Database Penalty Strategy

Question

For the online market under construction, I was faced with a situation requiring the implementation of a solution to fine the database. I am new to sharding, and after reading the posts on this forum, I believe that a template-based strategy using business units would be appropriate. But I still do not understand how best to use the methods of denormalization and data synchronization with such a stunning solution. There will be 3 main objects, supplier, customer and order. I plan to outline the database based on the vendor ID, since most of the processing of order data will be done by the vendor administrators. This will ensure that orders for the supplier are received from a single db instance, excluding cross db samples. However, in this case, when customers view their order information,the data will be in multiple db instances and will require a large database. What is usually done when such scenarios appear in a closed solution.

Answer 1

I think there is a 99.9% chance that you don’t need to scald.

You need to plaster if:

• Your database insert / update speed is close or exceeds the performance of the highest specification server, which you can save on buying AND
• You already process most of the read requests, reports, backups, etc. for replicated read-only slave data.
• You performed functional partitioning to transfer any non-essential or unrelated heavy workloads from your primary server.

Unless you can definitely say yes to all three of the above, you do not need to cheat.

Read

http://www.mysqlperformanceblog.com/2009/08/06/why-you-dont-want-to-shard/

Answer 2

, , TB. , , , / , , MySQL, .

.

• ( "Go Fish" ). . , .

, , , - , .

, , , .

Answer 3

You can also try nosql DB like mongodb or Cassandra

You can also use memcache to cache data for quick access.

You can also view replication of the master slave with several slaves.