/* * Generate/analyze pareto/zipf distributions to better understand * what an access pattern would look like. * * For instance, the following would generate a zipf distribution * with theta 1.2, using 262144 (1 GiB / 4096) values and split the * reporting into 20 buckets: * * ./t/fio-genzipf -t zipf -i 1.2 -g 1 -b 4096 -o 20 * * Only the distribution type (zipf or pareto) and spread input need * to be given, if not given defaults are used. * */ #include #include #include #include #include "../lib/zipf.h" #include "../lib/gauss.h" #include "../flist.h" #include "../hash.h" #define DEF_NR_OUTPUT 20 struct node { struct flist_head list; unsigned long long val; unsigned long hits; }; static struct flist_head *hash; static unsigned long hash_bits = 24; static unsigned long hash_size = 1 << 24; enum { TYPE_NONE = 0, TYPE_ZIPF, TYPE_PARETO, TYPE_NORMAL, }; static const char *dist_types[] = { "None", "Zipf", "Pareto", "Normal" }; enum { OUTPUT_NORMAL, OUTPUT_CSV, }; static int dist_type = TYPE_ZIPF; static unsigned long gib_size = 500; static unsigned long block_size = 4096; static unsigned long output_nranges = DEF_NR_OUTPUT; static double percentage; static double dist_val; static int output_type = OUTPUT_NORMAL; #define DEF_ZIPF_VAL 1.2 #define DEF_PARETO_VAL 0.3 static unsigned int hashv(unsigned long long val) { return jhash(&val, sizeof(val), 0) & (hash_size - 1); } static struct node *hash_lookup(unsigned long long val) { struct flist_head *l = &hash[hashv(val)]; struct flist_head *entry; struct node *n; flist_for_each(entry, l) { n = flist_entry(entry, struct node, list); if (n->val == val) return n; } return NULL; } static void hash_insert(struct node *n, unsigned long long val) { struct flist_head *l = &hash[hashv(val)]; n->val = val; n->hits = 1; flist_add_tail(&n->list, l); } static void usage(void) { printf("genzipf: test zipf/pareto values for fio input\n"); printf("\t-h\tThis help screen\n"); printf("\t-p\tGenerate size of data set that are hit by this percentage\n"); printf("\t-t\tDistribution type (zipf, pareto, or normal)\n"); printf("\t-i\tDistribution algorithm input (zipf theta, pareto power,\n" "\t\tor normal %% deviation)\n"); printf("\t-b\tBlock size of a given range (in bytes)\n"); printf("\t-g\tSize of data set (in gigabytes)\n"); printf("\t-o\tNumber of output rows\n"); printf("\t-c\tOutput ranges in CSV format\n"); } static int parse_options(int argc, char *argv[]) { const char *optstring = "t:g:i:o:b:p:ch"; int c, dist_val_set = 0; while ((c = getopt(argc, argv, optstring)) != -1) { switch (c) { case 'h': usage(); return 1; case 'p': percentage = atof(optarg); break; case 'b': block_size = strtoul(optarg, NULL, 10); break; case 't': if (!strncmp(optarg, "zipf", 4)) dist_type = TYPE_ZIPF; else if (!strncmp(optarg, "pareto", 6)) dist_type = TYPE_PARETO; else if (!strncmp(optarg, "normal", 6)) dist_type = TYPE_NORMAL; else { printf("wrong dist type: %s\n", optarg); return 1; } break; case 'g': gib_size = strtoul(optarg, NULL, 10); break; case 'i': dist_val = atof(optarg); dist_val_set = 1; break; case 'o': output_nranges = strtoul(optarg, NULL, 10); break; case 'c': output_type = OUTPUT_CSV; break; default: printf("bad option %c\n", c); return 1; } } if (dist_type == TYPE_PARETO) { if ((dist_val >= 1.00 || dist_val < 0.00)) { printf("pareto input must be > 0.00 and < 1.00\n"); return 1; } if (!dist_val_set) dist_val = DEF_PARETO_VAL; } else if (dist_type == TYPE_ZIPF) { if (dist_val == 1.0) { printf("zipf input must be different than 1.0\n"); return 1; } if (!dist_val_set) dist_val = DEF_ZIPF_VAL; } return 0; } struct output_sum { double output; unsigned int nranges; }; static int node_cmp(const void *p1, const void *p2) { const struct node *n1 = p1; const struct node *n2 = p2; return n2->hits - n1->hits; } static void output_csv(struct node *nodes, unsigned long nnodes) { unsigned long i; printf("rank, count\n"); for (i = 0; i < nnodes; i++) printf("%lu, %lu\n", i, nodes[i].hits); } static void output_normal(struct node *nodes, unsigned long nnodes, unsigned long nranges) { unsigned long i, j, cur_vals, interval_step, next_interval, total_vals; unsigned long blocks = percentage * nnodes / 100; double hit_percent_sum = 0; unsigned long long hit_sum = 0; double perc, perc_i; struct output_sum *output_sums; interval_step = (nnodes - 1) / output_nranges + 1; next_interval = interval_step; output_sums = malloc(output_nranges * sizeof(struct output_sum)); for (i = 0; i < output_nranges; i++) { output_sums[i].output = 0.0; output_sums[i].nranges = 0; } j = total_vals = cur_vals = 0; for (i = 0; i < nnodes; i++) { struct output_sum *os = &output_sums[j]; struct node *node = &nodes[i]; cur_vals += node->hits; total_vals += node->hits; os->nranges += node->hits; if (i == (next_interval) -1 || i == nnodes - 1) { os->output = (double) cur_vals / (double) nranges; os->output *= 100.0; cur_vals = 0; next_interval += interval_step; j++; } if (percentage) { if (total_vals >= blocks) { double cs = (double) i * block_size / (1024.0 * 1024.0); char p = 'M'; if (cs > 1024.0) { cs /= 1024.0; p = 'G'; } if (cs > 1024.0) { cs /= 1024.0; p = 'T'; } printf("%.2f%% of hits satisfied in %.3f%cB of cache\n", percentage, cs, p); percentage = 0.0; } } } perc_i = 100.0 / (double)output_nranges; perc = 0.0; printf("\n Rows Hits %% Sum %% # Hits Size\n"); printf("-----------------------------------------------------------------------\n"); for (i = 0; i < output_nranges; i++) { struct output_sum *os = &output_sums[i]; double gb = (double)os->nranges * block_size / 1024.0; char p = 'K'; if (gb > 1024.0) { p = 'M'; gb /= 1024.0; } if (gb > 1024.0) { p = 'G'; gb /= 1024.0; } perc += perc_i; hit_percent_sum += os->output; hit_sum += os->nranges; printf("%s %6.2f%%\t%6.2f%%\t\t%6.2f%%\t\t%8u\t%6.2f%c\n", i ? "|->" : "Top", perc, os->output, hit_percent_sum, os->nranges, gb, p); } printf("-----------------------------------------------------------------------\n"); printf("Total\t\t\t\t\t\t%8llu\n", hit_sum); free(output_sums); } int main(int argc, char *argv[]) { unsigned long offset; unsigned long long nranges; unsigned long nnodes; struct node *nodes; struct zipf_state zs; struct gauss_state gs; int i, j; if (parse_options(argc, argv)) return 1; if (output_type != OUTPUT_CSV) printf("Generating %s distribution with %f input and %lu GiB size and %lu block_size.\n", dist_types[dist_type], dist_val, gib_size, block_size); nranges = gib_size * 1024 * 1024 * 1024ULL; nranges /= block_size; if (dist_type == TYPE_ZIPF) zipf_init(&zs, nranges, dist_val, 1); else if (dist_type == TYPE_PARETO) pareto_init(&zs, nranges, dist_val, 1); else gauss_init(&gs, nranges, dist_val, 1); hash_bits = 0; hash_size = nranges; while ((hash_size >>= 1) != 0) hash_bits++; hash_size = 1 << hash_bits; hash = calloc(hash_size, sizeof(struct flist_head)); for (i = 0; i < hash_size; i++) INIT_FLIST_HEAD(&hash[i]); nodes = malloc(nranges * sizeof(struct node)); for (i = j = 0; i < nranges; i++) { struct node *n; if (dist_type == TYPE_ZIPF) offset = zipf_next(&zs); else if (dist_type == TYPE_PARETO) offset = pareto_next(&zs); else offset = gauss_next(&gs); n = hash_lookup(offset); if (n) n->hits++; else { hash_insert(&nodes[j], offset); j++; } } qsort(nodes, j, sizeof(struct node), node_cmp); nnodes = j; if (output_type == OUTPUT_CSV) output_csv(nodes, nnodes); else output_normal(nodes, nnodes, nranges); free(hash); free(nodes); return 0; }