As for distributed storage, affected by the Internet, some domestic customers think that since the software definition is distributed, distributed storage should be the general server hardware of distributed storage software, forcing the decoupling of software and hardware.
As a result, a small number of users bought out Distributed Storage Software: unlimited time, unlimited version, unlimited capacity & ldquo; Three Unlimited & rdquo; Permission, and the hardware below is a general-purpose server, distributed Storage software is required to adapt various network cards, RAID cards, HDD disks, SSD disks, and other components of various server manufacturers' hardware. The annual bidding results are different, and the specifications are repeated.
As a result, the compatibility, adaptation and reliability of distributed storage have fallen into the abyss of endless disaster, but the benefits are very small. How can this industry be self-hematopoietic and sustainable?
In fact, the distributed storage system is not a general-purpose OS. It is also a professional storage system based on the tailored OS and built with tens of millions of lines of code. It has high complexity and is highly reliable, highly available, for the pursuit of high-performance and efficient O & M Management, manufacturers should strictly screen and control hardware components, including selection during research and development, testing and verification, and quality evaluation of components and driver versions, during the production phase, the debugging and testing of the storage system software and the testing of the baking machine ensure that there is no problem in the delivery to the site, and the requirements of these server hardware are far lower.
However, most of the time, the server hardware components that customers bid for are from various sources, which requires on-site evaluation, testing and verification. The work is actually very low-level and boring, and has no value-added significance.
Moreover, the reason why software is well-known in China is that the value of software is difficult to realize, and it falls into extremely low-price competition. The hardware at the bottom counts as a server, we also need to do complex distributed storage systems and a lot of compatibility and adaptation verification work. How can this industry continue to be closed-loop?
Just like an independent person, he needs to do endless work, but he has only a few basically negligible gains. He cannot make continuous evolution of self-hematopoiesis and is soon killed.
Why not bid separately for distributed storage system devices? The storage system itself is not a general-purpose OS such as Windows and red hat OS, but a dedicated device and a software-hardware collaboration device, thus ensuring data security, the source of components is strictly controlled, and customers and manufacturers do not have to do a lot of low-level boring and meaningless adaptation verification work. Why not do it!
The reliability of the distributed storage system as the Data & ldquo; Home & rdquo;, is the first. If the manufacturer has no control over the following hardware sources, the performance, reliability requirements, and quality control are uneven, especially at present, the quality control of small brand parts in China is even more uneven, which is difficult to guarantee data security. Who can take the blame for data loss or business interruption? However, the current distributed database is likely to fall into a similar dilemma. In addition to the complex logic of the database itself, the database also needs to be highly reliable in the underlying hardware, which increases the complexity.
Source: Bi xunshuo
