The symbiosis of software and hardware-inspired by One Thousand and One Nights

The symbiosis of software and hardware-inspired by One Thousand and One Nights

Liao Heng

We have talked many times about the twists and turns of the development of hardware architecture. It is often like a twists and turns chase story. Because the decision makers of software architecture are often equipment buyers, the designers of the hardware architecture have become the upstream providers in the supply chain. Such a supply-demand relationship determines the pursuit and the Pursuit relationship between the two in a free-market economy.

In the era when the enterprise IT market is king, buyers are widely distributed all over the world, and the IT departments of countless large and small companies and institutions are integrated as the buyers; large computer equipment OEMs (such as HP, Dell, and IBM), with years of technical advantages, market capabilities, and global sales and service networks, occupy a dominant market share. In this way, suppliers have established a strong position and led the development of hardware technology. The root cause is that, as the next buyer, although the purchasing power is very strong in general, each individual cannot occupy a certain share of the market. Therefore, individual purchasers lack the power to dominate the hardware trend. In the long run, software architects did not expect that hardware could change their mindset. Instead, they forgot the possibility that hardware and software architectures could complement each other.

In the IT world after 2000, with the rapid growth of large Internet companies, the imbalance between hardware and software has quietly changed. When the top 10 Internet data centers in the world combined up and purchased devices to occupy more than 20% of the IT market, strong traditional markets and technologies (such as IBM, HP, and Dell)-these giant OEMs face enormous challenges and even fall into the crisis of survival. With unparalleled shopping power, giant customers of ultra-large scale data centers quickly seize the dominant position in the market and are seizing the dominant control of hardware technology. They can continuously lower the prices of hardware products through public bidding, and require the best product quality and service at the same time; it can also guide the market to develop the hardware system architecture most suitable for their applications through the development of hardware specifications; in addition, it can eliminate the extra costs of intermediate links by directly purchasing from the component supplier. In today's cloud era, large buyers have become new giants in the market, software and even the guide and guide of hardware technology.

In recent years, we have seen the initial efforts of Internet companies to guide the hardware architecture, such as projectscorpio, the leading bat program in China, and opencompute of Facebook. There are many similarities between them. We don't have to infer that the technology is advanced or backward, because the trend will not have a long-term impact in a few years, but if Scorpio can be truly used as the mainstream deployment of bat, in terms of the total market volume, opencompute is better. Therefore, opencompute emphasizes its openness and has even become an independent Open Industry Alliance from Facebook, hoping to attract more buyers by opening up the market, it also attracts many other technological innovations by opening up. Scorpio is still very closed, and even the chip vendors that play a key role in the hardware technology field (except Intel) Open up, because the preliminary work did not focus on architecture design, focusing on the mechanical and electrical design, it may also be caused by the lack of confidence in the hardware expertise and the lack of confidence in the hardware system. Similarly, opencompute is not like a mature standard organization (such as IEEE, IETF, and ITU) in its early and even current stages. It is more like a free market for family members. This is not unrelated to the lack of technical strength (lack of hardware personnel.

At this stage in 2014, the old balance has been broken and a new pattern is being formed. At the moment of this change, some old mindset still exists due to its inertia, which slowed down the pace of technological development to a certain extent. Even Google and Facebook, which are at the forefront of Hardware Innovation, have a hardware team of hundreds of people, while bat's Hardware team is also in 10. This kind of human resources is less than of the companies with tens of thousands of users. Compared with the scale of thousands of hardware project engineers in traditional hardware OEMs, it is not enough to undertake the capability to continuously develop new hardware architectures. While the ODM, Which is robbing traditional OEM market share, takes production as its foundation, and does not have enough hardware R & D scale to support the long-term development of hardware technology. Therefore, we are bold in pushing forward that this bypass OEM market is difficult to sustain, and the new generation of hardware OEMs that replace it will be centered on the data center market, therefore, it is no longer necessary to create a large sales and service system across the world, but it must be able to drive the continuous innovation of the new hardware system. They can be from the data center, the self-improvement ODM, the success OEM, the more upstream chip manufacturers, or a combination of four.

The author is in the most upstream chip manufacturers in the hardware industry. In this era of change, he often feels the anxiety and pain. Many of the giant manufacturers that have worked with the author well defined the SAS standards for many years are unknown today. I wonder if they are facing the pressure of layoff or forced early retirement. When none of them can continue to work, where will the next protocol standard to replace SAS come from? If the hardware vendors lose profits, who can pay for the further development of these complex technologies? The turning of the wheel of history is something that practitioners must face from time to time.

This reminds me of the Ancient Arabian story, the introduction of the one thousand and one night. The brutal king takes a bride every night and kills him before the sun rises. Later, a wise woman brought fascinating stories to the King. The fascinating king fell asleep and forgot everything. In this way, the smart bride tells the story of one thousand and one nights, and the king has changed his brutal habits.

On this new Arabian Night, Will giant Internet companies become the overlord of hardware manufacturers? How can hardware manufacturers and hardware practitioners who struggle to survive in the supply chain become the surviving bride? Perhaps the ancient wisdom of Arabia provided the only way to survive thousands of years ago, that is, continuous innovation-constantly providing the unique value required and irreplaceable by software systems, never walk in front of the wheel to avoid the fate of rolling.