On May 19, Dell Technologies launched the 3rd generation Intel Xeon Scalable Processor (code name: Ice Lake), which went on sale on the 13th of the new generation x86 server "Dell EMC PowerEdge" An online briefing session was held for the press about the five models.
17 new models of PowerEdge servers were announced by the US headquarters on March 18, including models equipped with 3rd generation AMD EPYC processors (code name: Milan) (Dell EMC PowerEdge C6525, R7525, R6525, R7515 and R6515) were released on the same day (Dell EMC PowerEdge XE8545 was released on March 29), and with the addition of models with Intel processors, the lineup of 3rd generation processors from both AMD and Intel has been completed. Become.
Mr. Mitsukichi Matsumoto, Executive Vice President of Dell Technologies, Mr. Iekazu Okano, Senior Product Manager of Product Division, Mr. Hiroshi Uehara, Executive Officer and General Manager of Product Division reviewed the company's server business strategy announced in December 2020, and introduced the Sustainable Development Goals (SDGs) and security initiatives as topics that have been of increasing social interest in recent years.In the server business strategy, the five basic measures for the new era are (1) expansion of the product portfolio, (2) introduction of a consumption model, (3) new pricing measures, (4) strengthening of customer support, and (5) sales ecosystem. reinforcement was launched. He said that the achievements at the moment, especially (1) and (3), have led to an increase in the share of the domestic server market, and that each of the initiatives is progressing steadily.
Looking back on the server business strategy announced in December 2020 and the current situationNext, he introduced the company's efforts regarding SDGs and security as urgent issues related to IT infrastructure. Among the SDGs-related initiatives, the company continues to make efforts to reduce carbon dioxide emissions with regard to the "Product Carbon Footprint," which is directly related to products. As an example, one of the company's high-end products, "PowerEdge R840" (at the maximum configuration), when operated for 5 years, which is the average service life, will emit 45,000 km of carbon dioxide equivalent. Introduced as a value comparable to the case. "We are using it as a measure to understand the actual situation and take measures," Matsumoto said.
Example of CO2 equivalent emissions of the company's high-end server "PowerEdge R840" calculated based on the concept of "Product Carbon Footprint". Based on this understanding of the current situation, we will work to reduce itThe graph "Dell Technologies Products Energy Intensity (Economic Efficiency)" was shown as a concrete result of our efforts. The term "Energy Intensity" is used, but in reality it can be viewed as the amount of power consumed to achieve the same performance. The reduction rate for monitors and PCs, which are devices for clients, did not reach the target value (80% reduction compared to fiscal 2012), so the overall average was also reduced by about 70%. Both storage and network have achieved their targets, and we can see that they are steadily producing results.
This is a graph showing the results of power consumption reduction for each type of company's products. We have already achieved results that exceed our goals for servers, storage, and networks for enterprisesIn addition, we are recycling resources related to product manufacturing and recovering marine plastics and reusing them as packaging materials and cushioning materials. We are also making efforts such as In addition, Hiroshi Uehara, executive officer and general manager of the product division of the company, points out, "There is a noticeable movement to select products with SDGs in mind when procuring products for domestic user companies." There is no doubt that further action will be required in the future. Finally, Mr. Matsumoto positioned the new-generation PowerEdge as "an 'innovation engine' that supports customer innovation."
An example of the company's SDGs-related initiatives. By the way, when I visited the company's U.S. headquarters in late fall 2019, I learned about "an initiative aimed at 100% recycling of carbon fiber mixed plastic materials used for the exterior of notebook PCs" and "used in enterprise products. They are listening to explanations about the promotion of recycling of packaging materials/cushioning materials, etc. It's not something that was suddenly brought up this time, but as a result of continuous efforts for many years, it can be said that various results are beginning to appear. He gave a product description of the new generation PowerEdge server. First, he listed three design concepts: "adaptive compute," "autonomous compute infrastructure," and "proactive resilience," and explained the corresponding design ideas and features.First, with regards to adaptive compute, various design variations are provided to suit different workloads. "Dell EMC PowerEdgeXR11" and "Dell EMC PowerEdgeXR12" (scheduled to be sold in mid-July), which are highly durable servers optimized for edge/telecommunications carriers, and GPUs for artificial intelligence/machine learning (AI/ML) In addition to models with characteristic designs such as the optimized "Dell EMC PowerEdge R750xa" and "Same XE8545", general rackmount servers are also divided into scale-out configurations and cloud/high-density implementations. .
Portfolio of new generation PowerEdge servers. Following the AMD processor-equipped model, the Intel processor-equipped model has been added this time, and the planned portfolio is almost complete (some models are scheduled to be released in mid-July) Overview of models for AI/ML. It is divided into R750xa, which emphasizes implementation efficiency, and XE8545, which specializes in performance. Another point is that iDRAC9 can monitor the GPU statusAs a unique initiative of the company, for example, the R750xa is a standard 2U rackmount server, but can be equipped with 4 double-wide GPUs. The XE8545, which features a combination of two generation AMD EPYC processors and NVIDIA HGX A100, is 4U size, and based on cooperation with NVIDIA, the maximum power supply to the GPU is 500W, which exceeds the catalog specification of 400W. A 15% improvement in performance was achieved. In addition, the company's unique system management processor "iDRAC9" also monitors GPUs, enabling real-time visualization of power consumption and temperature for each GPU, making it an extremely attractive hardware for users who want to make full use of GPUs. finished in clothing.
An overview of cooling/thermal management, a showcase for server vendors. There are three types of cooling fans with different maximum rotation speeds and air volumes, and the most suitable fan is selected according to the configuration of the server. In addition, the system board layout is left-right symmetrical, with a layout that does not interfere with airflow
For example, in the latest R650, the power supply unit, which was placed on one side in the previous generation R640, is split into two. It is said that even airflow is realized by making the entire arrangement symmetrical by installing it on both the left and right ends. In addition, by placing the cooling fan on the front side of the housing and placing the processor and memory immediately after the cooling fan, cool air is efficiently applied to both parts that are heat sources, while cold air also flows to the outside of the shroud that covers them. is secured, and consideration is given to the power supply unit and various expansion boards placed behind the processor/memory so that cold air hits them.
Let's take a look at the airflow ingenuity in line with the actual machine. The upper left is the front side and the lower right is the back side. The cooling fan is located on the front side, just before the CPU/memory, and efficiently blows cool air to the CPU/memory, which is the heat source. In reality, a shroud that covers the CPU / memory at the last minute is installed, but because the fan diameter has a margin, the air flow that does not involve cooling the CPU / memory is separated on the top of the shroud, and this is the rear expansion device turn to cooling. The power supply units are distributed to the left and right of the rear, aiming for the effect of quickly discharging hot air from the center after cooling the CPU / memory. Not only the memory module, but also the heat sink fins of the CPU, etc. are all arranged in an orderly manner so that they are parallel to the air flow. Detailed settings such as specifying the upper limit of temperature and setting the amount of airflow to PCIe are also possible, so it is possible to improve the efficiency and sophistication of cooling management not only for a single server but also for the machine room and the entire data center. Such detailed information grasped by iDRAC9 can be referenced from the outside via API, so by combining it with environment management tools in the machine room, it will be possible to link with air conditioning management on the facility side and improve efficiency. Regarding these efforts, he said, "Convenience when operated by large-scale users such as cloud providers is also considered, which will lead to differentiation from other companies' products."In addition to pursuing mechanical cooling efficiency, intelligent cooling control using iDRAC9 etc. also plays an important roleiDRAC9 is also used to strengthen security. The company provides "end-to-end supply chain security from product development to delivery", and ensures product safety and trust through silicon-level root of trust, firmware protection functions, and the formulation and operation of its own supplier security standards. We are striving to guarantee sexuality. As part of these efforts, we also provide "Secured Component Verification" that proves that unauthorized access has not occurred after the product is shipped from the company until it is delivered to the user. This function is realized by writing an encrypted digital signature to the internal tamper-proof storage area of the iDRAC9 at the time of shipment, and is provided as a paid option.
An overview of supply chain security. It is said that it has built a system that ``can guarantee that no one has made unauthorized access until the user turns on the power since it was manufactured in a factory certified by Dell''Finally, the new generation PowerEdge Mr. Hiroshi Uehara, executive officer and general manager of the product division, explained the measures to expand the server business. First, in light of the growing use of AI/ML, the company has prepared a server verification/PoC facility in Japan (in Tokyo), as well as a large-scale cluster environment that can be used remotely at the US headquarters. introduced the fact that the use of domestic users is also expanding. It is said that it is ready to support verification / PoC of R750xa / XE8545, etc., which is a server specialized for installing GPUs in the new generation server lineup.
Examples of efforts for AI/ML. In addition to providing specialized hardware, we also support the preparation of verification/PoC facilities and the use of IPUIn addition, the IPU (Intelligence Processing Unit) developed as a dedicated accelerator for AI/ML processing It also introduces that the company's PowerEdge R6525 is the only certified host server equipped with the "IPU-Machine M-2000" (1 unit achieves 1 petaflops computing performance) provided by GPARPHCORE as an accelerator module equipped with He emphasized that the company has a unique edge in the field of AI/ML processing. At the same time, it was announced that the "Dell de AI-Dell Encounter Program" will be launched as a program to match partners with excellent technology for AI / ML and user companies.
We also provide a matching service for partner companies with specialized knowledge and user companiesMr. Matsumoto made an additional comment, considering that the sales scale of domestic manufacturers is considerably inferior to that of global vendors. "We are in a situation where high-mix low-volume production is required, such as the introduction of new technology and the expansion of product variations according to user needs, and we are entering an era where scale matters more than ever before." He says that he will strive to provide customer value by taking advantage of global economies of scale, including breaking away from old Japanese business practices.
On the other hand, recently it has become impossible to ignore geopolitical risks regarding server hardware, and the social responsibility of continuing to be a "safe supplier" from the user's point of view has become heavier. there is Based on this situation, the company has made it clear that it will continue to take pride in manufacturing better and safer hardware.
The company has already announced "APEX" as an as-a-service model, but they do not think that hardware is no longer an issue when computing is provided as a service. Mr. Uehara explains this point with an analogy, "There is no one who does not care about the type of car when renting a car."
Some people may say, "It's the cheapest as long as it runs, so there's no problem", but even in that case, it's thought that they chose the "lowest cost model". The consolidation of IT vendors is progressing, and the number of companies that can provide hardware is narrowing down. As stated, it is expected that the hardware technology that supports the IT infrastructure will not be neglected in the future.
Category
Related Articles
Hot Articles