1, RAC (or OPS in Oracle 8i)
Rac is a myth that has been a long time, because RAC has high requirements for hardware, especially in the 9i version, almost all running in the small machine, and there are many people who want RAC to become a stable and fast synonym. In fact, is this? In fact, we should first look at what the RAC's design is designed: in order to solve high availability and high concurrency, it is not an overall performance. To a certain extent, RAC is slow than a single node because it is more communication and management than a single node. However, it is precisely because of its high availability and a large concurrent connection, the RAC application is so wide, and the average distribution load can be made on a machine comparison machine, and other nodes can be made when a node is faulty. take over. However, the RAC brought about the complex management, the performance is reduced (this is especially obvious in the OPS OLTP system, and there are many in RAC).
2, partition
Sometimes, someone will ask me: "I put my big table partition, how is the speed of the application slow, I think it should be fast?", Will I ask his index Is the local index? Is most a partition keyword as a query condition? Answer is often: "Right, for the convenience of management, we use the partition index, we have only a small part of the query is partitioned, most of it." Yes, here is the reason, the partition is not a fast = true parameter, in fact, there is no this parameter at all. Let's take a look at the purpose of the partition, for the convenience of large-scale object management such as DSS. The two main two effects are high management and scalability, and there is no performance improvement. Because of the partition of the big table, we can easily manage a partition and not to be a tang, but if it is not in accordance with the query of the partition condition, the index is a local index, the query needs to be separated to multiple local index trees to scan Not an index tree.
3, parallel (Parallel)
Today, I turned over the Oracle book, and I actually translated this word as a clue. At that time, I didn't have a faint back. Parallel is also a myth that is widely circulated. Before you say parallel, let's talk about a work allocation, there is a departmental leader, with 12 small soldiers (first said), one day, come to a small task, in fact, he is finished alone, However, he is not to distribute the following 12 people to do at the same time. The reason is blind division of labor. The parallel query is the case, suitable for a small amount of concurrent connection to the user in the data warehouse environment, performs a large query. As with the example above, if a big job is allocated to 12 people. Therefore, the problem that needs to be considered before deciding to use parallel query is: will not affect other users because of more and more resources, it is not a big, and the parallel can improve the query speed statement. In addition to parallel queries, parallel DDL is a means of more commonly reduced maintenance time for management.
