bloom america

Chapter 874 Clay’s new supercomputer plan

Seymour Cray, the father of supercomputers, was tossing and turning in the laboratory at this time. After finding that he couldn't sleep, Seymour Cray got up again at three o'clock in the middle of the night, and then looked at his design plan.

"Why doesn't it still work?"

A new generation of supercomputers will use Capricorn as their core.

When Seymour Clay got this 32-bit microprocessor, he was fascinated by this magical processor. It adopted the latest technology and the best architecture, and had ten times the performance of the fc series.

performance……

These gave Seymour Cray a surge of inspiration when designing a new generation of supercomputers.

At this time, Master Clay suddenly thought of the new supercomputer architecture that he had discussed with Catherine, and the combination of this architecture and the current new processor seemed to be a good decision!

Thinking of this, Clay immediately organized manpower and began the research and development of supercomputers.

However, during the research process, Clay encountered unprecedented obstacles.

The new supercomputer adopts a turbine design. Each blade is a server, which can work together and can be increased or decreased at any time according to application needs.

This type of blade server is what Cray plans to use.

A blade server refers to a server unit that can be plugged into a standard-height rack-mounted chassis. It is a low-cost server platform that implements hahd (high availability and high density). This is a high-density computing environment.

A specially designed server will be more efficient than the original one.

A blade server is like a "blade", and each "blade" is actually a system motherboard.

These blade servers can start their own operating systems through the "onboard" hard disk, such as Phoenix system, cloud system, etc., similar to an individual server. In this mode, each motherboard runs its own system and services.

Different specified user groups are not related to each other.

However, administrators can use system software to integrate these motherboards into a server cluster. In cluster mode, all motherboards can be connected to provide a high-speed network environment and share resources at the same time to serve the same user group. In a cluster

By inserting a new "blade" into the system, the overall performance can be improved. And because each "blade" is hot-swappable, the system can be easily replaced and maintenance time is reduced to a minimum.

In the previous generation of Galaxy supercomputers, this concept was used to add modules to the computer.

But that kind of technology is insufficient compared to the current new technology perfected by Clay.

New technologies can use the computing power of supercomputers more thoroughly and skillfully to improve the efficiency of machines.

These blade servers were designed with low power consumption, small space, and low stand-alone price. At the same time, they also inherited some of the technical indicators of the Galaxy series servers, such as applying hot-swappability and redundancy to blade servers.

Among them, these designs meet the requirements for server performance in dense computing environments; some also effectively improve the stability of the server and core network performance through built-in load balancing technology. Moreover, blade servers can save the use space of the server to the maximum extent.

and costs, and provide users with flexible and convenient expansion and upgrade methods.

Having said that, Clay encountered various problems during the design.

First of all, this is a completely new architecture.

Secondly, 32-bit processors are also a brand-new architecture.

But at this time, Clay's previous experience was useless. Because at this time, he was exposed to a completely new technology. Such technology could not be achieved by Clay alone, or even by a team.

is very difficult.

Clay realized that he needed a big team and a long period of research to achieve his goals.

One year...no, maybe two or three years will be enough to achieve your goal.

Everything in the past needs to be demolished and redesigned, so it is impossible to do it without such a long time.

In the 21st century, blade servers have become the mainstream of high-performance computing clusters. Among the world's top 500 supercomputers, many new cluster systems adopt blade architecture. The use of blade servers can greatly reduce the number of

The number of external cables can greatly reduce the hidden dangers caused by cable connection failures and improve system reliability.

When Catherine was discussing with Clay, the two sides were talking about a supercomputer composed of such blade servers.

In terms of servers, Catherine's classification is similar to that in her previous life. This is entirely the result of Katherine's imitation.

According to different appearance and structure, servers are divided into three types: tower, rack, and blade servers.

——But in fact, there are only the first two, because the third type had not yet appeared when Catherine proposed this concept. It only existed in Catherine's notebook and Clay's mind.

Tower servers are generally the most common. Their appearance and structure are similar to ordinary PCs, but they are slightly larger. There is no unified standard for their dimensions.

The motherboard of the tower server is highly expandable and has many slots, and a lot of space is often reserved inside the chassis of the tower server for redundant expansion of hard drives, power supplies, etc. This kind of server does not require additional equipment and saves space.

There are not many requirements, and it has good scalability and high configuration. Therefore, it has a very wide range of applications and can meet the needs of common server applications.

This type of server is especially suitable for common entry-level and workgroup-level server applications, and its cost is relatively low. Its performance can meet the requirements of most small and medium-sized enterprise users. The current market demand is still huge.

For example, the server of Catherine's current personal website basically uses a tower server.

Moreover, some small companies also choose to use this kind of servers. In fact, their demand for such computers is not large.

Computer requirements are not necessarily higher, the cost is also an important reference.

However, this type of server also has many limitations. When multiple servers need to work at the same time to meet higher server application requirements, it is not suitable because it is relatively large, takes up a lot of space, and is inconvenient to manage.

Rack servers are actually products under industrial standardization. Their appearance is designed according to unified standards and used uniformly with cabinets to meet the enterprise's server intensive deployment needs. The main function of rack servers is to save space, because it can integrate multiple

Installing the server in a cabinet not only takes up less space, but also facilitates unified management.

The width of the rack server is 19 inches, and the height is in u (1u=1.75 inches=44.45mm). There are usually several standard servers of 1u, 2u, 3u, 4u, 5u, and 7u.

If it were the 21st century, the most popular would be 1u and 2u chassis, but it was different in Catherine's era. Because motherboards were all "silly, big, black and thick", 3u and 4u chassis were almost the choice for people.

And in the eyes of many people, this is already "small".

The advantage of this kind of server is that it takes up little space and is easy to manage uniformly. However, due to internal space limitations, its scalability is relatively limited.

This kind of server is mostly used by large enterprises with a large number of servers. There are also many enterprises that use this type of server, but hand over the server to a specialized server hosting agency for hosting. Especially at present, many website servers use this method.

.

If you want a better company, it is to use a supercomputer. The first-generation Celeron is no longer produced by the company, but the second-generation Galaxy is the mainstream in the market, and orders have even been queued up.

1975.

Even the top ten consortiums are using such supercomputers at this time, but such supercomputers are not regular "servers". They are just powerful enough to be able to do the job.

After all, at the beginning, what Catherine needed was stability, and the cloud system was a typical example of this. Such a machine could of course be used as a server.

But speaking of it, this "orthodox" choice is within the system that Seymour Clay is currently studying.

“How can we bring out the efficiency of the entire system?”

According to Catherine, such a system is very efficient, but during the research process, Clay found that this was not the case.

Because compared to "Galaxy modules" - individual modules of Galaxy computers - blade servers are more densely packed.

The core of the blade server is a 32-bit core with better technology, but due to space constraints, the heat dissipation is higher compared to the "Galaxy module".

Not only this problem, during the first design, Clay found that the architecture he designed could only have at most 50% of the performance he expected, and more performance was wasted.

"Let's do it again..."

Clay could only think like this.

"Maybe... I should change my thinking..."

The new core comes with its own graphics card, but my thinking is still stuck on the old computer. Why can't I think differently?

If you can also use the performance of the GPU, it should be even better...

…

By the way, our new computer is configured with a server chassis, 4u standard, using gtx670 graphics card, Laogun 5 full speed 60 frames...

Three updates today, the first update~

It seems that I have eaten something bad and my stomach hurts...

!#

www.ptwxz.com