Ceph BackoffThrottle分析

来源：互联网发布：农村淘宝怎么买东西编辑：程序博客网时间：2024/06/06 01:00

本文讨论下Ceph在Jewel中引入的 dynamic throttle：BackoffThrottle；分析后优化Ceph filestore，journal相关的throttle配置；

参考文章：

http://blog.wjin.org/posts/ceph-dynamic-throttle.html

https://fossies.org/linux/ceph/src/doc/dynamic-throttle.txt

BackoffThrottle

Jewel引入了dynamic的throttle，就是代码中BackoffThrottle，现在filestore和Journal都是使用它来做throttle的；

class FileStore
{
    BackoffThrottle throttle_ops, throttle_bytes;
}
 
class JournalThrottle {
    BackoffThrottle throttle;
…
}

BackoffThrottle定义和相关参数如下：

/**
 * BackoffThrottle
 *
 * Creates a throttle which gradually induces delays when get() is called
 * based on params low_threshhold, high_threshhold, expected_throughput,
 * high_multiple, and max_multiple.
 *
 * In [0, low_threshhold), we want no delay.
 *
 * In [low_threshhold, high_threshhold), delays should be injected based
 * on a line from 0 at low_threshhold to
 * high_multiple * (1/expected_throughput) at high_threshhold.
 *
 * In [high_threshhold, 1), we want delays injected based on a line from
 * (high_multiple * (1/expected_throughput)) at high_threshhold to
 * (high_multiple * (1/expected_throughput)) +
 * (max_multiple * (1/expected_throughput)) at 1.
 *
 * Let the current throttle ratio (current/max) be r, low_threshhold be l,
 * high_threshhold be h, high_delay (high_multiple / expected_throughput) be e,
 * and max_delay (max_muliple / expected_throughput) be m.
 *
 * delay = 0, r \in [0, l)
 * delay = (r - l) * (e / (h - l)), r \in [l, h)
 * delay = h + (r - h)((m - e)/(1 - h))
 */
class BackoffThrottle {
…
    /// see above, values are in [0, 1].
    double low_threshhold = 0;
    double high_threshhold = 1;
 
    /// see above, values are in seconds
    double high_delay_per_count = 0;
    double max_delay_per_count = 0;
 
    /// Filled in in set_params
    double s0 = 0; ///< e / (h - l), l != h, 0 otherwise
    double s1 = 0; ///< (m - e)/(1 - h), 1 != h, 0 otherwise
 
    /// max
    uint64_t max = 0;
    uint64_t current = 0;
…
}

filestore throttle举例分析

下面以使用BackoffThrottle的filestore throttle举例分析下其参数配置

filestore throttle的相关配置项

OPTION(filestore_expected_throughput_bytes, OPT_DOUBLE, 200 << 20)
OPTION(filestore_expected_throughput_ops, OPT_DOUBLE, 200)
  
OPTION(filestore_queue_max_bytes, OPT_U64, 100 << 20)
OPTION(filestore_queue_max_ops, OPT_U64, 50)
  
OPTION(filestore_queue_max_delay_multiple, OPT_DOUBLE, 0)
OPTION(filestore_queue_high_delay_multiple, OPT_DOUBLE, 0)
  
OPTION(filestore_queue_low_threshhold, OPT_DOUBLE, 0.3)
OPTION(filestore_queue_high_threshhold, OPT_DOUBLE, 0.9)

根据配置项初始化BackoffThrottle

bool BackoffThrottle::set_params(
    double _low_threshhold,
    double _high_threshhold,
    double _expected_throughput,
    double _high_multiple,
    double _max_multiple,
    uint64_t _throttle_max,
    ostream *errstream)
{
    low_threshhold = _low_threshhold;
    high_threshhold = _high_threshhold;
    high_delay_per_count = _high_multiple / _expected_throughput;
    max_delay_per_count = _max_multiple / _expected_throughput;
    max = _throttle_max;
 
    if (high_threshhold - low_threshhold > 0) {
        s0 = high_delay_per_count / (high_threshhold - low_threshhold);
    } else {
        low_threshhold = high_threshhold;
        s0 = 0;
    }
 
    if (1 - high_threshhold > 0) {
        s1 = (max_delay_per_count - high_delay_per_count)
             / (1 - high_threshhold);
    } else {
        high_threshhold = 1;
        s1 = 0;
    }
}
 
int FileStore::set_throttle_params()
{
    stringstream ss;
    bool valid = throttle_bytes.set_params(
                     g_conf->filestore_queue_low_threshhold,
                     g_conf->filestore_queue_high_threshhold,
                     g_conf->filestore_expected_throughput_bytes,
                     g_conf->filestore_queue_high_delay_multiple,
                     g_conf->filestore_queue_max_delay_multiple,
                     g_conf->filestore_queue_max_bytes,
                     &ss);
 
    valid &= throttle_ops.set_params(
                 g_conf->filestore_queue_low_threshhold,
                 g_conf->filestore_queue_high_threshhold,
                 g_conf->filestore_expected_throughput_ops,
                 g_conf->filestore_queue_high_delay_multiple,
                 g_conf->filestore_queue_max_delay_multiple,
                 g_conf->filestore_queue_max_ops,
                 &ss);
…
}

获取delay值

std::chrono::duration<double> BackoffThrottle::_get_delay(uint64_t c) const
{
    if (max == 0)
        return std::chrono::duration<double>(0);
    double r = ((double)current) / ((double)max);
    if (r < low_threshhold) {
        return std::chrono::duration<double>(0);
    } else if (r < high_threshhold) {
        return c * std::chrono::duration<double>(
                   (r - low_threshhold) * s0);
    } else {
        return c * std::chrono::duration<double>(
                   high_delay_per_count + ((r - high_threshhold) * s1));
    }
}

如上述函数描述，分四种情况计算delay值：

max = 0时：永远返回 0
current/max < low_threshhold时：返回 0
low_threshhold <= current/max < high_threshhold时：计算一值
high_threshhold <= current/max时：计算一值

如图所示，在第一个区间的时候，也就是压力不大的情况下，delay值为0，是不需要wait的。当压力增大，x落入第二个区间后，delay值开始起作用，并且逐步增大，当压力过大的时候，会落入第三个区间，这时候delay值增加明显加快，wait值明显增大，尽量减慢io速度，减缓压力，故而得名dynamic throttle。

默认情况下filestore throttle分析

filestore有bytes和ops两个throttle，这里以bytes为例分析：

默认情况下：filestore_queue_high_delay_multiple = 0， filestore_queue_max_delay_multiple = 0；

相当于BackoffThrottle中的值如下：

low_threshhold = 0.3
high_threshhold = 0.9
high_delay_per_count = 0
max_delay_per_count = 0
s0 = 0
s1 = 0
max = 100 << 20

所以默认配置下，是关闭dynamic delay的；

开启dynamic throttle

参考最早的代码，配置：

filestore_queue_high_delay_multiple = 2
filestore_queue_max_delay_multiple = 10

其他使用默认值是，BackoffThrottle中的值如下：

low_threshhold = 0.3
high_threshhold = 0.9
high_delay_per_count = 2/(200 << 20)
max_delay_per_count = 10/(200 << 20)
s0 = (2/(200 << 20))/0.6
s1 = (8/(200 << 20))/0.1
max = 100 << 20

则此时的delay分为如下几种：

c： op->bytes，即一次请求的数据量

current：当前filestore queue的数据量，初始化为 0，每次调用：throttle_bytes.get(o->bytes)；{ current + = c;}

current/max < low_threshhold时：此时 current < (30 << 20)；delay = 0
low_threshhold <= current/max < high_threshhold时：
此时 (30 << 20) <= current < (90 << 20)
delay = c * ((current/max - 0.3) * s0)
a）current = 30 << 20时：delay = 0
b）current = 90 << 20时：delay = c / (100 << 20)
high_threshhold <= current/max时:
此时 (90 << 20) < current
delay = c * (2/(200 << 20) + (current/max - 0.9) * s1)
a）current = 90 << 20时：delay = c / (100 << 20)
b）current = 100 << 20时：delay = 5 * c / (100 << 20)

优化配置下的dynamic throttle

配置如下：

filestore_expected_throughput_bytes =  536870912    // 512M
filestore_queue_max_bytes= 1048576000               // 1000M
filestore_queue_low_threshhold = 0.6
filestore_queue_high_threshhold = 0.9   // 默认值
filestore_queue_high_delay_multiple = 2
filestore_queue_max_delay_multiple = 10

BackoffThrottle中的值如下：

low_threshhold = 0.6
high_threshhold = 0.9
high_delay_per_count = 2/(512 << 20)
max_delay_per_count = 10/(512 << 20)
s0 = (2/(512 << 20))/0.3
s1 = (8/(512 << 20))/0.1
max = 1000 << 20

则此时的delay分为如下几种：

current/max < low_threshhold时：此时 current < (600 << 20)；delay = 0
low_threshhold <= current/max < high_threshhold时：
此时 (600 << 20) <= current < (900 << 20)
delay = c * ((current/max - 0.6) * s0)
a）current = 600 << 20时：delay = 0
b）current = 900 << 20时：delay = c / (256 << 20)
high_threshhold <= current/max时:
此时 (900 << 20) < current
delay = c * (2/(512 << 20) + (current/max - 0.9) * s1)
a）current = 900 << 20时：delay = c / (256 << 20)
b）current = 1000 << 20时：delay = 5 * c / (256 << 20)

结论：这里的参数配置不是很合理；600M之前的delay都是0；后续随着current的增大，delay的值小于默认时候的值，可能会加大filestore的压力；

0 0