1 00:00:00,150 --> 00:00:04,320 ‫Let's talk about a little bit Synchronoss versus asynchronous application, which is the same, the 2 00:00:04,320 --> 00:00:09,750 ‫problem we originally raised in the first and the presentation. 3 00:00:09,750 --> 00:00:19,200 ‫So Synchronoss application is when you issue a right to the red master replica master node, the client 4 00:00:19,200 --> 00:00:24,300 ‫has to wait until that transaction is returned to the master, obviously. 5 00:00:24,300 --> 00:00:24,600 ‫Right. 6 00:00:24,720 --> 00:00:29,730 ‫Because if you commit a transaction, you have to wait for the authorization to commit in the master 7 00:00:29,730 --> 00:00:30,900 ‫or in any database. 8 00:00:30,900 --> 00:00:31,500 ‫You have to wait. 9 00:00:31,510 --> 00:00:31,730 ‫Right. 10 00:00:32,040 --> 00:00:41,550 ‫Once that's done, synchronous replication also forces you to commit to that child replicas or the standby 11 00:00:41,550 --> 00:00:45,660 ‫replicas before the client can be unblocked. 12 00:00:46,050 --> 00:00:48,810 ‫Cassandra does something similar like this called Khorram. 13 00:00:49,320 --> 00:00:56,430 ‫And now you get to a conversion like, OK, Hosain, I have seven standby replicas on my master note. 14 00:00:56,700 --> 00:00:59,360 ‫Which ones should I should I wait for the seven of them? 15 00:00:59,850 --> 00:01:02,190 ‫That's a little bit of an overkill, if you think about it. 16 00:01:02,190 --> 00:01:02,410 ‫Right. 17 00:01:02,820 --> 00:01:09,120 ‫That's why Oscars or other database gives you these options, like, OK, wait for the first two or 18 00:01:09,120 --> 00:01:18,330 ‫wait for the first one or whenever any of them completes unblock immediately or whatever two of them 19 00:01:18,330 --> 00:01:21,810 ‫or any of them or like, for example, these are more important than those. 20 00:01:22,710 --> 00:01:28,470 ‫Those two is more important than those three standby replicas like you have done on a one standby replica 21 00:01:28,740 --> 00:01:33,450 ‫is a backup and the other are actually active people actually acquiring organs. 22 00:01:33,750 --> 00:01:38,100 ‫So you can have the flexibility here so you can control this stuff. 23 00:01:38,100 --> 00:01:46,950 ‫You can say, no, I want this as if I am actually want this everything to be fully consistent. 24 00:01:46,950 --> 00:01:52,320 ‫So you're going to take the hit, obviously, but you get full consistency, not evangel consistency, 25 00:01:52,590 --> 00:01:59,520 ‫because once you write, the transaction cannot fully commit until everything is committed in the master 26 00:01:59,520 --> 00:02:04,860 ‫and all the note, if you just turn everything on, you are essentially in a single cluster almost. 27 00:02:05,040 --> 00:02:06,900 ‫By the way, this whole thing is called the cluster. 28 00:02:07,620 --> 00:02:14,700 ‫Asynchronous replication is what by default, some of the default is like, OK, once you write to my 29 00:02:14,700 --> 00:02:18,180 ‫master, I'm done, the client is done. 30 00:02:18,180 --> 00:02:20,400 ‫You just can't unlock to the client. 31 00:02:20,400 --> 00:02:21,000 ‫This is done. 32 00:02:21,480 --> 00:02:22,840 ‫But the data is now. 33 00:02:22,850 --> 00:02:26,160 ‫It now has to do more work on a background. 34 00:02:27,230 --> 00:02:31,950 ‫Process that writes periodically to those standby replicas. 35 00:02:32,390 --> 00:02:33,590 ‫This, obviously. 36 00:02:35,720 --> 00:02:41,780 ‫All of a sudden now you have faster throughput, I have faster right, but at the back end, I have 37 00:02:41,780 --> 00:02:49,370 ‫a little bit some puppet processes that are doing this, asynchronous jobs and that could eat off the 38 00:02:49,370 --> 00:02:51,650 ‫CPU and a lot of work. 39 00:02:51,660 --> 00:02:51,900 ‫Right. 40 00:02:52,220 --> 00:02:57,290 ‫If you do a lot of these and it could scale and could slow things down, you could definitely get to 41 00:02:57,300 --> 00:02:58,730 ‫get eventually consistency. 42 00:02:59,000 --> 00:03:01,790 ‫If you can't live with that, you know, fine. 43 00:03:01,790 --> 00:03:06,610 ‫But as something people prefer synchronous now over asynchronous forces for certain.