The Story of IOE's golden partner and three brothers has proved to be successful after 30 years of separation. Back to the 19 s and 80 s when storage and computing were integrated, how to make the core database "safe"?
毕须说  2024-07-30 18:55   published in China

Let's take a look at the evolution of domestic and foreign technology architectures, as can be seen from the above figure. Before 1990, IBM dominated the country at that time. IBM hosts included servers, databases DB2, and local disk-frame storage, which was an integrated storage solution.

In 1990, in order to improve the high reliability, high availability and high performance of the data storage system, the storage system was independent from the server. Since then, EMC has initiated and led the rapid development of the storage industry for 30 years. Later, the database evolved into the Oracle era, and the computing evolved from a large machine to a small IBM Power Machine to a PC Server, resulting in a classic IOE architecture. Of course, the reliability of computing is becoming cheaper and cheaper from large and small computers to PC servers, and the reliability is continuously decreasing. With the continuous investment of EMC, Huawei, HDS and other manufacturers, the price of storage systems is becoming more and more favorable. However, the overall hardware and software architecture is continuously optimized and upgraded, and the reliability is continuously improved, the high availability of the IOE architecture is completed.

In fact, the IOE architecture of country A is very in line with human nature and management philosophy. There are professional companies and professional jobs in every field.

The database has professional companies such as Oracle and IBM (DB2) to solve the logic and consistency of the database itself, and to connect with key capabilities such as application SQL engine, storage engine, file ASM, replication switching, etc; storage is provided by vendors such as EMC, HW, and HDS to solve the problems of high reliability, high availability, and high performance of storage. It corresponds to the storage RAID pooling, A- A key capabilities such as multi-control balancing architecture and dual-active replication; computing provides CPU computing power and brain resources, and is pooled.

In the past 30 years, the storage and calculation have been separated, and the boundary is clear. Practice has proved to be successful.

However, under the agitation of some vendors, the IOE architecture is being removed. In order to achieve high concurrency and high performance, this architecture needs to be overturned, fully evolved into the so-called distributed architecture (but all master-slave multi-replica architecture). The core database is deployed on the local disk of the server, and the multi-replica storage and computing architecture is integrated. It does not focus on the high reliability of the underlying storage, is it really stable and reliable?

As can be seen from this figure, isn't it back to the 1880 s? Go through the process from integration of storage and calculation to separation of storage and calculation again!

The core logic of this architecture defined by Silicon Valley bosses is that professional companies do professional work, each field has a clear boundary, each field does professional things well, the division of labor is clear, and mutual coordination. However, as a platform software on the OS, the database does not care about the reliability of server hardware, local disk storage and RAID. If something goes wrong, no professional company will be responsible for it. As the disk is used for a long time, it will be broken. Slow disks, timeout disks, bugs, sub-health, etc. will occur. If the number is too large, it may fail every week, this brings about a series of complex O & M problems such as reliability and switching nodes to restore and rebuild replicas.

To make a similar analogy, the database is similar to a group of people, and the storage space is similar to a building. If the building is unstable, the pillar will fall down every week, the roof will leak, or the sewer will be blocked, maybe someday the pillar wall may collapse, can people still dare to live in the building?

Therefore, the stability and security of the overall architecture is the first priority, and how to live in a database is the first priority.

To remove IOE is to remove the corresponding products, not to deny the architecture. In other words, which domestic manufacturer has the ability to define it technical architectures and standards? Do you have such successful experience? Have you achieved the global top 1? At present, the so-called whole family bucket scheme is to think about the binding of food and food for the sake of interests, and it is coupled from top to bottom. This is really not professional! People have ten fingers, long and short. Some systems must have short boards, which may become the source of systemic disasters.

Source: Bi xunshuo

毕须说公众号二维码.jpg

Replies(
Sort By   
Reply
Reply