文档章节

Fio随机读IOPS测试值可能偏大的原因分析

LastRitter
 LastRitter
发布于 11/21 21:29
字数 5930
阅读 146
收藏 0
fio

问题描述:

  • 在使用fio进行虚拟机磁盘(CephRBD,格式化为ext4文件系统)的IOPS测试时,发现randread比预估值高许多;
  • 在使用相同参数进行randwrite测试之后,再进行randread时会出现此现象;
  • 而使用dd构建测试文件后,再进行randread就不会出现这种情况,IOPS数值正常。

初步推测,可能fio的随机是伪随机,导致前后的randwriterandread使用了相同的伪随机序列。文件系统在进行物理块分配时从前往后分配,在逻辑上随机的块实际上是顺序写入物理磁盘,最终的随机读实际上是顺序读,导致IO被磁盘调度器合并,实际IO次数变少,所以测试的IOPS偏大,为此进行详细分析测试。

打印Debug日志

  • 打开fiodebug模式,执行测试,输出日志:
$ fio -direct=1 -iodepth=128 -rw=randwrite -ioengine=libaio -bs=4k -size=1G -numjobs=1 -runtime=10 -group_reporting -filename=iotest -name=Rand_Write_Testing --debug=random > rand_write_offset.log

$ fio -direct=1 -iodepth=128 -rw=randread -ioengine=libaio -bs=4k -size=1G -numjobs=1 -runtime=10 -group_reporting -filename=iotest -name=Rand_Read_Testingg --debug=random > rand_read_offset.log
  • 查看日志:
$ head -n30 rand_write_offset.log 
fio: set debug option random
Rand_Write_Testing: (g=0): rw=randwrite, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=128
fio-3.1
Starting 1 process
random   4057532 off rand 259043585
random   4057532 off rand 3179521932
random   4057532 off rand 3621444214
random   4057532 off rand 2018697059
random   4057532 off rand 1726199243
random   4057532 off rand 3608323581
random   4057532 off rand 1634212905
random   4057532 off rand 1518359867
random   4057532 off rand 3921331707
random   4057532 off rand 287004724
random   4057532 off rand 3673173177
random   4057532 off rand 2796675757
random   4057532 off rand 3988051731
random   4057532 off rand 1060357494
random   4057532 off rand 1685717462
random   4057532 off rand 2400737531
random   4057532 off rand 1891936796
random   4057532 off rand 3455447349
random   4057532 off rand 1553547805
random   4057532 off rand 2660809810
random   4057532 off rand 17263379
random   4057532 off rand 1823528783
random   4057532 off rand 1355450167
random   4057532 off rand 2956359995
random   4057532 off rand 3392712188
random   4057532 off rand 4240594610

$ $ head -n30 rand_read_offset.log 
fio: set debug option random
Rand_Read_Testingg: (g=0): rw=randread, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=libaio, iodepth=128
fio-3.1
Starting 1 process
random   4057831 off rand 259043585
random   4057831 off rand 3179521932
random   4057831 off rand 3621444214
random   4057831 off rand 2018697059
random   4057831 off rand 1726199243
random   4057831 off rand 3608323581
random   4057831 off rand 1634212905
random   4057831 off rand 1518359867
random   4057831 off rand 3921331707
random   4057831 off rand 287004724
random   4057831 off rand 3673173177
random   4057831 off rand 2796675757
random   4057831 off rand 3988051731
random   4057831 off rand 1060357494
random   4057831 off rand 1685717462
random   4057831 off rand 2400737531
random   4057831 off rand 1891936796
random   4057831 off rand 3455447349
random   4057831 off rand 1553547805
random   4057831 off rand 2660809810
random   4057831 off rand 17263379
random   4057831 off rand 1823528783
random   4057831 off rand 1355450167
random   4057831 off rand 2956359995
random   4057831 off rand 3392712188
random   4057831 off rand 4240594610
  • 日志对比,发现右侧的随机偏移都是一样的:

获取Fio源码

  • 分析的源码来源和版本如下:
$ git clone https://github.com/axboe/fio.git
$ cd fio
$ git branch -av
* master                       ee636f3 libaio: switch to newer libaio polled IO API
  remotes/origin/HEAD          -> origin/master
  remotes/origin/latency-probe fcd4e74 target: fixes
  remotes/origin/master        ee636f3 libaio: switch to newer libaio polled IO API

分析debug选项

  • 查找debug选项的定义和引用位置:
$ grep -rHn \"debug\"
init.c:176:		.name		= (char *) "debug",
  • 查找random参数的定义和引用位置,可以看到random参数使用FD_RANDOM宏或者枚举值进行定义:
$ grep -rHn \"random\" -A5 init.c 
init.c:2260:	{ .name = "random",
init.c-2261-	  .help = "Random generation logging",
init.c-2262-	  .shift = FD_RANDOM,
init.c-2263-	},
init.c-2264-	{ .name = "parse",
init.c-2265-	  .help = "Parser logging",
  • 查找FD_RANDOM这个宏开关的定义和引用位置,可以发现定义在debug.h中,在io_u.c中被引用,是用来开关debug打印的,其中第98行的和之前的Debug日志格式相同:
$ grep -rHn FD_RANDOM
debug.h:13:	FD_RANDOM,
init.c:2262:	  .shift = FD_RANDOM,
io_u.c:98:		dprint(FD_RANDOM, "off rand %llu\n", (unsigned long long) r);
io_u.c:124:	dprint(FD_RANDOM, "get_next_rand_offset: offset %llu busy\n",
  • 查看FD_RANDOM引用处附近的额源码,第96行处就是生成随机数的地方,第98行对生成的随机数进行打印:
$ grep -rHn FD_RANDOM io_u.c -C12
io_u.c-86-
io_u.c-87-static int __get_next_rand_offset(struct thread_data *td, struct fio_file *f,
io_u.c-88-				  enum fio_ddir ddir, uint64_t *b,
io_u.c-89-				  uint64_t lastb)
io_u.c-90-{
io_u.c-91-	uint64_t r;
io_u.c-92-
io_u.c-93-	if (td->o.random_generator == FIO_RAND_GEN_TAUSWORTHE ||
io_u.c-94-	    td->o.random_generator == FIO_RAND_GEN_TAUSWORTHE64) {
io_u.c-95-
io_u.c-96-		r = __rand(&td->random_state);
io_u.c-97-
io_u.c:98:		dprint(FD_RANDOM, "off rand %llu\n", (unsigned long long) r);
io_u.c-99-
io_u.c-100-		*b = lastb * (r / (rand_max(&td->random_state) + 1.0));
io_u.c-101-	} else {
io_u.c-102-		uint64_t off = 0;
io_u.c-103-
io_u.c-104-		assert(fio_file_lfsr(f));
io_u.c-105-
io_u.c-106-		if (lfsr_next(&f->lfsr, &off))
io_u.c-107-			return 1;
io_u.c-108-
io_u.c-109-		*b = off;
io_u.c-110-	}
--
io_u.c-112-	/*
io_u.c-113-	 * if we are not maintaining a random map, we are done.
io_u.c-114-	 */
io_u.c-115-	if (!file_randommap(td, f))
io_u.c-116-		goto ret;
io_u.c-117-
io_u.c-118-	/*
io_u.c-119-	 * calculate map offset and check if it's free
io_u.c-120-	 */
io_u.c-121-	if (random_map_free(f, *b))
io_u.c-122-		goto ret;
io_u.c-123-
io_u.c:124:	dprint(FD_RANDOM, "get_next_rand_offset: offset %llu busy\n",
io_u.c-125-						(unsigned long long) *b);
io_u.c-126-
io_u.c-127-	*b = axmap_next_free(f->io_axmap, *b);
io_u.c-128-	if (*b == (uint64_t) -1ULL)
io_u.c-129-		return 1;
io_u.c-130-ret:
io_u.c-131-	return 0;
io_u.c-132-}
io_u.c-133-
io_u.c-134-static int __get_next_rand_offset_zipf(struct thread_data *td,
io_u.c-135-				       struct fio_file *f, enum fio_ddir ddir,
io_u.c-136-				       uint64_t *b)

分析dprint函数

  • 查找dprint函数或者宏的定义及引用处,定义在debug.h中:
$ grep -rHn " dprint"
debug.h:62:#define dprint(type, str, args...)			\
debug.h:71:static inline void dprint(int type, const char *str, ...)
gettime.c:320:                dprint(FD_TIME, "tmp=%llu, sft=%u\n", tmp, sft);
io_u.h:153:static inline void dprint_io_u(struct io_u *io_u, const char *p)
io_u.h:170:#define dprint_io_u(io_u, p)
t/time-test.c:88:#define dprintf(...) if (DEBUG) { printf(__VA_ARGS__); }
  • 查看debug.h中的dprint定义内容:
$ grep -rHn " dprint" -C7 debug.h
debug.h-55-};
debug.h-56-extern const struct debug_level debug_levels[];
debug.h-57-
debug.h-58-extern unsigned long fio_debug;
debug.h-59-
debug.h-60-void __dprint(int type, const char *str, ...) __attribute__((format (printf, 2, 3)));
debug.h-61-
debug.h:62:#define dprint(type, str, args...)			\
debug.h-63-	do {						\
debug.h-64-		if (((1 << type) & fio_debug) == 0)	\
debug.h-65-			break;				\
debug.h-66-		__dprint((type), (str), ##args);	\
debug.h-67-	} while (0)					\
debug.h-68-
debug.h-69-#else
debug.h-70-
debug.h:71:static inline void dprint(int type, const char *str, ...)
debug.h-72-{
debug.h-73-}
debug.h-74-#endif
debug.h-75-
debug.h-76-#endif
  • 查看__dprint函数,使用log_prevalist进行最终的字符串打印:
$ grep -rHn " __dprint" -C10
debug.c-1-#include <assert.h>
debug.c-2-#include <stdarg.h>
debug.c-3-
debug.c-4-#include "debug.h"
debug.c-5-#include "log.h"
debug.c-6-
debug.c-7-#ifdef FIO_INC_DEBUG
debug.c:8:void __dprint(int type, const char *str, ...)
debug.c-9-{
debug.c-10-	va_list args;
debug.c-11-
debug.c-12-	assert(type < FD_DEBUG_MAX);
debug.c-13-
debug.c-14-	va_start(args, str);
debug.c-15-	log_prevalist(type, str, args);
debug.c-16-	va_end(args);
debug.c-17-}
debug.c-18-#endif
--
debug.h-50-struct debug_level {
debug.h-51-	const char *name;
debug.h-52-	const char *help;
debug.h-53-	unsigned long shift;
debug.h-54-	unsigned int jobno;
debug.h-55-};
debug.h-56-extern const struct debug_level debug_levels[];
debug.h-57-
debug.h-58-extern unsigned long fio_debug;
debug.h-59-
debug.h:60:void __dprint(int type, const char *str, ...) __attribute__((format (printf, 2, 3)));
debug.h-61-
debug.h-62-#define dprint(type, str, args...)			\
debug.h-63-	do {						\
debug.h-64-		if (((1 << type) & fio_debug) == 0)	\
debug.h-65-			break;				\
debug.h-66-		__dprint((type), (str), ##args);	\
debug.h-67-	} while (0)					\
debug.h-68-
debug.h-69-#else
debug.h-70-
--
t/debug.c-1-#include <stdio.h>
t/debug.c-2-
t/debug.c-3-FILE *f_err;
t/debug.c-4-struct timespec *fio_ts = NULL;
t/debug.c-5-unsigned long fio_debug = 0;
t/debug.c-6-
t/debug.c:7:void __dprint(int type, const char *str, ...)
t/debug.c-8-{
t/debug.c-9-}
t/debug.c-10-
t/debug.c-11-void debug_init(void)
t/debug.c-12-{
t/debug.c-13-	f_err = stderr;
t/debug.c-14-}
  • 查看log.c中的log_prevalist函数定义,首先打印了当前进程PID,然后是类型字符串,最后才是格式化字符串:
/* add prefix for the specified type in front of the valist */
void log_prevalist(int type, const char *fmt, va_list args)
{
	char *buf1, *buf2;
	int len;
	pid_t pid;

	pid = gettid();
	if (fio_debug_jobp && *fio_debug_jobp != -1U
	    && pid != *fio_debug_jobp)
		return;

	len = vasprintf(&buf1, fmt, args);
	if (len < 0)
		return;
	len = asprintf(&buf2, "%-8s %-5u %s", debug_levels[type].name,
		       (int) pid, buf1);
	free(buf1);
	if (len < 0)
		return;
	len = log_info_buf(buf2, len);
	free(buf2);
}
  • 查找debug_levels字符串数组的定义处:
$ grep -rHn debug_levels
debug.h:56:extern const struct debug_level debug_levels[];
gfio.c:1187:		buttons[i] = gtk_check_button_new_with_label(debug_levels[i].name);
gfio.c:1188:		gtk_widget_set_tooltip_text(buttons[i], debug_levels[i].help);
init.c:2144:	const struct debug_level *dl = &debug_levels[0];
init.c:2235:const struct debug_level debug_levels[] = {
init.c:2327:		for (i = 0; debug_levels[i].name; i++) {
init.c:2328:			dl = &debug_levels[i];
init.c:2344:		for (i = 0; debug_levels[i].name; i++) {
init.c:2345:			dl = &debug_levels[i];
log.c:59:	len = asprintf(&buf2, "%-8s %-5u %s", debug_levels[type].name,
  • init.c可以看到最终的定义:
#ifdef FIO_INC_DEBUG
const struct debug_level debug_levels[] = {
	{ .name = "process",
	  .help = "Process creation/exit logging",
	  .shift = FD_PROCESS,
	},
	{ .name = "file",
	  .help = "File related action logging",
	  .shift = FD_FILE,
	},
	{ .name = "io",
	  .help = "IO and IO engine action logging (offsets, queue, completions, etc)",
	  .shift = FD_IO,
	},
	{ .name = "mem",
	  .help = "Memory allocation/freeing logging",
	  .shift = FD_MEM,
	},
	{ .name = "blktrace",
	  .help = "blktrace action logging",
	  .shift = FD_BLKTRACE,
	},
	{ .name = "verify",
	  .help = "IO verification action logging",
	  .shift = FD_VERIFY,
	},
	{ .name = "random",
	  .help = "Random generation logging",
	  .shift = FD_RANDOM,
	},
	{ .name = "parse",
	  .help = "Parser logging",
	  .shift = FD_PARSE,
	},
	{ .name = "diskutil",
	  .help = "Disk utility logging actions",
	  .shift = FD_DISKUTIL,
	},
	{ .name = "job",
	  .help = "Logging related to creating/destroying jobs",
	  .shift = FD_JOB,
	},
	{ .name = "mutex",
	  .help = "Mutex logging",
	  .shift = FD_MUTEX
	},
	{ .name	= "profile",
	  .help = "Logging related to profiles",
	  .shift = FD_PROFILE,
	},
	{ .name = "time",
	  .help = "Logging related to time keeping functions",
	  .shift = FD_TIME,
	},
	{ .name = "net",
	  .help = "Network logging",
	  .shift = FD_NET,
	},
	{ .name = "rate",
	  .help = "Rate logging",
	  .shift = FD_RATE,
	},
	{ .name = "compress",
	  .help = "Log compression logging",
	  .shift = FD_COMPRESS,
	},
	{ .name = "steadystate",
	  .help = "Steady state detection logging",
	  .shift = FD_STEADYSTATE,
	},
	{ .name = "helperthread",
	  .help = "Helper thread logging",
	  .shift = FD_HELPERTHREAD,
	},
	{ .name = "zbd",
	  .help = "Zoned Block Device logging",
	  .shift = FD_ZBD,
	},
	{ .name = NULL, },
};

static int set_debug(const char *string)
{
	const struct debug_level *dl;
	char *p = (char *) string;
	char *opt;
	int i;

	if (!string)
		return 0;

	if (!strcmp(string, "?") || !strcmp(string, "help")) {
		log_info("fio: dumping debug options:");
		for (i = 0; debug_levels[i].name; i++) {
			dl = &debug_levels[i];
			log_info("%s,", dl->name);
		}
		log_info("all\n");
		return 1;
	}

	while ((opt = strsep(&p, ",")) != NULL) {
		int found = 0;

		if (!strncmp(opt, "all", 3)) {
			log_info("fio: set all debug options\n");
			fio_debug = ~0UL;
			continue;
		}

		for (i = 0; debug_levels[i].name; i++) {
			dl = &debug_levels[i];
			found = !strncmp(opt, dl->name, strlen(dl->name));
			if (!found)
				continue;

			if (dl->shift == FD_JOB) {
				opt = strchr(opt, ':');
				if (!opt) {
					log_err("fio: missing job number\n");
					break;
				}
				opt++;
				fio_debug_jobno = atoi(opt);
				log_info("fio: set debug jobno %d\n",
							fio_debug_jobno);
			} else {
				log_info("fio: set debug option %s\n", opt);
				fio_debug |= (1UL << dl->shift);
			}
			break;
		}

		if (!found)
			log_err("fio: debug mask %s not found\n", opt);
	}
	return 0;
}
#else
static int set_debug(const char *string)
{
	log_err("fio: debug tracing not included in build\n");
	return 1;
}
#endif

分析randwrite选项

  • 查找randwrite参数的定义和引用处,使用TD_DDIR_RANDWRITE作为参数值:
$ grep -rHn \"randwrite\" -C5
io_ddir.h-62-}
io_ddir.h-63-
io_ddir.h-64-static inline const char *ddir_str(enum td_ddir ddir)
io_ddir.h-65-{
io_ddir.h-66-	static const char *__str[] = { NULL, "read", "write", "rw", "rand",
io_ddir.h:67:				"randread", "randwrite", "randrw",
io_ddir.h-68-				"trim", NULL, "trimwrite", NULL, "randtrim" };
io_ddir.h-69-
io_ddir.h-70-	return __str[ddir];
io_ddir.h-71-}
io_ddir.h-72-
--
options.c-1690-			  },
options.c-1691-			  { .ival = "randread",
options.c-1692-			    .oval = TD_DDIR_RANDREAD,
options.c-1693-			    .help = "Random read",
options.c-1694-			  },
options.c:1695:			  { .ival = "randwrite",
options.c-1696-			    .oval = TD_DDIR_RANDWRITE,
options.c-1697-			    .help = "Random write",
options.c-1698-			  },
options.c-1699-			  { .ival = "randtrim",
options.c-1700-			    .oval = TD_DDIR_RANDTRIM,
--
profiles/act.c-182-
profiles/act.c-183-	if (act_add_opt("name=act-%s-%s", reads ? "read" : "write", dev))
profiles/act.c-184-		return 1;
profiles/act.c-185-	if (act_add_opt("filename=%s", dev))
profiles/act.c-186-		return 1;
profiles/act.c:187:	if (act_add_opt("rw=%s", reads ? "randread" : "randwrite"))
profiles/act.c-188-		return 1;
profiles/act.c-189-	if (reads) {
profiles/act.c-190-		int rload = ao->load * R_LOAD / ao->threads_per_queue;
profiles/act.c-191-
profiles/act.c-192-		if (act_add_opt("numjobs=%u", ao->threads_per_queue))
--
t/sgunmap-test.py-116-
t/sgunmap-test.py-117-
t/sgunmap-test.py-118-def runalltests(args, qd, batch):
t/sgunmap-test.py-119-    block = False
t/sgunmap-test.py-120-    for dev in [args.chardev, args.blockdev]:
t/sgunmap-test.py:121:        for rw in ["randread", "randwrite", "randtrim"]:
t/sgunmap-test.py-122-            parameters = ["--name=test",
t/sgunmap-test.py-123-                           "--time_based",
t/sgunmap-test.py-124-                           "--runtime=30s",
t/sgunmap-test.py-125-                           "--output-format=json",
t/sgunmap-test.py-126-                           "--ioengine=sg",
  • 查找TD_DDIR_RANDWRITE的定义,是由TD_DDIR_WRITETD_DDIR_RAND组成,我们应该关注TD_DDIR_RAND这个参数对程序执行的影响:
$ grep -rHn TD_DDIR_RANDWRITE
io_ddir.h:38:	TD_DDIR_RANDWRITE	= TD_DDIR_WRITE | TD_DDIR_RAND,
options.c:1696:			    .oval = TD_DDIR_RANDWRITE,
  • 查找TD_DDIR_RAND的定义和引用处,主要是被td_random宏引用,应该是作为标志位判断使用:
$ grep -rHn TD_DDIR_RAND -C3
io_ddir.h-31-enum td_ddir {
io_ddir.h-32-	TD_DDIR_READ		= 1 << 0,
io_ddir.h-33-	TD_DDIR_WRITE		= 1 << 1,
io_ddir.h:34:	TD_DDIR_RAND		= 1 << 2,
io_ddir.h-35-	TD_DDIR_TRIM		= 1 << 3,
io_ddir.h-36-	TD_DDIR_RW		= TD_DDIR_READ | TD_DDIR_WRITE,
io_ddir.h:37:	TD_DDIR_RANDREAD	= TD_DDIR_READ | TD_DDIR_RAND,
io_ddir.h:38:	TD_DDIR_RANDWRITE	= TD_DDIR_WRITE | TD_DDIR_RAND,
io_ddir.h:39:	TD_DDIR_RANDRW		= TD_DDIR_RW | TD_DDIR_RAND,
io_ddir.h:40:	TD_DDIR_RANDTRIM	= TD_DDIR_TRIM | TD_DDIR_RAND,
io_ddir.h-41-	TD_DDIR_TRIMWRITE	= TD_DDIR_TRIM | TD_DDIR_WRITE,
io_ddir.h-42-};
io_ddir.h-43-
--
io_ddir.h-45-#define td_write(td)		((td)->o.td_ddir & TD_DDIR_WRITE)
io_ddir.h-46-#define td_trim(td)		((td)->o.td_ddir & TD_DDIR_TRIM)
io_ddir.h-47-#define td_rw(td)		(((td)->o.td_ddir & TD_DDIR_RW) == TD_DDIR_RW)
io_ddir.h:48:#define td_random(td)		((td)->o.td_ddir & TD_DDIR_RAND)
io_ddir.h-49-#define file_randommap(td, f)	(!(td)->o.norandommap && fio_file_axmap((f)))
io_ddir.h-50-#define td_trimwrite(td)	(((td)->o.td_ddir & TD_DDIR_TRIMWRITE) \
io_ddir.h-51-					== TD_DDIR_TRIMWRITE)
--
options.c-1689-			    .help = "Sequential trim",
options.c-1690-			  },
options.c-1691-			  { .ival = "randread",
options.c:1692:			    .oval = TD_DDIR_RANDREAD,
options.c-1693-			    .help = "Random read",
options.c-1694-			  },
options.c-1695-			  { .ival = "randwrite",
options.c:1696:			    .oval = TD_DDIR_RANDWRITE,
options.c-1697-			    .help = "Random write",
options.c-1698-			  },
options.c-1699-			  { .ival = "randtrim",
options.c:1700:			    .oval = TD_DDIR_RANDTRIM,
options.c-1701-			    .help = "Random trim",
options.c-1702-			  },
options.c-1703-			  { .ival = "rw",
--
options.c-1709-			    .help = "Sequential read and write mix",
options.c-1710-			  },
options.c-1711-			  { .ival = "randrw",
options.c:1712:			    .oval = TD_DDIR_RANDRW,
options.c-1713-			    .help = "Random read and write mix"
options.c-1714-			  },
options.c-1715-			  { .ival = "trimwrite",
  • 接着查找td_random的定义和引用,重点关注在io_u.c中的影响,因为这里才是序列生成的主要位置,发现当时random时,调用了get_next_rand_block函数,应该就是生成随机数的位置:
$ grep -rHn td_random -C5 io_u.c
io_u.c-416-	assert(ddir_rw(ddir));
io_u.c-417-
io_u.c-418-	b = offset = -1ULL;
io_u.c-419-
io_u.c-420-	if (rw_seq) {
io_u.c:421:		if (td_random(td)) {
io_u.c-422-			if (should_do_random(td, ddir)) {
io_u.c-423-				ret = get_next_rand_block(td, f, ddir, &b);
io_u.c-424-				*is_random = true;
io_u.c-425-			} else {
io_u.c-426-				*is_random = false;
--
io_u.c-934-	}
io_u.c-935-
io_u.c-936-	/*
io_u.c-937-	 * mark entry before potentially trimming io_u
io_u.c-938-	 */
io_u.c:939:	if (td_random(td) && file_randommap(td, io_u->file))
io_u.c-940-		io_u->buflen = mark_random_map(td, io_u, offset, io_u->buflen);
io_u.c-941-
io_u.c-942-out:
io_u.c-943-	dprint_io_u(io_u, "fill");
io_u.c-944-	td->zone_bytes += io_u->buflen;
  • io_u.c中查看get_next_rand_block的代码,最终调用了之前已经分析到的调用dprint的位置,使用__randrand_max进行随机数计算:
static int get_next_rand_block(struct thread_data *td, struct fio_file *f,
			       enum fio_ddir ddir, uint64_t *b)
{
	if (!get_next_rand_offset(td, f, ddir, b))
		return 0;

	if (td->o.time_based ||
	    (td->o.file_service_type & __FIO_FSERVICE_NONUNIFORM)) {
		fio_file_reset(td, f);
		loop_cache_invalidate(td, f);
		if (!get_next_rand_offset(td, f, ddir, b))
			return 0;
	}

	dprint(FD_IO, "%s: rand offset failed, last=%llu, size=%llu\n",
			f->file_name, (unsigned long long) f->last_pos[ddir],
			(unsigned long long) f->real_file_size);
	return 1;
}

static int get_next_rand_offset(struct thread_data *td, struct fio_file *f,
				enum fio_ddir ddir, uint64_t *b)
{
	if (td->o.random_distribution == FIO_RAND_DIST_RANDOM) {
		uint64_t lastb;

		lastb = last_block(td, f, ddir);
		if (!lastb)
			return 1;

		return __get_next_rand_offset(td, f, ddir, b, lastb);
	} else if (td->o.random_distribution == FIO_RAND_DIST_ZIPF)
		return __get_next_rand_offset_zipf(td, f, ddir, b);
	else if (td->o.random_distribution == FIO_RAND_DIST_PARETO)
		return __get_next_rand_offset_pareto(td, f, ddir, b);
	else if (td->o.random_distribution == FIO_RAND_DIST_GAUSS)
		return __get_next_rand_offset_gauss(td, f, ddir, b);
	else if (td->o.random_distribution == FIO_RAND_DIST_ZONED)
		return __get_next_rand_offset_zoned(td, f, ddir, b);
	else if (td->o.random_distribution == FIO_RAND_DIST_ZONED_ABS)
		return __get_next_rand_offset_zoned_abs(td, f, ddir, b);

	log_err("fio: unknown random distribution: %d\n", td->o.random_distribution);
	return 1;
}

static int __get_next_rand_offset(struct thread_data *td, struct fio_file *f,
				  enum fio_ddir ddir, uint64_t *b,
				  uint64_t lastb)
{
	uint64_t r;

	if (td->o.random_generator == FIO_RAND_GEN_TAUSWORTHE ||
	    td->o.random_generator == FIO_RAND_GEN_TAUSWORTHE64) {

		r = __rand(&td->random_state);

		dprint(FD_RANDOM, "off rand %llu\n", (unsigned long long) r);

		*b = lastb * (r / (rand_max(&td->random_state) + 1.0));
	} else {
		uint64_t off = 0;

		assert(fio_file_lfsr(f));

		if (lfsr_next(&f->lfsr, &off))
			return 1;

		*b = off;
	}

	/*
	 * if we are not maintaining a random map, we are done.
	 */
	if (!file_randommap(td, f))
		goto ret;

	/*
	 * calculate map offset and check if it's free
	 */
	if (random_map_free(f, *b))
		goto ret;

	dprint(FD_RANDOM, "get_next_rand_offset: offset %llu busy\n",
						(unsigned long long) *b);

	*b = axmap_next_free(f->io_axmap, *b);
	if (*b == (uint64_t) -1ULL)
		return 1;
ret:
	return 0;
}

分析随机数计算

  • 查找__rand函数,是定义在lib/rand.h中的一个静态内联函数:
$ grep -rHn " __rand("
backend.c:1012:				io_u->rand_seed = __rand(&td->verify_state);
backend.c:1014:					io_u->rand_seed *= __rand(&td->verify_state);
engines/rdma.c:715:			index = __rand(&rd->rand_state) % rd->rmt_nr;
engines/rdma.c:725:			index = __rand(&rd->rand_state) % rd->rmt_nr;
filesetup.c:337:	r = __rand(&td->file_size_state);
io_u.c:96:		r = __rand(&td->random_state);
io_u.c:548:		r = __rand(&td->bsrange_state[ddir]);
io_u.c:1165:		r = __rand(&td->next_file_state);
lib/gauss.c:16:	r = __rand(&gs->r);
lib/gauss.c:28:		sum += __rand(&gs->r) % (gs->nranges + 1);
lib/rand.c:128:	unsigned long r = __rand(fs);
lib/rand.c:131:		r *= (unsigned long) __rand(fs);
lib/rand.c:190:	unsigned long r = __rand(fs);
lib/rand.c:193:		r *= (unsigned long) __rand(fs);
lib/rand.h:96:static inline uint64_t __rand(struct frand_state *state)
lib/zipf.c:32:	zs->rand_off = __rand(&zs->rand);
lib/zipf.c:55:	rand_uni = (double) __rand(&zs->rand) / (double) FRAND32_MAX;
lib/zipf.c:82:	double rand = (double) __rand(&zs->rand) / (double) FRAND32_MAX;
trim.c:77:	r = __rand(&td->trim_state);
verify.c:1370:			io_u->rand_seed = __rand(&td->verify_state);
verify.c:1372:				io_u->rand_seed *= __rand(&td->verify_state);
  • 查看该函数的实现,其最终计算结果只与输入形参有关,实实在在的伪随机:
struct taus88_state {
	unsigned int s1, s2, s3;
};

struct taus258_state {
	uint64_t s1, s2, s3, s4, s5;
};

struct frand_state {
	unsigned int use64;
	union {
		struct taus88_state state32;
		struct taus258_state state64;
	};
};

static inline unsigned int __rand32(struct taus88_state *state)
{
#define TAUSWORTHE(s,a,b,c,d) ((s&c)<<d) ^ (((s <<a) ^ s)>>b)

	state->s1 = TAUSWORTHE(state->s1, 13, 19, 4294967294UL, 12);
	state->s2 = TAUSWORTHE(state->s2, 2, 25, 4294967288UL, 4);
	state->s3 = TAUSWORTHE(state->s3, 3, 11, 4294967280UL, 17);

	return (state->s1 ^ state->s2 ^ state->s3);
}

static inline uint64_t __rand64(struct taus258_state *state)
{
	uint64_t xval;

	xval = ((state->s1 <<  1) ^ state->s1) >> 53;
	state->s1 = ((state->s1 & 18446744073709551614ULL) << 10) ^ xval;

	xval = ((state->s2 << 24) ^ state->s2) >> 50;
	state->s2 = ((state->s2 & 18446744073709551104ULL) <<  5) ^ xval;

	xval = ((state->s3 <<  3) ^ state->s3) >> 23;
	state->s3 = ((state->s3 & 18446744073709547520ULL) << 29) ^ xval;

	xval = ((state->s4 <<  5) ^ state->s4) >> 24;
	state->s4 = ((state->s4 & 18446744073709420544ULL) << 23) ^ xval;

	xval = ((state->s5 <<  3) ^ state->s5) >> 33;
	state->s5 = ((state->s5 & 18446744073701163008ULL) <<  8) ^ xval;

	return (state->s1 ^ state->s2 ^ state->s3 ^ state->s4 ^ state->s5);
}

static inline uint64_t __rand(struct frand_state *state)
{
	if (state->use64)
		return __rand64(&state->state64);
	else
		return __rand32(&state->state32);
}
  • 再次查找rand_max的实现,依旧定义在lib/rand.h中:
$ grep -rHn " rand_max("
filesetup.c:336:	frand_max = rand_max(&td->file_size_state);
io_u.c:546:	frand_max = rand_max(&td->bsrange_state[ddir]);
io_u.c:1162:		uint64_t frand_max = rand_max(&td->next_file_state);
lib/rand.h:27:static inline uint64_t rand_max(struct frand_state *state)
trim.c:76:	frand_max = rand_max(&td->trim_state);
  • 输出也是严重依赖输入,依旧不是随机计算:
#define FRAND32_MAX	(-1U)
#define FRAND64_MAX	(-1ULL)

static inline uint64_t rand_max(struct frand_state *state)
{
	if (state->use64)
		return FRAND64_MAX;
	else
		return FRAND32_MAX;
}

分析种子初始化

  • 从上面可以看出,整机计算结果只与输入的random_state有关,那么接下来查找一下random_state的初始化和引用处,最终发现random_state使用init_rand_seed进行初始化,之后只有__randrand_max函数会对其进行变更:
$ grep -rHn random_state
fio.h:357:	struct frand_state random_state;
init.c:1056:	init_rand_seed(&td->random_state, td->rand_seeds[FIO_RAND_BLOCK_OFF], use64);
io_u.c:96:		r = __rand(&td->random_state);
io_u.c:100:		*b = lastb * (r / (rand_max(&td->random_state) + 1.0));
verify.c:1623:		if (td->random_state.use64) {
verify.c:1624:			s->rand.state64.s[0] = cpu_to_le64(td->random_state.state64.s1);
verify.c:1625:			s->rand.state64.s[1] = cpu_to_le64(td->random_state.state64.s2);
verify.c:1626:			s->rand.state64.s[2] = cpu_to_le64(td->random_state.state64.s3);
verify.c:1627:			s->rand.state64.s[3] = cpu_to_le64(td->random_state.state64.s4);
verify.c:1628:			s->rand.state64.s[4] = cpu_to_le64(td->random_state.state64.s5);
verify.c:1632:			s->rand.state32.s[0] = cpu_to_le32(td->random_state.state32.s1);
verify.c:1633:			s->rand.state32.s[1] = cpu_to_le32(td->random_state.state32.s2);
verify.c:1634:			s->rand.state32.s[2] = cpu_to_le32(td->random_state.state32.s3);
  • 再查看init_rand_seed的实现,依旧是伪随机计算,只与输入形参seed有关:
// lib/rand.h
static inline uint64_t __rand64(struct taus258_state *state)
{
	uint64_t xval;

	xval = ((state->s1 <<  1) ^ state->s1) >> 53;
	state->s1 = ((state->s1 & 18446744073709551614ULL) << 10) ^ xval;

	xval = ((state->s2 << 24) ^ state->s2) >> 50;
	state->s2 = ((state->s2 & 18446744073709551104ULL) <<  5) ^ xval;

	xval = ((state->s3 <<  3) ^ state->s3) >> 23;
	state->s3 = ((state->s3 & 18446744073709547520ULL) << 29) ^ xval;

	xval = ((state->s4 <<  5) ^ state->s4) >> 24;
	state->s4 = ((state->s4 & 18446744073709420544ULL) << 23) ^ xval;

	xval = ((state->s5 <<  3) ^ state->s5) >> 33;
	state->s5 = ((state->s5 & 18446744073701163008ULL) <<  8) ^ xval;

	return (state->s1 ^ state->s2 ^ state->s3 ^ state->s4 ^ state->s5);
}

// lib/rand.c
static inline uint64_t __seed(uint64_t x, uint64_t m)
{
	return (x < m) ? x + m : x;
}

static void __init_rand32(struct taus88_state *state, unsigned int seed)
{
	int cranks = 6;

#define LCG(x, seed)  ((x) * 69069 ^ (seed))

	state->s1 = __seed(LCG((2^31) + (2^17) + (2^7), seed), 1);
	state->s2 = __seed(LCG(state->s1, seed), 7);
	state->s3 = __seed(LCG(state->s2, seed), 15);

	while (cranks--)
		__rand32(state);
}

static void __init_rand64(struct taus258_state *state, uint64_t seed)
{
	int cranks = 6;

#define LCG64(x, seed)  ((x) * 6906969069ULL ^ (seed))

	state->s1 = __seed(LCG64((2^31) + (2^17) + (2^7), seed), 1);
	state->s2 = __seed(LCG64(state->s1, seed), 7);
	state->s3 = __seed(LCG64(state->s2, seed), 15);
	state->s4 = __seed(LCG64(state->s3, seed), 33);
	state->s5 = __seed(LCG64(state->s4, seed), 49);

	while (cranks--)
		__rand64(state);
}

void init_rand(struct frand_state *state, bool use64)
{
	state->use64 = use64;

	if (!use64)
		__init_rand32(&state->state32, 1);
	else
		__init_rand64(&state->state64, 1);
}

void init_rand_seed(struct frand_state *state, unsigned int seed, bool use64)
{
	state->use64 = use64;

	if (!use64)
		__init_rand32(&state->state32, seed);
	else
		__init_rand64(&state->state64, seed);
}

分析种子来源

  • 再查看一下seed这个形参td->rand_seeds[FIO_RAND_BLOCK_OFF]的引用处,是在init.c1054处进行初始化的:
$ grep -rHn "rand_seeds\[FIO_RAND_BLOCK_OFF\]"
filesetup.c:1292:			seed = td->rand_seeds[FIO_RAND_BLOCK_OFF];
filesetup.c:1882:		lfsr_reset(&f->lfsr, td->rand_seeds[FIO_RAND_BLOCK_OFF]);
init.c:1054:		td->rand_seeds[FIO_RAND_BLOCK_OFF] = FIO_RANDSEED * td->thread_number;
init.c:1056:	init_rand_seed(&td->random_state, td->rand_seeds[FIO_RAND_BLOCK_OFF], use64);
  • 查看一下该出函数的实现,发现依赖进程的thread_number变量:
// init.c
static void td_fill_rand_seeds_internal(struct thread_data *td, bool use64)
{
	unsigned int read_seed = td->rand_seeds[FIO_RAND_BS_OFF];
	unsigned int write_seed = td->rand_seeds[FIO_RAND_BS1_OFF];
	unsigned int trim_seed = td->rand_seeds[FIO_RAND_BS2_OFF];
	int i;

	/*
	 * trimwrite is special in that we need to generate the same
	 * offsets to get the "write after trim" effect. If we are
	 * using bssplit to set buffer length distributions, ensure that
	 * we seed the trim and write generators identically. Ditto for
	 * verify, read and writes must have the same seed, if we are doing
	 * read verify.
	 */
	if (td->o.verify != VERIFY_NONE)
		write_seed = read_seed;
	if (td_trimwrite(td))
		trim_seed = write_seed;
	init_rand_seed(&td->bsrange_state[DDIR_READ], read_seed, use64);
	init_rand_seed(&td->bsrange_state[DDIR_WRITE], write_seed, use64);
	init_rand_seed(&td->bsrange_state[DDIR_TRIM], trim_seed, use64);

	td_fill_verify_state_seed(td);
	init_rand_seed(&td->rwmix_state, td->rand_seeds[FIO_RAND_MIX_OFF], false);

	if (td->o.file_service_type == FIO_FSERVICE_RANDOM)
		init_rand_seed(&td->next_file_state, td->rand_seeds[FIO_RAND_FILE_OFF], use64);
	else if (td->o.file_service_type & __FIO_FSERVICE_NONUNIFORM)
		init_rand_file_service(td);

	init_rand_seed(&td->file_size_state, td->rand_seeds[FIO_RAND_FILE_SIZE_OFF], use64);
	init_rand_seed(&td->trim_state, td->rand_seeds[FIO_RAND_TRIM_OFF], use64);
	init_rand_seed(&td->delay_state, td->rand_seeds[FIO_RAND_START_DELAY], use64);
	init_rand_seed(&td->poisson_state[0], td->rand_seeds[FIO_RAND_POISSON_OFF], 0);
	init_rand_seed(&td->poisson_state[1], td->rand_seeds[FIO_RAND_POISSON2_OFF], 0);
	init_rand_seed(&td->poisson_state[2], td->rand_seeds[FIO_RAND_POISSON3_OFF], 0);
	init_rand_seed(&td->dedupe_state, td->rand_seeds[FIO_DEDUPE_OFF], false);
	init_rand_seed(&td->zone_state, td->rand_seeds[FIO_RAND_ZONE_OFF], false);

	if (!td_random(td))
		return;

	if (td->o.rand_repeatable)
		td->rand_seeds[FIO_RAND_BLOCK_OFF] = FIO_RANDSEED * td->thread_number;

	init_rand_seed(&td->random_state, td->rand_seeds[FIO_RAND_BLOCK_OFF], use64);

	for (i = 0; i < DDIR_RWDIR_CNT; i++) {
		struct frand_state *s = &td->seq_rand_state[i];

		init_rand_seed(s, td->rand_seeds[FIO_RAND_SEQ_RAND_READ_OFF], false);
	}
}
  • 在查找thread_number变量的初始化和引用位置,发现HOWTO里面貌似有解释:
$ grep -rHn thread_number
HOWTO:1245:	* thread_number`, where the thread number is a counter that starts at 0 and
backend.c:64:unsigned int thread_number = 0;
backend.c:1624:			ret = fio_cpus_split(&o->cpumask, td->thread_number - 1);
backend.c:1899:		verify_save_state(td->thread_number);
backend.c:2131:					td->thread_number - 1, &data);
backend.c:2248:	todo = thread_number;
backend.c:2254:		print_status_init(td->thread_number - 1);
backend.c:2488:	if (!thread_number)
client.c:923:	pdu.thread_number = cpu_to_le32(client->thread_number);
client.c:948:	dst->thread_number	= le32_to_cpu(src->thread_number);
client.c:1078:	if (client->opt_lists && p->ts.thread_number <= client->jobs)
client.c:1079:		opt_list = &client->opt_lists[p->ts.thread_number - 1];
client.c:1095:	client_ts.thread_number = p->ts.thread_number;
client.c:1653:	ret->thread_number	= le32_to_cpu(ret->thread_number);
client.c:1832:		client->thread_number = le32_to_cpu(pdu->thread_number);
client.h:60:	uint32_t thread_number;
eta.c:41:	char c = __run_str[td->thread_number - 1];
eta.c:118:	__run_str[td->thread_number - 1] = c;
eta.c:411:	eta_secs = malloc(thread_number * sizeof(uint64_t));
eta.c:412:	memset(eta_secs, 0, thread_number * sizeof(uint64_t));
eta.c:530:	je->nr_threads = thread_number;
eta.c:704:	if (!thread_number)
filesetup.c:1203:	seed = jhash(f->file_name, strlen(f->file_name), 0) * td->thread_number;
fio.1:1012:* thread_number', where the thread number is a counter that starts at 0 and
fio.h:183:	unsigned int thread_number;
fio.h:509:extern unsigned int thread_number;
fio.h:701:	for ((i) = 0, (td) = &threads[0]; (i) < (int) thread_number; (i)++, (td)++)
gclient.c:299:	client_ts.thread_number = p->ts.thread_number;
gclient.c:578:	p->thread_number = le32_to_cpu(p->thread_number);
init.c:480:	if (thread_number >= max_jobs) {
init.c:486:	td = &threads[thread_number++];
init.c:505:	td->thread_number = thread_number;
init.c:536:	memset(&threads[td->thread_number - 1], 0, sizeof(*td));
init.c:537:	thread_number--;
init.c:1054:		td->rand_seeds[FIO_RAND_BLOCK_OFF] = FIO_RANDSEED * td->thread_number;
init.c:1073:			td->rand_seeds[i] = FIO_RANDSEED * td->thread_number
init.c:1235:		td->rand_seeds[i] = seed * td->thread_number + i;
init.c:1565:				td->thread_number, suf, o->per_job_logs);
init.c:1569:				td->thread_number, suf, o->per_job_logs);
init.c:1573:				td->thread_number, suf, o->per_job_logs);
init.c:1605:				td->thread_number, suf, o->per_job_logs);
init.c:1637:				td->thread_number, suf, o->per_job_logs);
init.c:1668:				td->thread_number, suf, o->per_job_logs);
init.c:3004:	if (!thread_number) {
libfio.c:160:	thread_number = 0;
server.c:758:	spdu.jobs = cpu_to_le32(thread_number);
server.c:801:	spdu.jobs = cpu_to_le32(thread_number);
server.c:842:	spdu.jobs = cpu_to_le32(thread_number);
server.c:943:	tnumber = le32_to_cpu(pdu->thread_number);
server.c:947:	if (!tnumber || tnumber > thread_number) {
server.c:1478:	p.ts.thread_number	= cpu_to_le32(ts->thread_number);
server.c:1958:		.thread_number		= cpu_to_le32(td->thread_number),
server.c:2029:		.thread_number = cpu_to_le32(td->thread_number),
server.h:172:	uint32_t thread_number;
server.h:192:	uint32_t thread_number;
stat.c:1782:			ts->thread_number = td->thread_number;
stat.c:1998:	rt = malloc(thread_number * sizeof(unsigned long long));
stat.h:152:	uint32_t thread_number;
stat.h:365:#define THREAD_RUNSTR_SZ	__THREAD_RUNSTR_SZ(thread_number)
verify.c:1168:	hdr->thread = td->thread_number;
verify.c:1797:	fd = open_state_file(td->o.name, prefix, td->thread_number - 1, 0);
  • HOWTO看看,这个变量其实就是一个进程编号,我们这个一直是单进程测试,这个值就是0
$ grep -rHn thread_number HOWTO -C5
HOWTO-1240-	offset is aligned to the minimum block size.
HOWTO-1241-
HOWTO-1242-.. option:: offset_increment=int
HOWTO-1243-
HOWTO-1244-	If this is provided, then the real offset becomes `offset + offset_increment
HOWTO:1245:	* thread_number`, where the thread number is a counter that starts at 0 and
HOWTO-1246-	is incremented for each sub-job (i.e. when :option:`numjobs` option is
HOWTO-1247-	specified). This option is useful if there are several jobs which are
HOWTO-1248-	intended to operate on a file in parallel disjoint segments, with even
HOWTO-1249-	spacing between the starting points.
HOWTO-1250-
  • init.c里查找一下他的引用处,第480行附近的语句貌似是主要修改这个值的位置:
$ grep -rHn thread_number init.c 
init.c:480:	if (thread_number >= max_jobs) {
init.c:486:	td = &threads[thread_number++];
init.c:505:	td->thread_number = thread_number;
init.c:536:	memset(&threads[td->thread_number - 1], 0, sizeof(*td));
init.c:537:	thread_number--;
init.c:1054:		td->rand_seeds[FIO_RAND_BLOCK_OFF] = FIO_RANDSEED * td->thread_number;
init.c:1073:			td->rand_seeds[i] = FIO_RANDSEED * td->thread_number
init.c:1235:		td->rand_seeds[i] = seed * td->thread_number + i;
init.c:1565:				td->thread_number, suf, o->per_job_logs);
init.c:1569:				td->thread_number, suf, o->per_job_logs);
init.c:1573:				td->thread_number, suf, o->per_job_logs);
init.c:1605:				td->thread_number, suf, o->per_job_logs);
init.c:1637:				td->thread_number, suf, o->per_job_logs);
init.c:1668:				td->thread_number, suf, o->per_job_logs);
init.c:3004:	if (!thread_number) {
  • init.c中进行详细查看,thread_number这个值的确是岁进程数的增加而进行递加的:
// init.c
/*
 * Return a free job structure.
 */
static struct thread_data *get_new_job(bool global, struct thread_data *parent,
				       bool preserve_eo, const char *jobname)
{
	struct thread_data *td;

	if (global)
		return &def_thread;
	if (setup_thread_area()) {
		log_err("error: failed to setup shm segment\n");
		return NULL;
	}
	if (thread_number >= max_jobs) {
		log_err("error: maximum number of jobs (%d) reached.\n",
				max_jobs);
		return NULL;
	}

	td = &threads[thread_number++];
	*td = *parent;

	INIT_FLIST_HEAD(&td->opt_list);
	if (parent != &def_thread)
		copy_opt_list(td, parent);

	td->io_ops = NULL;
	td->io_ops_init = 0;
	if (!preserve_eo)
		td->eo = NULL;

	td->o.uid = td->o.gid = -1U;

	dup_files(td, parent);
	fio_options_mem_dupe(td);

	profile_add_hooks(td);

	td->thread_number = thread_number;
	td->subjob_number = 0;

	if (jobname)
		td->o.name = strdup(jobname);

	if (!parent->o.group_reporting || parent == &def_thread)
		stat_number++;

	return td;
}

最终总结

  • 整个随机序列的数值生成只依赖于创建这个Job的序号,以及整个测试序列的数量,这是实实在在的伪随机。之前的--debug=random中第一列是PID,第二列就是最终生成的序列序号;
  • 先使用randwrite在文件系统中进行数据写入,如果文件系统为空,且写入期间没有其他文件被写入,那么虽然逻辑上是随机序列,但实际被文件系统分配到连续的物理扇区上(同时,写操作没有被块设备调度器进行合并,所以随机写的IO次数并未被变少);
  • 接着再进行randread进行IOPS测试时,由于伪随机的原因,和之前randwrite的逻辑随机序列是一样的,但经过文件系统到达块设备时,就是物理上连续的扇区,会被块设备调度器合并,变成顺序读,IO次数减少,所以这时测试的IOPS会偏大(至于为啥写没被合并,读却被合并,这跟文件系统和调度器算法中,读写合并等待时间以及队列长度等参数不同有关);
  • 在使用fio进行随机读测试之前,应该使用dd或者fio的顺序写初始化文件,使读写的序列顺序不同,即可规避这个问题;
  • 至于随机写的测试,最好也要在测试前对文件系统进行一些列不规则的文件增删读写,使得文件系统的文件扇区分配不再连续,以便得出更靠谱的测试结果。

© 著作权归作者所有

共有 人打赏支持
LastRitter
粉丝 37
博文 40
码字总数 190940
作品 0
武汉
高级程序员
私信 提问
磁盘io性能分析

一,概念 磁盘io,顾名思义就是磁盘的输入输出。即向磁盘写入数据和从磁盘读取数据。 I/O 读写的类型,大体上讲,I/O 的类型可以分为: 读 / 写 I/O、大 / 小块 I/O、连续 / 随机 I/O, 顺序 ...

leejia1989
2014/09/15
0
0
硬盘测试——fio

fio(Flexible IO Tester)FIO是测试IOPS的非常好的工具,用来对硬件进行压力测试和验证,支持13种不同的I/O引擎,包括:sync,mmap, libaio, posixaio, SG v3, splice, null, network, syslet,...

jesse-lin
2016/11/06
39
0
服务器性能分析

磁盘IO 在进行下列测试前,请确保磁盘已经 4K 对齐。 上述测试时 fio 相关参数说明:

295631788
2017/12/29
0
0
linux使用FIO测试磁盘的iops

FIO是测试IOPS的非常好的工具,用来对硬件进行压力测试和验证,支持13种不同的I/O引擎,包括:sync,mmap, libaio, posixaio, SG v3, splice, null, network, syslet, guasi, solarisaio 等等。...

qianghong000
06/29
0
0
磁盘性能指标--IOPS、吞吐量及测试

一、概念介绍: 磁盘性能指标--IOPS ---------------------------------------------------------- IOPS (Input/Output Per Second)即每秒的输入输出量(或读写次数),是衡量磁盘性能的主要指...

wushank
06/29
0
0

没有更多内容

加载失败,请刷新页面

加载更多

mac 下 mysql 8.0.13 安装并记录遇到的问题 以便以后查看

安装 官网mysql 下载地址 安装过程 省去 安装好之后 下载navicat 错误1 链接 遇到 mysql 2003 - Can't connect to MySQL server 错误, 解决方案 重启mysql 服务 #错误2 ERROR 1045: Acces...

杭州-IT攻城狮
26分钟前
3
0

中国龙-扬科
29分钟前
1
0
[Spring4.x]基于spring4.x纯注解的Web工程搭建

在前文中已经说明了如何基于 Spring4.x+ 版本开发纯注解的非web项目,链接如下: https://my.oschina.net/morpheusWB/blog/2985600 本文则主要说明,如何在Web项目中,"基于spring纯注解方式...

morpheusWB
58分钟前
13
0
基础编程题目集-7-13 日K蜡烛图

股票价格涨跌趋势,常用蜡烛图技术中的K线图来表示,分为按日的日K线、按周的周K线、按月的月K线等。以日K线为例,每天股票价格从开盘到收盘走完一天,对应一根蜡烛小图,要表示四个价格:开...

niithub
今天
5
0
Jenkins window 下的安装使用

1.下载:https://jenkins.io/download/ 双击安装完毕,将自动打开浏览器: http://localhost:8080 打开对应位置的文件,将初始密钥粘贴至输入框。 第一个是 安装默认的软件;第二个是 自定义...

狼王黄师傅
今天
7
0

没有更多内容

加载失败,请刷新页面

加载更多

返回顶部
顶部