Multi master test of AWS Aurora database

Time:2019-12-18
AWS Aurora MySQL finally launched multi master, facing the hard Oracle RAC. In addition to the excitement of database product selection, we also look at the features (including advantages and disadvantages) of the new multi master.
 
1. Aurora multi master currently only has the following regions available
US East (N. Virginia), US East (Ohio), US West (Oregon), and EU (Ireland)
 
2. Let’s take a picture first. It can be seen that the implementation mechanism of Oracle RAC is completely different from that of Oracle RAC. Oracle RAC is block cache fusion at the memory level, Aurora mm is storage replication, and it is basically staying at the theoretical level of Oracle 9i. There is no further comparison.
This paper mainly compares Aurora multi master and single master
For more installation / testing, refer to:
https://amazonaws-china.com/cn/blogs/database/building-highly-available-mysql-applications-using-amazon-aurora-mmsr/?from=timeline&isappinstalled=0
 
3. A unified endpoint can be accessed, and aurora can automatically load balance internally

 

4. You can also create customized endpoint and access only one instance
 
5. Multi master cannot create replica. There are only two master nodes in the whole cluster.
 

If you want to try to add nodes, you will encounter the following prompt.

 

6. The single master can add 15 replicas, cross region replicas and replica auto scaling, etc

 

7. Currently, multi master is only available in mysql-5.6.10a

 

8. In fact, there are not many optional versions of single master

 

9. There are only three types of machines available for multi master. There are neither very small nor very large models.

 
10. The optional single master models will be rich

 

 
 
In summary, MySQL’s open source gene also determines that there is a big gap with Oracle in many advanced functions. After all, they are all products of Oracle company. One is very expensive and the other is free.
However, with the secondary development of many large companies, the database products based on MySQL in many aspects also show a lot of concerns.
 
Aurora multi master, as a new product, estimates that considering the data consistency problem of dual masters, slave is not supported at present. For a multi master cluster, there are currently at most two nodes.
Aurora single master can have 15 slaves. In the query scenario with large data volume and high concurrency, the advantages can be seen at a glance. Since Aurora’s storage is shared, when master fails, slave takes over quickly. The main delay is in storing the synchronized gap. In theory, it’s a second level failover.
 
Therefore, in the process of use, database availability tolerance is required according to the business situation. To decide whether to use multi master or single master
For financial business with strong consistency, and for this kind of traditional business, the data volume will not be large. The multi master of two nodes can not only ensure strong consistency, but also meet the performance requirements.
If there is a system with a certain tolerance for downtime, single master is a better choice. After all, master doesn’t often hang up when it’s OK.
 
Moreover, the more beautiful and technical functions are, the more time is needed to settle and stabilize.
So Aurora multi master, on the way to the future, whether the two masters can coexist peacefully or not is a very important point.

Recommended Today

MySQL partition

MySQL partition Advantages of partitioning Store more data To optimize the query, only scan one or more necessary partitions. For count () and sum (), only summarize the partition statistics For data that is out of date or does not need to be saved, partitioning is faster Data queries are distributed across multiple disks to […]