Top of the big era

Zhou Buqi called He Yang, Meng Houkun, Yang Zhenkun, Zhou Feng and others to go to Lenovo to investigate their supercomputer project.

Along the way, expert Yang Zhenkun focused on introducing the concept of supercomputing HPC.

At that time, he was the chief scientist of Lenovo Research Institute and the core backbone of Lenovo's HPC team. Due to differences in technical directions, he left Lenovo to join Ziweixing to deepen distributed research and lay the foundation for cloud computing.

In some ways, supercomputing and cloud computing are a left-hand relationship.

They all use parallel relationships to connect some low-priced and low-performance hardware together through technical means to achieve supercomputing effects.

The difference is that HPC focuses more on computing. In addition to computing, cloud computing also includes a series of supporting ecosystems such as infrastructure, software, and platforms.

But this does not mean that the technical level of cloud computing is higher than that of HPC supercomputing.

The ecological nature of cloud computing is better than that of supercomputing, and the computing power of supercomputing is stronger than that of cloud computing.

When encountering extremely large amounts of computing, in order to save costs, the cloud computing platform must also cooperate with supercomputer providers to deploy HPC supercomputers on the cloud computing platform.

For example, an animation company receives an order and needs to use a lot of computing power to complete the animation production within 3 months. But after 3 months, they no longer need these abilities.

The cost of purchasing equipment for maintenance is very high. What should I do?

The answer is cloud computing.

For cloud computing platforms, this kind of calculation amount is also very large. It needs to deploy too much computing power, has to be turned on 24 hours a day, and the maintenance cost is also too high. In order to save costs, cloud computing platforms can seek help from HPC to reduce expenses.

In particular, it can be seen that HPC is an unpopular and high-end technical direction. The scope of marketization is extremely small, and only a few industries can apply it, such as weather prediction, oil exploration, chip mapping, aerospace and national defense, gene sequencing, and image processing. , biopharmaceuticals, etc. Not to mention ordinary people, 99% of companies have never heard of HPC in their lives.

Yang Zhenkun said: "There are more than a dozen supercomputing platforms in China, such as the two deployed by the Meteorological Bureau and Daqing Oilfield. The projects are really good."

"What about domestic products?"

"Differs greatly."

"I heard that there is already a supercomputing center in Shanghai."

"No one uses it at all."

Yang Zhenkun came from this field, so he understands it very well.

Zhou Buqi asked: "What's wrong?"

Yang Zhenkun said: "The two HPCs deployed by the Meteorological Bureau and Daqing Oilfield are produced by IBM and have a high degree of application. The supercomputing center in Shanghai is a collaboration between Sugon and some universities. The calculation amount is quite high, but there is no way Use it, but the compatibility is not good. The supercomputing center in Shanghai has a very good idea and is provided to fields such as chips, genes, pharmaceuticals, etc., but even the operating system and software are not deployed well."

Zhou Feng nodded, "Technology must compromise with the market. No matter how high the technology is, it will not work if the market does not accept it. Hardware, operating system and software must be highly coordinated."

Yang Zhenkun shook his head and sighed, "The technology is not good in the first place, but considering the market, it is even worse."

Zhou Buqi asked: "I heard that the domestically developed 'Tianhe-1' is one of the top five in the world? Isn't this amazing?"

"That was done by the national team. It is said that the investment is tens of billions..." Yang Zhenkun was a little helpless, "Actually, the principle of this thing is not complicated. The more chips, the greater the amount of calculation. Don't consider the market, don't pay attention to sales, don't consider Cost performance does not matter, just stack chips. The size of the Tianhe-1 HPC is larger than this building. The prospect of marketization is zero, the possibility of compatibility is zero, and all industrial software on the market cannot run. "

Zhou Buqi frowned, "What does that mean? Technology for the sake of technology?"

Yang Zhenkun said: "For aerospace and national defense, it is safer. However, if it is to be truly applied, it is impossible without continuous investment for ten or eight years. The overall technology is too immature."

Zhou Buqi nodded.

If it is the military industry, there is no reason to argue, and it cannot be judged by market logic.

Yang Zhenkun continued: "In the past few years, Lenovo's main PC business has poor technical content and low R\u0026D investment. Most of the R\u0026D funds have been invested in supercomputers, taking the market-oriented route. If Lenovo's products If none of them are suitable, there is no suitable one in China, so just look for HP.”

If it were ten years later, it would definitely be Lenovo. They started developing HPC in 2002. After more than ten years of research, they have reached the number one position in the world.

But we are still far from it now, and the research time is too short.

In the current list of the world's top 500 supercomputers, 212 are from HP, 188 are from IBM, and only 6 are from Lenovo. There is a huge gap.

Fortunately, the performance requirements for this batch of HPCs that cooperate with the public security system are not that high, that is, data query, DNA and fingerprint comparison, and image processing. It is enough to purchase a batch of low-cost and low-performance ones.

But the HPC donated to the Hong Kong Multimedia Laboratory is specially used for scientific research experiments, so it can't be too bad.

…

When I went to Lenovo, Yang Yuanqing came to greet me and accompanied me personally.

Recently, Mr. Yang's life has been a mixture of joy and sorrow.

Fortunately, Lenovo has entered the Fortune 500 list, ranking 499th. It is the only domestic private enterprise on the list. In the past few years, there was a lot of talk that Founder and Haier, which are the most likely to become the first domestic company to enter the Fortune 500 with non-monopoly market operations, failed to make the list. This wave is quite dazzling.

Sadly, in the past fiscal year amid the economic crisis, Lenovo, which he led, suffered its first major loss since its listing. In an emergency, Mr. Liu came back and re-planned and adjusted the corporate strategy. He could only take action again.

However, in front of the guests, Mr. Yang still showed a spring-like smile.

"Eh? Mr. Liu isn't here?"

"He's on the holding side."

"oh."

Zhou Buqi nodded, this matter has been causing quite a stir recently. Mr. Liu will take advantage of the economic crisis to take control of Legend Holdings in his own hands. Instead of having more than 50% of the state-owned assets as in the past, the state has the final say on decision-making, personnel, and financial rights. He has fostered a group of cronies. , the tissue structure is too bloated.

This kind of thing is too risky, and most people don't dare to participate. Not to mention participating, I didn't even dare to touch it. Wang Shi even left home early. However, Mr. Fanhai Lu is not afraid. He is a member of the standing committee of the applause club and is very confident.

Zhou Buqi is not involved in these things and mainly focuses on HPC.

If everything goes well, this wave of orders may reach about 500 million. In the future, there will be about 100 million annual equipment updates, maintenance, and upgrades.

This money is really a boost for Lenovo's HPC business, which has been investing for many years but has yet to produce results, and can also support their development in this direction.

The HPC market will have no effect in the short term. The hottest areas in the next few years will be cloud computing, big data, and the Internet of Things. But in a few years, with the acceleration of innovation and breakthroughs in new generation information and communication technologies such as 5G, Internet of Things, cloud computing, and big data, computing power will become a new productivity in the digital economy era, and the big stage of HPC will arrive, probably It will take 15-20 years.

Just wait and see!

Persistence is victory.

Fortunately, Lenovo paid enough attention to this field. Yang Yuanqing invited a foreigner with blue eyes and a big nose, saying that he came from the United States on a special trip. His name was Brad Smith, and he was the chief scientist of Lenovo's American Research Institute.

Then, he muttered a lot.

It is said that in the development process of HPC, several different representative technologies have emerged, such as vector computers, multi-processor shared memory HPC, and cluster systems built with multiple nodes.

From the current point of view, due to various reasons, vector computers and pure memory-sharing HPC systems are gradually being eliminated. Now cluster architecture is the mainstream and has absolute dominance. Basically 90% of HPC This architecture…

Zhou Buqi couldn't understand the other words either.

But he understands cluster architecture very well.

This is the technical direction of cloud computing, which is to connect countless servers together and work in a clustered manner.

Yang Yuanqing said with a smile: "Back then, we actually had two options when it came to choosing the direction of technology. One was cloud computing and the other was computer supercomputing. After careful consideration, we decided to develop HPC. If Lenovo did cloud computing, Maybe there will be no Ziweiyun."

Zhou Buqi laughed a few times but did not expose him.

Even if he is telling the truth, it is impossible for Lenovo to do cloud computing with its strength.

Not to mention Lenovo, IBM, HP, and Dell are all doing super computing. The development of cloud computing relies on Internet companies such as Amazon, Google, and Microsoft.

Supercomputing is hardware, an extension from PC to HPC, and the core is the computer.

Cloud computing is about ecology, which is a platform, system and application software deployed on cloud servers. The core is the Internet.

Immediately, Brad, an expert from the United States, talked about a very important point - the combination of software and hardware.

"HPC products have many parameters. Under the same cost constraints, there is room for optimization in parameter selection and software and hardware configuration. Usually when the customer's programs and applications remain unchanged, the optimized system may It brings 10% to 20% performance improvement or convenience in usage and management to customers.”

"Depend on!"

Zhou Buqi cursed in his heart, saying that Americans are indeed monkeys, and their sales tactics are quite powerful.

Now I can only buy Lenovo.

You really can't be careless about this kind of thing.

The image processing system developed by Ziweixing is world-class. It combines the research results of Shen Xiangyang's artificial intelligence vision team and the Hong Kong Multimedia Laboratory. It can identify this person through any deterministic blurred picture. His identity will greatly assist the police in handling cases.

The computer science department of the Chinese University of Hong Kong is ranked 27th in the world, and Shen Xiangyang represents the highest level in the field of computer vision. The latest results they have researched may not be much better even if Americans have them.

Software and hardware optimization means that the software side and the hardware side cooperate to make adjustments based on software requirements and hardware performance. If you are purchasing from foreign brands such as IBM, HP, Dell, Sony, etc., you must cooperate with them to conduct software analysis... Wouldn't this be exposed?

Just like the national team has to spend huge sums of money and endure huge waste to develop "Tianhe One", some things really cannot be left to the hands of foreigners.