"Chairman, the second batch of 300 servers have been verified and passed. After 30 days of actual operation, the general failure rate is 17%, the serious failure rate is 4%, and the stability rate is 97.79%!" In the Jiangnan Network Center in Cao County, the person in charge here reported to Huang Hehui.

"No, thank you for your hard work!" Huang He nodded. Huang He was very satisfied with this result.

Although the failure rate of 7% sounds a bit high, and there is a serious failure rate of 1.8%, this means that 4 of every 100 servers have serious failures, which are not the kind that cannot be repaired in ten days and half a month, or even directly scrapped.

If the server purchased from outside had such a high failure rate, Huang He would have long been clamoring to return the product.

However, if these servers are produced by Jiangnan Group, it would be a different matter.

Yes, this batch of 300 servers is actually the product of Jiangnan Group, and it is the second batch of servers produced. The first batch has been sent to the Jiangnan Network Center in Wenzhou. There are a total of 600 servers in one south and one north, providing guarantees for Jiangnan Group's network services.

Because it is the first batch of products produced by oneself, it is the product of crossing the river by feeling the stones, and it is natural to have a high failure rate. If you want to produce it, you can have amazing quality. This is a situation that only happens inside.

When any new product is just developed, it is accompanied by various failures. Only by constantly looking for failures during use, accumulating experience, improving production processes, and avoiding those areas where problems are prone to occur can the quality of the server be gradually improved.

In fact, the failure rate of the first batch was higher, reaching more than 25%, so Huang He was already very satisfied with the failure rate of the second batch. Next, it only needs to keep improving.

At the same time, one thing must be noted is that although the failure rate is very high, the stability rate is also high.

The stability rate of 97.79% is almost comparable to that of servers from international first-tier brands, such as the HP servers purchased by Jiangnan Group before.

So the question is, why are the quality of these servers so bad and the stability rate is amazing?

That is because these servers took into account the problem of single server failure at the beginning of their design, so these servers designed a complete set of linkage systems, with 300 servers forming a whole. If any server fails, the task of this server will be quickly handed over to other servers for execution.

At the same time, because these 300 servers are actually responsible for the work of about 200 servers, even if dozens of servers are interrupted in an instant, they can continue to operate without any impact.

The reason why this time there is a 2% unstable effect is that there is a large-scale regional power outage, and there are two times when the local network has failed. In addition, there is another time when the storage server has failed, resulting in data loss and network fluctuations.

In short, these computing servers actually maintain a stable operation efficiency of 100%.

Of course, one thing must be admitted is that compared with international first-class server manufacturers... no, in fact, even compared with international second-class server manufacturers, Jiangnan Group's servers are still very different.

First of all, for data volume like Jiangnan Group, normally, purchasing 50 HP servers can be maintained normally. However, Huang He invested 600 servers in the two network centers of the north and south, so 200 of them were added to maintain stability, but 400 were actually working.

In other words, the efficiency of one HP server is equivalent to that of eight Jiangnan servers.

However, there is no way to do this. After all, the chips used by Jiangnan servers are all mid-range chips designed by themselves. They are two levels different from foreign servers in terms of computing core. In addition, the problems with other electronic components and the server design are still too immature. Therefore, the level that these Jiangnan servers can reach is the highest level that Jiangnan Group can manufacture. If you want to catch up with the international high-end level, it is still quite a long way to go.

Of course, Jiangnan Group's servers also have their own advantages, that is, the price is quite cheap.

For example, an HP server costs $1.2 million, which is the cheapest of all high-end servers.

But what about the price of Jiangnan servers?

For pricing, there is no public pricing because it has not been sold to the public yet, but Huang He is expected to sell for 500,000 yuan per unit.

Yes, you only need only 500,000 yuan to buy an enterprise-level server. If you convert it into US dollars, it only looks like a little over 60,000 US dollars, which is equivalent to 120 of HP servers.

The price of an HP server can buy 20 Jiangnan servers, and the computing effect of 20 Jiangnan service areas is equivalent to the efficiency of 2.5 HP servers. Everyone can calculate this cost account clearly.

But this is Huang He sets the external selling price, so what about the actual production cost?

If the cost of purchasing equipment is not considered, the production cost of a Jiangnan server is only about 200,000 yuan, of which 100,000 yuan is used to purchase various parts that Jiangnan Group cannot produce, and the remaining 100,000 yuan is almost used to purchase chips and labor costs.

However, the chip cost is a game where the left hand is to change to the right hand, because Nikon's ten lithography machines have been in place, and now all the chips used in Jiangnan servers can be self-sufficient.

For example, in the second batch of Jiangnan servers conducting experiments, 200 of them used chips produced by Jiangnan Group. After that, Jiangnan Group will completely stop purchasing chips, and the server group will all use chips developed and produced by itself.

By then, the production cost of Jiangnan servers can be further compressed. Selling one server can make a profit of more than 300,000 yuan. This is indeed a very amazing industry.

But thinking about HP buying $1.2 million, you can imagine how much money they made.

If Brother Ma could visit this base, most of his doubts would be answered. Because Jiangnan Group invested a large number of servers to maintain Facebook's operation at a very low cost.

If Tengda wants to maintain a server group of the same size and computing power, it may have to spend about 200 million US dollars, and Tengda does not have that much money at all.

So this is the fundamental gap in strength. Jiangnan and Tengda are no longer companies in the same dimension. Tengda is still just a simple network enterprise, but Jiangnan has evolved to the point of network operators + network hardware manufacturers. People can produce servers themselves, which is a problem that Tengda cannot solve at all.

It is also worth mentioning that although Jiangnan Group's computing servers are at the third-rate international level, Jiangnan Group's storage servers have reached the top international level.

This is mainly attributed to Tong He, and his graphene.

The black crystal that had caused a stir in the R&D center before was actually just the second graphene material with high practical value developed by Tong He. The first graphene material with high practical value was actually a storage particle.

Storage particles made of this special material can store data of 20 times the best storage particles in the world at this time at the same size.

Yes, it is 20 times the data, so even if the design of storage servers is far behind the international advanced level in other aspects, the storage servers produced by Jiangnan Group can achieve 5 times the effect of the best storage servers in the world.

Whether it is capacity, read and write speed, or size, all of them are more than 5 times the gap, which is approximately equivalent to the level of storage servers around 2008.

At this time, there were about 50 storage servers incorporated into the server group, which could store hundreds of millions of tbs of data. At least for two years, Jiangnan Group had no need for storage server expansion.

At the same time, Jiangnan Server is also planning to use storage servers as their first product to promote the international market. After all, the computing server is really unprofitable at this time.

So back to the point, as for Brother Xiao Ma’s other question, is the Internet problem.

First, because Jiangnan has established two network centers in the north and south, they can share all network speed needs and improve network efficiency.

But this is not the main reason, the main reason is the design optimization of Facebook itself.

In fact, it only took more than two months from the development of Facebook to the birth of the first version, and the next four months were all spent on continuing to polish Facebook.

Because during the actual test, the Jiangnan team found that the network that Facebook consumed was too huge. So after 4 months of optimization, Facebook stored almost all the pictures in the offline center, which consumed the most network.

In this way, except for opening the space for the first time, every time you open a web page, you can call the image data directly from the offline center, and you don’t need to connect to the network to obtain the data.

Second, Facebook is a variety of message circles carefully designed. In the message circle, the recommended space dynamics will be published directly on the circle page. You need to click on the space to read directly. This also omits the step of retrieving space data. It is enough to just retrieve part of text data or picture data, which naturally greatly saves the Internet speed.

In addition, there are dozens of carefully polished designs that can save traffic in details, which are all 4 months of research and development results.

As for Tengda, there are only 15 days in total, so the QQ space is basically based on the oo space. This one was directly sentenced to death as soon as it was born and was copied without any polished samples.

Although the code has been changed, the essential functional design is exactly the same. Therefore, if you want to see the dynamics of the qq space, you must first enter the qq space, read a large amount of spatial data from the server, and then read the dynamic data, so that the content of the space can be presented.

These steps greatly consume the Internet speed, which naturally leads to failure.

In fact, Brother Ma later discovered the problems, but all of these problems need time to polish them, but Brother Ma has no time anymore.

Purple-brush literature
Chapter completed!