"... In PAGE 3: ... Resources Collectively, resources at these sites represent a significant pool of heterogeneous computing power. The Level 2 Grid consists of a large number of testbed systems plus the larger ( gt;32 processor) servers shown in Table1 . Other resources are diverse in nature, including a Sony Playstation at University of Manchester and an SGI Onyx visualisation engine at University... ..."

"... In PAGE 6: ... The rst analysis we did was to determine how evenly NTP usage was spread across the New Zealand second and third level domains. Table3 summarises for each level 2 domain the number of NTP servers in that level 2 domain, and the number of di erent level 3 domains these servers belonged to. Information is given for both datasets.... In PAGE 7: ...nd 11 were in comnet.co.nz. In the gures presented in Table3 , all 16 121.203 hosts were counted as belonging to govt.... ..."

"... In PAGE 10: ...ytes by less than 0.3%). However, it is larger than any other job size in the empirical workload, and depending on the probe arrival time, could extend the simulated completion time slightly. Table2 summarizes the factors and levels used in our experiments. For space reasons, only a subset of the experiments is reported in this paper.... ..."

"... In PAGE 3: ... The system load is xed at 95% for all of the experiments. Table2 sum- marizes the factors and levels used in the trace-driven simulation experiments. We consider two main performance metrics: Slowdown: The slowdown metric is de ned as the response time of a job divided by the ideal response time if it were the sole job in the system.... ..."

"... In PAGE 54: ...Table16 (continued) Initialization Functions /* Function pthread_ss_init */ /****************************/ int pthread_ss_init (pthread_ss_t *ss, struc timespec *period, struc timespec *budget, int normal_priority, int background_priority) static pthread_once_t ss_manager=pthread_once_init; pthread_mutex_t *ss_thread_mutex; pthread_cond_t *ss_thread_cond; { /* Create the mutex and cond objects to manage the thread*/ pthread_mutex_init(ss_thread_mutex,pthread_mutexattr_default); pthread_cond_init(ss_thread_cond, pthread_condattr_default); /* Start the sporadic_server_manager (once) */ pthread_once( amp;ss_manager, amp;sporadic_server_init); initialize private attributes of sporadic server control block (period, budget, priorities, mutex, condition, owner id .... In PAGE 55: ...Table16 (continued) Sporadic Server Manager Thread void sporadic_server_manager () { while (1) { get next replenishment from queue; /* Wait until replenishment time or ss request */ pthread_cond_timedwait(ss_list_cond,ss_list_mutex, replenishment_time); /* Replenish due tickets */ while (there are replenishments due) { replenish tickets; /* Replenishment Action: Indivisibly test amp; increase priority */ pthread_mutex_lock (ss_thread_mutex); if (pthread_getprio(ss_thread)=background p.) { if (available_tickets gt;=request_size) { decrement available_tickets by request_size; set replenishment time into queue; pthread_setprio(normal_priority,ss_thread); } /* end if */ } /* end if */ pthread_mutex_unlock (ss_thread_mutex); } /* end while */ /* Signal continuation to waiting sporadic servers */ while (there are ss_threads waiting in ss_list) { if (request_type=process) { /* Determine priority and replenishment actions */ if (available_tickets gt;=request_size) { decrement available_tickets by request_size; set replenishment time into queue; prio:=normal_priority; } else { prio:=background_priority; } /* end if */ } else { /*request_type=detach*/ detach ss from replenishment_queue; } /*end if*/ /* Lock ss thread mutex, set priority and signal */ pthread_mutex_lock (ss_thread_mutex); pthread_setprio(prio,ss_thread); clear waiting_flag; pthread_mutex_unlock(ss_thread_mutex); pthread_cond_signal(ss_thread_cond); } /* end while */ } /* end while */ } /* end sporadic_server_manager */ CMU/SEI-91-TR-26... In PAGE 56: ...Table16 (continued) Sporadic Server Detach Function /* Function pthread_ss_detach */ /********************************/ int pthread_ss_detach (pthread_ss_t *ss) { /* Request detach to Sporadic_server_manager */ pthread_mutex_lock(ss_list_mutex); set request type (=detach) and waiting_flag in ss_list; pthread_mutex_unlock(ss_list_mutex); pthread_cond_signal(ss_list_cond); /* Wait for manager continuation signal */ pthread_mutex_lock (ss_thread.mutex); if (waiting_flag set) { pthread_cond_wait(ss_thread.... In PAGE 57: ...Table16 (continued) Sporadic Server Request Function /* Function pthread_ss_request */ /****************************/ int pthread_ss_request (pthread_ss_t *ss_thread struc timespec *request_size) { /* Request processing by Sporadic_server_manager */ pthread_mutex_lock(ss_list_mutex); set request_size and waiting_flag in ss_list; pthread_mutex_unlock(ss_list_mutex); pthread_cond_signal(ss_list_cond); /* Wait for manager continuation signal */ pthread_mutex_lock (ss_thread.mutex); if (waiting_flag set) { pthread_cond_wait(ss_thread.... In PAGE 57: ...mutex); } /* end if */ pthread_mutex_unlock (ss_thread.mutex); } /* end pthread_ss_request */ Table16 (continued) Sporadic Server Arm Function /* Function pthread_ss_arm */ /***************************/ int pthread_ss_arm (pthread_ss_t *ss_thread) { /* Set priority to high value */ pthread_mutex_lock (ss_thread.mutex); pthread_setprio(ss_thread,PRIO_MAX); pthread_mutex_unlock (ss_thread.... ..."

"... In PAGE 5: ... The main activity in the index server consists of decoding compressed information in the inverted index and finding matches against a set of documents that could satisfy a query. Table1 shows some basic instruction-level measurements of the index server program running on a 1-GHz dual- processor Pentium III system. The application has a moderately high CPI, considering that the Pentium III is capable of issuing three instructions per cycle.... In PAGE 6: ... The avail- able thread-level parallelism should allow near-linear speedup with the number of cores, and a shared L2 cache of reasonable size would speed up interprocessor communication. Memory system Table1 also outlines the main memory sys- tem performance parameters. We observe good performance for the instruction cache and instruction translation look-aside buffer, a result of the relatively small inner-loop code size.... ..."

"... In PAGE 5: ... The main activity in the index server consists of decoding compressed information in the inverted index and finding matches against a set of documents that could satisfy a query. Table1 shows some basic instruction-level measurements of the index server program running on a 1-GHz dual- processor Pentium III system. The application has a moderately high CPI, considering that the Pentium III is capable of issuing three instructions per cycle.... In PAGE 6: ... The avail- able thread-level parallelism should allow near-linear speedup with the number of cores, and a shared L2 cache of reasonable size would speed up interprocessor communication. Memory system Table1 also outlines the main memory sys- tem performance parameters. We observe good performance for the instruction cache and instruction translation look-aside buffer, a result of the relatively small inner-loop code size.... ..."

"... In PAGE 7: ... We use a request file size as 16K bytes, and do similar experiments as the previous one. Table4 shows the function level (protocol level) measurement for processing one data packet sent from the server (its size is determined by the MTU, which is 1500 bytes for Ethernet). The functions we list here are encountered on a path of the re- sponse packet.... ..."