1
00:00:00,100 --> 00:00:01,440
‫So, a quick lecture on something

2
00:00:01,440 --> 00:00:03,230
‫called DynamoDB Write Sharding.

3
00:00:03,230 --> 00:00:04,350
‫So, let's assume this use case

4
00:00:04,350 --> 00:00:05,900
‫where we have a voting application,

5
00:00:05,900 --> 00:00:08,450
‫and voting application people can vote

6
00:00:08,450 --> 00:00:10,300
‫for candidate A and candidate B.

7
00:00:10,300 --> 00:00:11,600
‫So, two candidates.

8
00:00:11,600 --> 00:00:14,730
‫Now, if we used the partition key to be candidate ID,

9
00:00:14,730 --> 00:00:16,270
‫the problem is that all the results

10
00:00:16,270 --> 00:00:18,230
‫will go into two partitions only

11
00:00:18,230 --> 00:00:20,320
‫because, well, the data is going to be partitioned

12
00:00:20,320 --> 00:00:22,980
‫by either candidate A or candidate B,

13
00:00:22,980 --> 00:00:25,240
‫which will generate issues for writes and for reads

14
00:00:25,240 --> 00:00:27,620
‫so we're gonna get hot partition issues.

15
00:00:27,620 --> 00:00:29,950
‫So, how do we solve this issue?

16
00:00:29,950 --> 00:00:33,440
‫Well, a strategy is to distribute the candidate ID

17
00:00:33,440 --> 00:00:35,870
‫better across partitions and to do so,

18
00:00:35,870 --> 00:00:37,330
‫we're going to add a suffix,

19
00:00:37,330 --> 00:00:39,920
‫or a prefix to the partition key value.

20
00:00:39,920 --> 00:00:41,400
‫So, the idea is that we're now we're going to get

21
00:00:41,400 --> 00:00:43,190
‫the candidate ID, and then we'll add,

22
00:00:43,190 --> 00:00:44,290
‫for example, the number 11.

23
00:00:44,290 --> 00:00:46,920
‫So candidate a 11, and add a row,

24
00:00:46,920 --> 00:00:49,010
‫and then we're going to have candidate b 17,

25
00:00:49,010 --> 00:00:51,540
‫candidate b 18, candidate a 20,

26
00:00:51,540 --> 00:00:53,360
‫and the idea is that now the partition key,

27
00:00:53,360 --> 00:00:54,460
‫is a lot more distributed,

28
00:00:54,460 --> 00:00:56,570
‫because it takes more unique values

29
00:00:56,570 --> 00:00:59,820
‫and therefore your data set is going to be fully written

30
00:00:59,820 --> 00:01:02,410
‫and fully readable from your Dynamo DB table.

31
00:01:02,410 --> 00:01:05,530
‫Now, two methods for creating this suffix or this prefix.

32
00:01:05,530 --> 00:01:07,560
‫You can use a random suffix

33
00:01:07,560 --> 00:01:10,860
‫or you can calculate it using a hashing algorithm.

34
00:01:10,860 --> 00:01:11,780
‫Both of these things work.

35
00:01:11,780 --> 00:01:13,620
‫The idea is that you want to get a very,

36
00:01:13,620 --> 00:01:15,200
‫very distributed partition key.

37
00:01:15,200 --> 00:01:16,480
‫That's all, it's just a theory lecture,

38
00:01:16,480 --> 00:01:17,930
‫but you know this strategy now.

39
00:01:17,930 --> 00:01:20,683
‫I hope you liked it and I will see you in the next lecture.