Table Pattern: Rotating Log / Ring Buffer


原文  http://kejser.org/table-pattern-rotating-log-ring-buffer/ 添加到推刊

Most database systems need some form of log table to keep track of events, for example for auditing purposes. To avoid the log growing forever, it is often a good idea to regularly rotate old log entries out of this table. For small log tables, running a DELETE statement works well for this purpose. However, as the log throughput grows, it is often preferable to use partition switching instead. In this blog, I will show you an implementation of a rotating log table.

The Table Structure

A rotating log is essentially a ring buffer where the writes happen at one part of the buffer and the old records are removed at the other end. For the implementation example, we will assume you wish to rotate the log daily, removing and aggregating old entries and making sure that the new entries are added to a place that is easy to rotate in the future.

It is possible to traditional partition “sliding window” techniques and add new partitions daily to achieve this. However, I find the splitting, merging and switching a bit clunky for this purpose. Instead, I prefer to implement a rotating log like this:

Transact-SQL
CREATE PARTITION FUNCTION PF_RingBuffer (TINYINT)
AS RANGE RIGHT FOR VALUES (0,1,2,3,4,5,6,7,8,9,10,11,12)

CREATE PARTITION SCHEME PS_RingBuffer 
AS PARTITION PF_RingBuffer ALL TO ([DATA])

CREATE TABLE RingBuffer
(
  EventTime DATETIME NOT NULL DEFAULT GETUTCDATE()
  , Attribute VARCHAR(50) NOT NULL 
  , Value BIGINT NOT NULL
  , Offset AS CAST(DATEDIFF(day, 0, EventTime) % 14 AS TINYINT) PERSISTED
) ON PS_RingBuffer (Offset)

CREATE PARTITION FUNCTION PF_RingBuffer ( TINYINT )

AS RANGE RIGHT FOR VALUES ( 0 , 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 )

CREATE PARTITION SCHEME PS_RingBuffer

AS PARTITION PF_RingBuffer ALL TO ( [ DATA ] )

CREATE TABLE RingBuffer

(

   EventTime DATETIME NOT NULL DEFAULT GETUTCDATE ( )

   , Attribute VARCHAR ( 50 ) NOT NULL

   , Value BIGINT NOT NULL

   , Offset AS CAST ( DATEDIFF ( day , 0 , EventTime ) % 14 AS TINYINT ) PERSISTED

) ON PS_RingBuffer ( Offset )

With this design, I never have to alter the partition function or scheme, I can simply switch out the old parts of the ring buffer.

For the purposes of this implementation, let me add a few more tables:

Transact-SQL
CREATE TABLE RingBuffer_SwitchTarget
(
  EventTime DATETIME NOT NULL DEFAULT GETUTCDATE()
  , Attribute VARCHAR(50) NOT NULL 
  , Value BIGINT NOT NULL
  , Offset AS CAST(DATEDIFF(day, 0, EventTime) % 14 AS TINYINT) PERSISTED
) ON PS_RingBuffer (Offset)

CREATE TABLE RingBuffer_Aggregate
(
  EventTime DATE NOT NULL
  , Attribute VARCHAR(50) NOT NULL
  , Value BIGINT NOT NULL
) ON [DATA]
ALTER TABLE RingBuffer_Aggregate ADD CONSTRAINT PK_RingBuffer_Aggregate
PRIMARY KEY (EventTime, Attribute, Value)

CREATE TABLE RingBuffer_SwitchTarget

(

   EventTime DATETIME NOT NULL DEFAULT GETUTCDATE ( )

   , Attribute VARCHAR ( 50 ) NOT NULL

   , Value BIGINT NOT NULL

   , Offset AS CAST ( DATEDIFF ( day , 0 , EventTime ) % 14 AS TINYINT ) PERSISTED

) ON PS_RingBuffer ( Offset )

CREATE TABLE RingBuffer_Aggregate

(

   EventTime DATE NOT NULL

   , Attribute VARCHAR ( 50 ) NOT NULL

   , Value BIGINT NOT NULL

) ON [ DATA ]

ALTER TABLE RingBuffer_Aggregate ADD CONSTRAINT PK_RingBuffer_Aggregate

PRIMARY KEY ( EventTime , Attribute , Value )

The  RingBuffer_SwitchTarget  will be used to switch out the old entries in RingBuffer . The RingBuffer_Aggregate is used to keep an aggregate of old data, which of course will be much smaller.

To illustrate, here is the flow we are implementing:

Ring Buffer Implementation

Rotating the Log

With the table structures in place, we can now write the procedure that rotates the log. Using a few simple math tricks, we can rotate and the log with a simple piece of code:

Transact-SQL
CREATE PROCEDURE RingbufferRotate
  @Now DATETIME = NULL
  , @PartitionsToKeep INT = 7
AS
SET NOCOUNT ON
BEGIN
  IF @Now IS NULL SET @Now = GETUTCDATE()

  DECLARE @BufferSize INT
  SELECT @BufferSize = COUNT(*)
  FROM sys.partitions P 
  JOIN sys.tables T
    ON P.object_id = T.object_id 
  WHERE T.name = 'RingBuffer' 

  IF @PartitionsToKeep > @BufferSize BEGIN
    RAISERROR ('Can''t keep more partitions than the current buffer size of: %i', 16, 1, @BufferSize)
    RETURN
  END

  DECLARE @CurrentPartition INT
  SET @CurrentPartition = DATEDIFF(day, 0, @Now) % @BufferSize

  DECLARE @OldestPartitionToKeep INT
  SET @OldestPartitionToKeep 
    = (@CurrentPartition + @BufferSize - @PartitionsToKeep) % @BufferSize

  /* Start from the next partition up from current and move forward */
  DECLARE @P INT = (@CurrentPartition + 1) % @BufferSize
  DECLARE @I INT = 0
  WHILE @I < @BufferSize - @PartitionsToKeep BEGIN 
    DECLARE @Sql NVARCHAR(4000)
      = 'ALTER TABLE RingBuffer SWITCH PARTITION <p> TO RingBuffer_SwitchTarget PARTITION <p>'
    SET @Sql = REPLACE(@Sql, '<p>', CAST(@P + 1 AS NVARCHAR)) /* Partition are 1 offset */

    EXEC sp_executesql @Sql

    SET @I = @I + 1;
    SET @P = (@P + 1) % @BufferSize
  END
    
  MERGE INTO RingBuffer_Aggregate AS T
  USING (SELECT CAST(EventTime AS DATE) AS EventTime
           , Attribute
           , SUM(Value) AS Value
         FROM RingBuffer_SwitchTarget
         GROUP BY CAST(EventTime AS DATE), Attribute) AS A
    ON A.EventTime = T.EventTime
    AND A.Attribute = T.Attribute
    WHEN MATCHED THEN
      UPDATE SET T.Value = T.Value + A.Value
    WHEN NOT MATCHED THEN
      INSERT (EventTime, Attribute, Value)
      VALUES (A.EventTime, A.Attribute, A.Value);

    TRUNCATE TABLE RingBuffer_SwitchTarget;
END

CREATE PROCEDURE RingbufferRotate

   @ Now DATETIME = NULL

   , @ PartitionsToKeep INT = 7

AS
SET NOCOUNT ON
BEGIN

   IF @ Now IS NULL SET @ Now = GETUTCDATE ( )

   DECLARE @ BufferSize INT

   SELECT @ BufferSize = COUNT ( * )

   FROM sys . partitions P

   JOIN sys . tables T

     ON P . object_id = T . object_id

   WHERE T . name = 'RingBuffer'

   IF @ PartitionsToKeep > @ BufferSize BEGIN

     RAISERROR ( 'Can' 't keep more partitions than the current buffer size of: %i' , 16 , 1 , @ BufferSize )

     RETURN
   END

   DECLARE @ CurrentPartition INT

   SET @ CurrentPartition = DATEDIFF ( day , 0 , @ Now ) % @ BufferSize

   DECLARE @ OldestPartitionToKeep INT

   SET @ OldestPartitionToKeep

     = ( @ CurrentPartition + @ BufferSize - @ PartitionsToKeep ) % @ BufferSize

   /* Start from the next partition up from current and move forward */

   DECLARE @ P INT = ( @ CurrentPartition + 1 ) % @ BufferSize

   DECLARE @ I INT = 0

   WHILE @ I < @ BufferSize - @ PartitionsToKeep BEGIN

     DECLARE @ Sql NVARCHAR ( 4000 )

       = 'ALTER TABLE RingBuffer SWITCH PARTITION <p> TO RingBuffer_SwitchTarget PARTITION <p>'

     SET @ Sql = REPLACE ( @ Sql , '<p>' , CAST ( @ P + 1 AS NVARCHAR ) ) /* Partition are 1 offset */

     EXEC sp_executesql @ Sql

     SET @ I = @ I + 1 ;

     SET @ P = ( @ P + 1 ) % @ BufferSize

   END
    

   MERGE INTO RingBuffer_Aggregate AS T

   USING ( SELECT CAST ( EventTime AS DATE ) AS EventTime

           , Attribute

           , SUM ( Value ) AS Value

         FROM RingBuffer_SwitchTarget

         GROUP BY CAST ( EventTime AS DATE ) , Attribute ) AS A

     ON A . EventTime = T . EventTime

     AND A . Attribute = T . Attribute

     WHEN MATCHED THEN

       UPDATE SET T . Value = T . Value + A . Value

     WHEN NOT MATCHED THEN

       INSERT ( EventTime , Attribute , Value )

       VALUES ( A . EventTime , A . Attribute , A . Value ) ;

     TRUNCATE TABLE RingBuffer_SwitchTarget ;

END

And that’s really all there is to it.

Summary

In this blog, I have shown you how to implement a rotating log without the need for a lot of partition function management. Using a ring buffer of dates, a full day of data can be rotated around with a single SWITCH statement.

The ring buffer mechanism can also be used for heaps or tables with only a cluster index that are so large that online index rebuilds don’t fit in memory and tempdb . Instead of rebuilding the entire table, each entry in the ring buffer can be rebuilt, one at a time, in a rotating manner. It is trivial to modify the above stored procedure to implement this functionality.

# 高校智慧校园解决方案摘要 智慧校园解决方案是针对高校信息化建设的核心工程,旨在通过物联网技术实现数字化校园的智能化升级。该方案通过融合计算机技术、网络通信技术、数据库技术和IC卡识别技术,初步实现了校园一卡通系统,进而通过人脸识别技术实现了更精准的校园安全管理、生活管理、教务管理和资源管理。 方案包括多个管理系统:智慧校园管理平台、一卡通卡务管理系统、一卡通人脸库管理平台、智能人脸识别消费管理系统、疫情防控管理系统、人脸识别无感识别管理系统、会议签到管理系统、人脸识别通道管理系统和图书馆对接管理系统。这些系统共同构成了智慧校园的信息化基础,通过统一数据库和操作平台,实现了数据共享和信息一致性。 智能人脸识别消费管理系统通过人脸识别终端,在无需接触的情况下快速完成消费支付过程,提升了校园服务效率。疫情防控管理系统利用热成像测温技术、视频智能分析等手段,实现了对校园人员体温监测和疫情信息实时上报,提高了校园公共卫生事件的预防和控制能力。 会议签到管理系统和人脸识别通道管理系统均基于人脸识别技术,实现了会议的快速签到和图书馆等场所的高效通行管理。与图书馆对接管理系统实现了一卡通系统与图书馆管理系统的无缝集成,提升了图书借阅的便捷性。 总体而言,该智慧校园解决方案通过集成的信息化管理系统,提升了校园管理的智能化水平,优化了校园生活体验,增强了校园安全,并提高了教学和科研的效率。
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值