2. 程式人生 > >第三章：Time Series (A Collection of Observations)

第三章：Time Series (A Collection of Observations)

阿新 • • 發佈：2019-02-08

Chapter 3. Time Series (A Collection of Observations)

時間序列（Time Series）是在一段時間範圍內排序的時間點的值，可用於統計，社交網路，電信工程等一切需要時間度量的場景。它可以來進行預測，如股市走向，環境條件變化等。stock market changes, real estate trends, environmental conditions, and more.

時間序列的例子包括：
* 隨時間變化，報紙中特定詞語的使用
* 年度最低工資變化
* 每日股價波動
* 按月產品購買趨勢
* 氣候變化

時間序列面臨儲存的挑戰，例如如果每秒存一個時間，每天則需86,400個時間點，長時間儲存壓力更大，對於記憶體資料庫如Redis更是如此。

另一個問題是，隨時間推移，之前細粒度的時間點失去價值，例如在後期的統計中，秒級的時間點已沒有意義，通常用小時就夠了。

New York Times提供了一個有趣的工具，可統計自1851年起，在報紙中使用的特定詞語的統計圖。可參見http://chronicle.nytlabs.com
下面是對於china關鍵詞的年度統計圖，不能按月或按天：

Building the foundation

本節演示用Strings, Hash, Sorted Sets 和 HyperLogLogs實現時間序列。

時間粒度支援日，小時，分鐘，秒，每一個粒度一個key，當事件發生時，每一個key都需要更新

時間按時間粒度分組，例如某一秒發生的時間會同時歸類到相應的分鐘，小時和天。

此解決方案非常適合於實時計數，如頁面瀏覽，視訊觀看，點選數，購買商品數等。

此程式的執行方式如下，在命令列中可指定資料型別從而對應相應的實現，但結果是一樣的。

[[email protected] examples]$ cd chapter\ 3
[[email protected] chapter 3]$ ls
timeseries-hash.js         timeseries-sorted-set.js  using-timeseries.js
timeseries-hyperloglog.js  timeseries-string.js      using-timeseries-unique.js
[
 
[email protected] chapter 3]$ node using-timeseries.js string
Results from 1sec:
Timestamp   | Value
--------------- | ------
  0 | 1
  1 | 2
  2 | 0
  3 | 1
  4 | 0

Results from 1min:
Timestamp   | Value
--------------- | ------
  0 | 4
  60  | 1
  120 | 0

[[email protected] chapter 3]$ node using-timeseries.js hash
Results from 1sec:
Timestamp   | Value
--------------- | ------
  0 | 1
  1 | 2
  2 | 0
  3 | 1
  4 | 0

Results from 1min:
Timestamp   | Value
--------------- | ------
  0 | 4
  60  | 1
  120 | 0

為了避免存太多的資料，每一個key都設定了過期時間，如秒級只保留2小時，分鐘級保留7天等，小時級保留60天，天級永不過期.

當事件發生時，每一個key都會被更新, 如下:

TimeSeries.prototype.insert = function(timestampInSeconds) { // 1
  for (var granularityName in this.granularities) { // 2
    var granularity = this.granularities[granularityName]; // 3
    var key = this._getKeyName(granularity, timestampInSeconds); // 4
    this.client.incr(key); // 5
    if (granularity.ttl !== null) { // 6
      this.client.expire(key, granularity.ttl); // 7
    }
  }
};

執行以下的程式碼產生時間序列：

item1Purchases.insert(beginTimestamp); // 7
item1Purchases.insert(beginTimestamp + 1); // 8
item1Purchases.insert(beginTimestamp + 1); // 9
item1Purchases.insert(beginTimestamp + 3); // 10
item1Purchases.insert(beginTimestamp + 61); // 11

通過redis-cli monitor可以看到實際執行的命令:

1462417516.878865 [0 127.0.0.1:63610] "info"
1462417516.884225 [0 127.0.0.1:63610] "flushall"
1462417516.917339 [0 127.0.0.1:63610] "incr" "purchases:item1:1sec:0"
1462417516.917352 [0 127.0.0.1:63610] "expire" "purchases:item1:1sec:0" "7200"
1462417516.917359 [0 127.0.0.1:63610] "incr" "purchases:item1:1min:0"
1462417516.917363 [0 127.0.0.1:63610] "expire" "purchases:item1:1min:0" "604800"
1462417516.917369 [0 127.0.0.1:63610] "incr" "purchases:item1:1hour:0"
1462417516.917373 [0 127.0.0.1:63610] "expire" "purchases:item1:1hour:0" "5184000"
1462417516.917378 [0 127.0.0.1:63610] "incr" "purchases:item1:1day:0"

1462417516.917382 [0 127.0.0.1:63610] "incr" "purchases:item1:1sec:1"
1462417516.917386 [0 127.0.0.1:63610] "expire" "purchases:item1:1sec:1" "7200"
1462417516.917417 [0 127.0.0.1:63610] "incr" "purchases:item1:1min:0"
1462417516.917425 [0 127.0.0.1:63610] "expire" "purchases:item1:1min:0" "604800"
1462417516.917430 [0 127.0.0.1:63610] "incr" "purchases:item1:1hour:0"
1462417516.917435 [0 127.0.0.1:63610] "expire" "purchases:item1:1hour:0" "5184000"
1462417516.917440 [0 127.0.0.1:63610] "incr" "purchases:item1:1day:0"

1462417516.917444 [0 127.0.0.1:63610] "incr" "purchases:item1:1sec:1"
1462417516.917448 [0 127.0.0.1:63610] "expire" "purchases:item1:1sec:1" "7200"
1462417516.917467 [0 127.0.0.1:63610] "incr" "purchases:item1:1min:0"
1462417516.917471 [0 127.0.0.1:63610] "expire" "purchases:item1:1min:0" "604800"
1462417516.917476 [0 127.0.0.1:63610] "incr" "purchases:item1:1hour:0"
1462417516.917481 [0 127.0.0.1:63610] "expire" "purchases:item1:1hour:0" "5184000"
1462417516.917486 [0 127.0.0.1:63610] "incr" "purchases:item1:1day:0"

1462417516.917490 [0 127.0.0.1:63610] "incr" "purchases:item1:1sec:3"
1462417516.917494 [0 127.0.0.1:63610] "expire" "purchases:item1:1sec:3" "7200"
1462417516.917499 [0 127.0.0.1:63610] "incr" "purchases:item1:1min:0"
1462417516.917503 [0 127.0.0.1:63610] "expire" "purchases:item1:1min:0" "604800"
1462417516.917508 [0 127.0.0.1:63610] "incr" "purchases:item1:1hour:0"
1462417516.917513 [0 127.0.0.1:63610] "expire" "purchases:item1:1hour:0" "5184000"
1462417516.917518 [0 127.0.0.1:63610] "incr" "purchases:item1:1day:0"

1462417516.917522 [0 127.0.0.1:63610] "incr" "purchases:item1:1sec:61"
1462417516.917526 [0 127.0.0.1:63610] "expire" "purchases:item1:1sec:61" "7200"
1462417516.917530 [0 127.0.0.1:63610] "incr" "purchases:item1:1min:60"
1462417516.917534 [0 127.0.0.1:63610] "expire" "purchases:item1:1min:60" "604800"
1462417516.917539 [0 127.0.0.1:63610] "incr" "purchases:item1:1hour:0"
1462417516.917550 [0 127.0.0.1:63610] "expire" "purchases:item1:1hour:0" "5184000"
1462417516.917556 [0 127.0.0.1:63610] "incr" "purchases:item1:1day:0"

1462417516.917562 [0 127.0.0.1:63610] "mget" "purchases:item1:1sec:0" "purchases:item1:1sec:1" "purchases:item1:1sec:2" "purchases:item1:1sec:3" "purchases:item1:1sec:4"
1462417516.917576 [0 127.0.0.1:63610] "mget" "purchases:item1:1min:0" "purchases:item1:1min:60" "purchases:item1:1min:120"

這個例子可以很方便的改成記錄股價變動，只需把INCRBY改為SET，把計數改為設定股價。

Optimizing with Hashes

前面String的實現每天會產生87,865個key (60 * 60 * 24 + 60 * 24 + 24 + 1)，太多的key除錯不方便，也消耗記憶體（每日11MB），因此我們用Hash的ziplist來優化(每日800KB)。
具體的benchmark可參見：https://gist.github.com/hltbra/2fbf5310aabbecee68c5

Hash要使用ziplist需要同時滿足以下條件, 否則就會用hash table:

It must have fewer fields than the threshold Set in the configuration hash-max-ziplist-entries. The default value for hash-max-ziplist-entries is 512.
No field value can be bigger than hash-max-ziplist-value. The default value for hash-max-ziplist-value is 64 bytes.

Note

前述String的實現每秒至少產生一個key，為了使用Hash介紹記憶體空間，我們需要把多個key置於一個組中, 每組中key的數量小於512。分組的依據如下：

1sec granularity: Stores a maximum of 300 timestamps of 1 second each (5 minutes of data points)
1min granularity: Stores a maximum of 480 timestamps of 1 minute each (8 hours of data points)
1hour granularity: Stores a maximum of 240 timestamps of 1 hour each (10 days of data points)
1day granularity: Stores a maximum of 30 timestamps of 1 day each (30 days of data points)

大部分程式碼是相同的，關鍵是利用hash進行了分割槽，相當於建立多個bucket。
通過redis-cli monitor可以看到實際執行的命令:

[[email protected] ~]$ redis-cli monitor
OK
1462418086.839729 [0 127.0.0.1:63620] "info"
1462418086.843865 [0 127.0.0.1:63620] "flushall"
1462418086.894362 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1sec:0" "0" "1"
1462418086.894425 [0 127.0.0.1:63620] "expire" "purchases:item1:1sec:0" "7200"
1462418086.894437 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1min:0" "0" "1"
1462418086.894446 [0 127.0.0.1:63620] "expire" "purchases:item1:1min:0" "604800"
1462418086.894455 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1hour:0" "0" "1"
1462418086.894464 [0 127.0.0.1:63620] "expire" "purchases:item1:1hour:0" "5184000"
1462418086.894473 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1day:0" "0" "1"

1462418086.894483 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1sec:0" "1" "1"
1462418086.894492 [0 127.0.0.1:63620] "expire" "purchases:item1:1sec:0" "7200"
1462418086.894500 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1min:0" "0" "1"
1462418086.894511 [0 127.0.0.1:63620] "expire" "purchases:item1:1min:0" "604800"
1462418086.894520 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1hour:0" "0" "1"
1462418086.894529 [0 127.0.0.1:63620] "expire" "purchases:item1:1hour:0" "5184000"
1462418086.894537 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1day:0" "0" "1"

1462418086.894546 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1sec:0" "1" "1"
1462418086.894556 [0 127.0.0.1:63620] "expire" "purchases:item1:1sec:0" "7200"
1462418086.894564 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1min:0" "0" "1"
1462418086.894573 [0 127.0.0.1:63620] "expire" "purchases:item1:1min:0" "604800"
1462418086.894581 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1hour:0" "0" "1"
1462418086.894590 [0 127.0.0.1:63620] "expire" "purchases:item1:1hour:0" "5184000"
1462418086.894598 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1day:0" "0" "1"

1462418086.894606 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1sec:0" "3" "1"
1462418086.894615 [0 127.0.0.1:63620] "expire" "purchases:item1:1sec:0" "7200"
1462418086.894623 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1min:0" "0" "1"
1462418086.894631 [0 127.0.0.1:63620] "expire" "purchases:item1:1min:0" "604800"
1462418086.894639 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1hour:0" "0" "1"
1462418086.894648 [0 127.0.0.1:63620] "expire" "purchases:item1:1hour:0" "5184000"
1462418086.894656 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1day:0" "0" "1"

1462418086.894665 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1sec:0" "61" "1"
1462418086.894673 [0 127.0.0.1:63620] "expire" "purchases:item1:1sec:0" "7200"
1462418086.894681 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1min:0" "60" "1"
1462418086.894690 [0 127.0.0.1:63620] "expire" "purchases:item1:1min:0" "604800"
1462418086.894699 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1hour:0" "0" "1"
1462418086.894709 [0 127.0.0.1:63620] "expire" "purchases:item1:1hour:0" "5184000"
1462418086.894718 [0 127.0.0.1:63620] "hincrby" "purchases:item1:1day:0" "0" "1"
1462418086.894726 [0 127.0.0.1:63620] "multi"
1462418086.894759 [0 127.0.0.1:63620] "hget" "purchases:item1:1sec:0" "0"
1462418086.894768 [0 127.0.0.1:63620] "hget" "purchases:item1:1sec:0" "1"
1462418086.894774 [0 127.0.0.1:63620] "hget" "purchases:item1:1sec:0" "2"
1462418086.894780 [0 127.0.0.1:63620] "hget" "purchases:item1:1sec:0" "3"
1462418086.894786 [0 127.0.0.1:63620] "hget" "purchases:item1:1sec:0" "4"
1462418086.894793 [0 127.0.0.1:63620] "exec"
1462418086.894795 [0 127.0.0.1:63620] "multi"
1462418086.894800 [0 127.0.0.1:63620] "hget" "purchases:item1:1min:0" "0"
1462418086.894806 [0 127.0.0.1:63620] "hget" "purchases:item1:1min:0" "60"
1462418086.894812 [0 127.0.0.1:63620] "hget" "purchases:item1:1min:0" "120"
1462418086.894826 [0 127.0.0.1:63620] "exec"

可以看到，本例所有的秒級都存入了key: purchases:item1:1sec:0中，可以存300個時間點，然後具體的秒在field中表示。

Adding uniqueness with Sorted Sets and HyperLogLog

本節使用Sorted Sets 和 HyperLogLog實現統計唯一(distinct)訪問, 利用了這兩個資料型別沒有重複值的特性。Sorted Set 100%準確，而HyperLogLog使用較少記憶體，但準確率只有99.19%。具體使用哪個就看你的需要了。

這兩個例子和前兩個例子的區別是，前面的是計訪問總數，而這兩個只計唯一的訪問（同一使用者的重複訪問不會計入）

要讓Sorted Sets(zset)有效率的使用記憶體，需要一個zset中的field數量小於128，類似於前面Hash的例子，這裡也做了key grouping，相當於做了分割槽。

通過redis-cli monitor可以看到實際執行的命令:

[[email protected] chapter 3]$ node using-timeseries-unique.js sorted-set

[[email protected] ~]$ redis-cli monitor
OK
1462419249.646618 [0 127.0.0.1:63625] "info"
1462419249.650224 [0 127.0.0.1:63625] "flushall"
1462419249.689717 [0 127.0.0.1:63625] "zadd" "concurrentplays:1sec:0" "0" "0:user:max"
1462419249.690007 [0 127.0.0.1:63625] "expire" "concurrentplays:1sec:0" "7200"
1462419249.690033 [0 127.0.0.1:63625] "zadd" "concurrentplays:1min:0" "0" "0:user:max"
1462419249.690053 [0 127.0.0.1:63625] "expire" "concurrentplays:1min:0" "604800"
1462419249.690067 [0 127.0.0.1:63625] "zadd" "concurrentplays:1hour:0" "0" "0:user:max"
1462419249.690083 [0 127.0.0.1:63625] "expire" "concurrentplays:1hour:0" "5184000"
1462419249.690098 [0 127.0.0.1:63625] "zadd" "concurrentplays:1day:0" "0" "0:user:max"

1462419249.690114 [0 127.0.0.1:63625] "zadd" "concurrentplays:1sec:0" "0" "0:user:max"
1462419249.690130 [0 127.0.0.1:63625] "expire" "concurrentplays:1sec:0" "7200"
1462419249.690144 [0 127.0.0.1:63625] "zadd" "concurrentplays:1min:0" "0" "0:user:max"
1462419249.690159 [0 127.0.0.1:63625] "expire" "concurrentplays:1min:0" "604800"
1462419249.690173 [0 127.0.0.1:63625] "zadd" "concurrentplays:1hour:0" "0" "0:user:max"
1462419249.690187 [0 127.0.0.1:63625] "expire" "concurrentplays:1hour:0" "5184000"
1462419249.690201 [0 127.0.0.1:63625] "zadd" "concurrentplays:1day:0" "0" "0:user:max"

1462419249.690215 [0 127.0.0.1:63625] "zadd" "concurrentplays:1sec:0" "1" "1:user:hugo"
1462419249.690235 [0 127.0.0.1:63625] "expire" "concurrentplays:1sec:0" "7200"
1462419249.690248 [0 127.0.0.1:63625] "zadd" "concurrentplays:1min:0" "0" "0:user:hugo"
1462419249.690265 [0 127.0.0.1:63625] "expire" "concurrentplays:1min:0" "604800"
1462419249.690279 [0 127.0.0.1:63625] "zadd" "concurrentplays:1hour:0" "0" "0:user:hugo"
1462419249.690296 [0 127.0.0.1:63625] "expire" "concurrentplays:1hour:0" "5184000"
1462419249.690309 [0 127.0.0.1:63625] "zadd" "concurrentplays:1day:0" "0" "0:user:hugo"

1462419249.690325 [0 127.0.0.1:63625] "zadd" "concurrentplays:1sec:0" "1" "1:user:renata"
1462419249.690342 [0 127.0.0.1:63625] "expire" "concurrentplays:1sec:0" "7200"
1462419249.690382 [0 127.0.0.1:63625] "zadd" "concurrentplays:1min:0" "0" "0:user:renata"
1462419249.690401 [0 127.0.0.1:63625] "expire" "concurrentplays:1min:0" "604800"
1462419249.690415 [0 127.0.0.1:63625] "zadd" "concurrentplays:1hour:0" "0" "0:user:renata"
1462419249.690431 [0 127.0.0.1:63625] "expire" "concurrentplays:1hour:0" "5184000"
1462419249.690445 [0 127.0.0.1:63625] "zadd" "concurrentplays:1day:0" "0" "0:user:renata"

1462419249.690461 [0 127.0.0.1:63625] "zadd" "concurrentplays:1sec:0" "3" "3:user:hugo"
1462419249.690477 [0 127.0.0.1:63625] "expire" "concurrentplays:1sec:0" "7200"
1462419249.690490 [0 127.0.0.1:63625] "zadd" "concurrentplays:1min:0" "0" "0:user:hugo"
1462419249.690504 [0 127.0.0.1:63625] "expire" "concurrentplays:1min:0" "604800"
1462419249.690518 [0 127.0.0.1:63625] "zadd" "concurrentplays:1hour:0" "0" "0:user:hugo"
1462419249.690532 [0 127.0.0.1:63625] "expire" "concurrentplays:1hour:0" "5184000"
1462419249.690545 [0 127.0.0.1:63625] "zadd" "concurrentplays:1day:0" "0" "0:user:hugo"

1462419249.690559 [0 127.0.0.1:63625] "zadd" "concurrentplays:1sec:0" "61" "61:user:kc"
1462419249.690576 [0 127.0.0.1:63625] "expire" "concurrentplays:1sec:0" "7200"
1462419249.690589 [0 127.0.0.1:63625] "zadd" "concurrentplays:1min:0" "60" "60:user:kc"
1462419249.690605 [0 127.0.0.1:63625] "expire" "concurrentplays:1min:0" "604800"
1462419249.690618 [0 127.0.0.1:63625] "zadd" "concurrentplays:1hour:0" "0" "0:user:kc"
1462419249.690634 [0 127.0.0.1:63625] "expire" "concurrentplays:1hour:0" "5184000"
1462419249.690648 [0 127.0.0.1:63625] "zadd" "concurrentplays:1day:0" "0" "0:user:kc"
1462419249.690669 [0 127.0.0.1:63625] "multi"
1462419249.690695 [0 127.0.0.1:63625] "zcount" "concurrentplays:1sec:0" "0" "0"
1462419249.690842 [0 127.0.0.1:63625] "zcount" "concurrentplays:1sec:0" "1" "1"
1462419249.690860 [0 127.0.0.1:63625] "zcount" "concurrentplays:1sec:0" "2" "2"
1462419249.690872 [0 127.0.0.1:63625] "zcount" "concurrentplays:1sec:0" "3" "3"
1462419249.690883 [0 127.0.0.1:63625] "zcount" "concurrentplays:1sec:0" "4" "4"
1462419249.690897 [0 127.0.0.1:63625] "exec"
1462419249.690902 [0 127.0.0.1:63625] "multi"
1462419249.690911 [0 127.0.0.1:63625] "zcount" "concurrentplays:1min:0" "0" "0"
1462419249.690923 [0 127.0.0.1:63625] "zcount" "concurrentplays:1min:0" "60" "60"
1462419249.690936 [0 127.0.0.1:63625] "zcount" "concurrentplays:1min:0" "120" "120"
1462419249.690949 [0 127.0.0.1:63625] "exec"

可以看到，key的設計和前面的例子是一樣的，score中存的是時間，而不是累計的訪問次數，field中存的使用者名稱，但不明白為何要加一個時間字首。
幾個命令解釋一下，zcount用來計數，multi和exec用於交易

127.0.0.1:6379> help zcount

  ZCOUNT key min max
  summary: Count the members in a sorted set with scores within the given values
  since: 2.0.0
  group: sorted_set

127.0.0.1:6379> help multi

  MULTI -
  summary: Mark the start of a transaction block
  since: 1.2.0
  group: transactions

127.0.0.1:6379> help exec

  EXEC -
  summary: Execute all commands issued after MULTI
  since: 1.2.0
  group: transactions

HyperLogLog 的實現無需key grouping, 類似於String, 實現非常簡潔！每一個timestamp一個key。並用PFADD代替了INCRBY, 以及PFCOUNT 替代了 MGET.

通過redis-cli monitor可以看到實際執行的命令:

[[email protected] chapter 3]$ node using-timeseries-unique.js hyperloglog

[[email protected] ~]$ redis-cli monitor
OK
1462421278.494144 [0 127.0.0.1:63631] "info"
1462421278.498276 [0 127.0.0.1:63631] "flushall"
1462421278.556209 [0 127.0.0.1:63631] "pfadd" "concurrentplays:1sec:0" "user:max"
1462421278.556277 [0 127.0.0.1:63631] "expire" "concurrentplays:1sec:0" "7200"
1462421278.556300 [0 127.0.0.1:63631] "pfadd" "concurrentplays:1min:0" "user:max"
1462421278.556315 [0 127.0.0.1:63631] "expire" "concurrentplays:1min:0" "604800"
1462421278.556330 [0 127.0.0.1:63631] "pfadd" "concurrentplays:1hour:0" "user:max"
1462421278.556344 [0 127.0.0.1:63631] "expire" "concurrentplays:1hour:0" "5184000"
1462421278.556358 [0 127.0.0.1:63631] "pfadd" "concurrentplays:1day:0" "user:max"

1462421278.556371 [0 127.0.0.1:63631] "pfadd" "concurrentplays:1sec:0" "user:max"
1462421278.556386 [0 127.0.0.1:63631] "expire" "concurrentplays:1sec:0" "7200"
1462421278.556399 [0 127.0.0.1:63631] "pfadd" "concurrentplays:1min:0" "user:max"
1462421278.556412 [0 127.0.0.1:63631] "expire" "concurrentplays:1min:0" "604800"
1462421278.556427 [0 127.0.0.1:63631] "pfadd" "concurrentplays:1hour:0" "user:max"
1462421278.556440 [0 127.0.0.1:63631] "expire" "concurrentplays:1hour:0" "5184000"
1462421278.556454 [0 127.0.0.1:63631] "pfadd" "concurrentplays:1day:0" "user:max"

1462421278.556466 [0 127.0.0.1:63631] "pfadd" "concurrentplays:1sec:1" "user:hugo"
1462421278.556481 [0 127.0.0.1:63631] "expire" "concurrentplays:1sec:1" "7200"
1462421278.556495 [0 127.0.0.1:63631] "pfadd" "concurrentplays:1min:0" "user:hugo"
1462421278.556509 [0 127.0.0.1:63631] "expire" "concurrentplays:1min:0" "604800"
1462421278.556523 [0 127.0.0.1:63631] "pfadd" "concurrentplays:1hour:0" "user:hugo"
1462421278.556537 [0 127.0.0.1:63631] "expire" "concurrentplays:1hour:0" "5184000"
1462421278.556550 [0 127.0.0.1:63631] "pfadd<

相關推薦

第三章：Time Series (A Collection of Observations)

Chapter 3. Time Series (A Collection of Observations) 時間序列（Time Series）是在一段時間範圍內排序的時間點的值，可用於統計，社交網路，電信工程等一切需要時間度量的場景。它可以來進行預測，如股市走

讀構建之法 第三章：軟件工程師的成長

知識點 可維護 vid -s 評估 不同 fun 可靠 科研 本章理論和知識點：評價軟件工程師水平的主要方法 軟件工程把相關的技術和過程統一到一個體系中，叫“軟件開發流程”，軟件開發流程的目的是為了提高軟件開發、運營、維護的效率，以及提升用戶滿意度、軟件的可靠性和可維護性。

第三章：基本概念

npr bject 字面量 enume 不能 數據 開頭 clas int() 語法 借鑒了C語言以及類C語言（如java,perl）的語法 區分大小寫 標記符 字母數字下劃線$ 開頭不為數字 嚴格模式 use strict; 語句以分號結尾 關鍵

第三章：使用listview控件展示數據

復數 ted ima 深度 lar row listview items uil 第三章：使用listview控件展示數據 1.圖像列表控件Imagelist Images 儲存的所有圖片的復數 ImageSize 圖片的大小 transparentColor 被視

《Python基礎教程》第三章：使用字符串

方法的參數 基礎 指定 方法 添加 repl pre replace this find方法可以在一個較長的字符串中查找子字符串。它返回子串所在位置的最左端索引。如果沒有找到則返回-1 join方法用來在隊列中添加元素，需要添加的隊列元素都必須是字符串 >>

第三章：TestOps方法論

測試數據 bug 重要 性能 開發 良好的 image 工具 ued 某些質量目標只能在TestOps中可行，這就是為什麽TestOps成功的一個重要因素，也這是一個開發團隊和測試團隊與Ops具有相同目標的範例。 TestOps需要一個全面的方

《數據庫設計入門經典》讀書筆記——第三章：工作場所中的數據庫建模

中間 特定 理論 大學 並且 外鍵 另一個 必須 所有 規範化用於粒度化和組織在數據庫中使用的數據。 在第4章中將詳細介紹規範化和應用範式的過程。在這個階段只需要知道規範化是用於將數據劃分到單獨表中的方法或公式——根據一組規則。 不信任將視圖用於除了安全性目標之外的任何事情

Node入門教程(4)第三章：第一個 Nodejs 程序

tps con javascrip 第三章 body linux 一定的 ava UC 第一個 Nodejs 程序 本教程僅適合您已經有一定的JS編程的基礎或者是後端語言開發的基礎。如果您是零基礎，建議您先學一下老馬的前端免費視頻教程 第一步：創建項目文件夾 首先創建

第三章：軟件也要拼臉蛋——UI 開發的點點滴滴

layout 開發 ssd 幀布局 edittext PE ogre 點點滴滴 tex 常用控件 常用控件有：按鈕 Button、文本顯示框 TextView、圖片顯示框 ImageView、文本編輯框 EditText、進度條 ProgressBar、提示框 Alert

（轉）Django學習之 第三章：動態Web頁面基礎

只有一個 即使 typeerror 方法 對象傳遞 power int() 擔心 意圖 上一章我們解釋了怎樣開始一個Django項目和運行Django服務器 當然了，這個站點實際上什麽也沒有做------除了顯示了"It worked"這條信息以外。 這一章我們介紹怎樣使用

現代軟件工程—構建之法---第三章：練習與討論

員工 軟件行業 別人 經典 可能 能力 現在 必備 似的 1.選哪一種醫生？ （1）.如果是我的話，我會選擇C類型的醫生。因為c類型的醫生看著比較靠譜，首先他的從業經驗比較豐富，遇見過很多類似的病歷，對病情包括手術比較有把握，對患者可能會比較了解；其次就是他可以一邊手術一邊

加密與解密第三章：IDA的基本操作

substr line for src 用法 標準庫 set poi 調用 更改數據為結構體和枚舉類型 查看輸入和輸出的idc腳本 //Imports.idc ?г??????????????? //(c) www.PEDIY.com 2000-2008 #includ

第三章：模板擴展

gif 但是 pro 第三章 沖突 totally self 訪問網站 tel 在第二章中，我們看到了Tornado模板系統如何簡單地傳遞信息給網頁，使你在插入動態數據時保持網頁標記的整潔。然而，大多數站點希望復用像header、footer和布局網格這樣的內容。在這一章中

第三章：數組和字符串

ans can 輸入 是否 讀取字符串 下一步 字符數組 lag getc 2018-10-21-18:37:41 隨學筆記 小計 <1>：比較大的數組一般定義在main函數外面，否則程序很容易出錯。 <2>：memset 函數： 　　原型：void

第三章：如何建模服務

問題 快的 left 上下文 結構 保持 減少 都是 部分 什麽好的服務？ 松耦合 一個松耦合的服務應該盡可能的少知道與之協作的那些服務的信息。 如果做到了服務之間的松耦合，那麽修改一個服務就不需要修改另一個服務。 使用微服務的特定就是可以獨立的修改和部署單個服務而不需要修

Introduction to 3D Game Programming with DirectX 12 學習筆記之 --- 第三章：變換

學習目標 理解如何用矩陣表示線性變換和仿射變換； 學習在座標系中縮放，旋轉和移動幾何體； 學習利用矩陣的乘法合併幾個變換矩陣； 學習如何在座標系之間轉換，並且表示為轉換矩陣；斜體樣式 學習如何利用DirectX Math庫提供的方法構造轉換矩陣。

玩轉資料結構——第三章：最基礎的動態資料結構：連結串列

內容概括： 3-1.什麼是連結串列 3-2.在連結串列中新增元素 3-3.使用連結串列的虛擬頭結點 3-4.連結串列的遍歷，查詢和修改 3-5.從連結串列中刪除元素 3-6.使用連結串列實現棧 3-7.帶有尾指標的連結串列：使用連結串列實現佇列

《SQL入門經典》筆記（第三章：建立資料庫之管理資料庫物件）

1. 什麼是資料庫物件？ 資料庫物件用於儲存或引用資料，需要被定義，例如表、試圖、簇、序列、索引和異名（讓表具有另一個名稱）。   2. 什麼是規劃？ 規劃是與資料庫某個使用者名稱相關聯的資料庫物件集合。相應的使用者名稱被稱為“規劃所有人”/“關聯物件組的所有人”。 同

《惡意程式碼分析實戰》--第三章：動態基礎分析

一、虛擬網路環境配置 配置環境：伺服器端kali2.0 客戶端win7 1、配置inetsim kali自帶inetsim，因此只需配置就可以了（但是好像不配置也是可以的……） 配置連結：http://www.cnblogs.com/hyq20135317/p/5515675.h

演算法導論 第三章：函式的增長 筆記（Θ記號、O記號、Ω記號、o記號、ω記號、漸近記號的性質、標準記號與常用函式）

Θ記號： 該記號圓圈中是個M。Θ記號漸近地給出一個函式的上界和下界。 對於一個給定的函式g(n)，我們用Θ(g(n))來表示以下函式的集合： Θ(g(n))={f(n)：存在正常量c1、c2和n0，使得對於所有n⩾n0，有0⩽c1g(n)⩽f(n)⩽c2g(n)}。 即若存在正常