation(comm->super.obj_class->cls_name, 0, 0);
     comm->procs = (mca_pml_ob1_comm_proc_t **) calloc(size, sizeof (mca_pml_ob1_comm_proc_t *));
+    Tau_start_class_allocation("mca_pml_ob1_comm_proc_t", size * sizeof(mca_pml_ob1_comm_proc_t *), 0);
+    Tau_stop_class_allocation("mca_pml_ob1_comm_proc_t", 1);
+    Tau_stop_class_allocation(comm->super.obj_class->cls_name, 0);
     if(NULL == comm->procs) {
         return OMPI_ERR_OUT_OF_RESOURCE;
     }
diff --git a/ompi/mca/pml/yalla/pml_yalla.c b/ompi/mca/pml/yalla/pml_yalla.c
index 11026b39c20..617406c6464 100644
--- a/ompi/mca/pml/yalla/pml_yalla.c
+++ b/ompi/mca/pml/yalla/pml_yalla.c
@@ -256,7 +256,11 @@ int mca_pml_yalla_add_procs(struct ompi_proc_t **procs, size_t nprocs)
             return OMPI_ERROR;
         }
 
+        Tau_start_class_allocation(procs[i]->super.super.super.obj_class->cls_name, 0, 0);
+        Tau_start_class_allocation("mxm_conn_h", sizeof(mxm_conn_h), 0);
         procs[i]->proc_endpoints[OMPI_PROC_ENDPOINT_TAG_PML] = conn;
+        Tau_stop_class_allocation("mxm_conn_h", 1);
+        Tau_stop_class_allocation(procs[i]->super.super.super.obj_class->cls_name, 0);
     }
 
     return OMPI_SUCCESS;
diff --git a/ompi/win/win.c b/ompi/win/win.c
index 6ce756e6e87..1939b7043d0 100644
--- a/ompi/win/win.c
+++ b/ompi/win/win.c
@@ -144,6 +144,7 @@ static int alloc_window(struct ompi_communicator_t *comm, ompi_info_t *info, int
 
     /* create the object */
     win = OBJ_NEW(ompi_win_t);
+    Tau_start_class_allocation(win->w_base.obj_class->cls_name, 0, 0);
     if (NULL == win) {
         return OMPI_ERR_OUT_OF_RESOURCE;
     }
@@ -177,6 +178,7 @@ static int alloc_window(struct ompi_communicator_t *comm, ompi_info_t *info, int
     win->w_group = group;
 
     *win_out = win;
+    Tau_stop_class_allocation(win->w_base.obj_class->cls_name, 0);
 
     return OMPI_SUCCESS;
 }
diff --git a/opal/class/opal_object.c b/opal/class/opal_object.c
index d36ab208b6c..87b01391f7a 100644
--- a/opal/class/opal_object.c
+++ b/opal/class/opal_object.c
@@ -33,6 +33,30 @@
 #include "opal/class/opal_object.h"
 #include "opal/constants.h"
 
+void __attribute__((weak)) Tau_track_class_allocation(const char * name, size_t size) {
+
+}
+
+void __attribute__((weak)) Tau_track_class_deallocation(const char * name, size_t size) {
+
+}
+
+void __attribute__((weak)) Tau_start_class_allocation(const char * name, size_t size, int include_in_parent) {
+
+}
+
+void __attribute__((weak)) Tau_stop_class_allocation(const char * name, int record) { 
+
+}
+
+void __attribute__((weak)) Tau_start_class_deallocation(const char * name, size_t size, int include_in_parent) {
+
+}
+
+void __attribute__((weak)) Tau_stop_class_deallocation(const char * name, int record) { 
+
+}
+
 /*
  * Instantiation of class descriptor for the base class.  This is
  * special, since be mark it as already initialized, with no parent
diff --git a/opal/class/opal_object.h b/opal/class/opal_object.h
index 8539f2bf872..e8e6c2a047c 100644
--- a/opal/class/opal_object.h
+++ b/opal/class/opal_object.h
@@ -132,6 +132,13 @@ BEGIN_C_DECLS
 #define OPAL_OBJ_MAGIC_ID ((0xdeafbeedULL << 32) + 0xdeafbeedULL)
 #endif
 
+void __attribute__((weak)) Tau_track_class_allocation(const char * name, size_t size);
+void __attribute__((weak)) Tau_track_class_deallocation(const char * name, size_t size);
+void __attribute__((weak)) Tau_start_class_allocation(const char * name, size_t size, int include_in_parent);
+void __attribute__((weak)) Tau_stop_class_allocation(const char * name, int record);
+void __attribute__((weak)) Tau_start_class_deallocation(const char * name, size_t size, int include_in_parent);
+void __attribute__((weak)) Tau_stop_class_deallocation(const char * name, int record);
+
 /* typedefs ***********************************************************/
 
 typedef struct opal_object_t opal_object_t;
@@ -457,6 +464,7 @@ static inline void opal_obj_run_destructors(opal_object_t * object)
 
     assert(NULL != object->obj_class);
 
+    Tau_track_class_deallocation(object->obj_class->cls_name, object->obj_class->cls_sizeof);
     cls_destruct = object->obj_class->cls_destruct_array;
     while( NULL != *cls_destruct ) {
         (*cls_destruct)(object);
@@ -480,6 +488,8 @@ static inline opal_object_t *opal_obj_new(opal_class_t * cls)
     opal_object_t *object;
     assert(cls->cls_sizeof >= sizeof(opal_object_t));
 
+    Tau_start_class_allocation(cls->cls_name, cls->cls_sizeof, 0);
+
 #if OPAL_WANT_MEMCHECKER
     object = (opal_object_t *) calloc(1, cls->cls_sizeof);
 #else
@@ -493,6 +503,7 @@ static inline opal_object_t *opal_obj_new(opal_class_t * cls)
         object->obj_reference_count = 1;
         opal_obj_run_constructors(object);
     }
+    Tau_stop_class_allocation(cls->cls_name, 1);
     return object;
 }
 
diff --git a/oshmem/proc/proc.c b/oshmem/proc/proc.c
index 665dd10caed..f6bf7a3b259 100644
--- a/oshmem/proc/proc.c
+++ b/oshmem/proc/proc.c
@@ -136,6 +136,7 @@ oshmem_group_t* oshmem_proc_group_create(int pe_start,
     assert(oshmem_proc_local());
 
     group = OBJ_NEW(oshmem_group_t);
+    Tau_start_class_allocation(group->base.obj_class->cls_name, 0, 0);
 
     if (group) {
         cur_pe = 0;
@@ -145,7 +146,10 @@ oshmem_group_t* oshmem_proc_group_create(int pe_start,
 
         /* allocate an array */
         proc_array = (ompi_proc_t**) malloc(pe_size * sizeof(ompi_proc_t*));
+        Tau_start_class_allocation("ompi_proc_t **", pe_size * sizeof(ompi_proc_t*), 0);
+        Tau_stop_class_allocation("ompi_proc_t **", 1);
         if (NULL == proc_array) {
+            Tau_stop_class_allocation(group->base.obj_class->cls_name, 0);
             OBJ_RELEASE(group);
             OPAL_THREAD_UNLOCK(&oshmem_proc_lock);
             return NULL ;
@@ -197,6 +201,7 @@ oshmem_group_t* oshmem_proc_group_create(int pe_start,
         if (OSHMEM_SUCCESS != mca_scoll_base_select(group)) {
             opal_output(0,
                         "Error: No collective modules are available: group is not created, returning NULL");
+            Tau_stop_class_allocation(group->base.obj_class->cls_name, 0);
             oshmem_proc_group_destroy(group);
             OPAL_THREAD_UNLOCK(&oshmem_proc_lock);
             return NULL;
@@ -204,6 +209,7 @@ oshmem_group_t* oshmem_proc_group_create(int pe_start,
         OPAL_THREAD_UNLOCK(&oshmem_proc_lock);
     }
 
+    Tau_stop_class_allocation(group->base.obj_class->cls_name, 0);
     return group;
 }