Each data center has its own constructor
Often, in the harsh Russian reality, the number of computers in office centers exceeds the planned number by 3-4 times. As a result, the server is overloaded, cooling systems do not really cope, and one can only dream of observing fire safety standards. Now imagine all this in the scale of the data center ...
It is no secret that for the average businessman-owner in the first place there will always be a dual goal: how to spend less and earn more. The goal, of course, is remarkable, but only if you try to link it with the wishes of the flexibility of the infrastructure for productive and hyper-scalable applications and tasks, there is a high probability that it will fail. But, as is known, there are no hopeless situations, and DELL have found their own way.
Following the needs of the market, the company introduced, including in Russia, new server chassis and PowerEdge C8000 modules.
')
Below we will try to figure out what kind of animal it is and what it is eaten with.
Perhaps, each of the respected harabrachiteli in childhood was some kind of designer. Someone built spaceships from LEGO, someone twisted a domestic "immortal" metal constructor, and someone even had transformers. In fact, the PowerEdge C8000 is nothing more than a constructor for “adult sysadmins,” allowing you to combine compute modules, storage modules, modules with a GPU, and power supplies in one compact 4U form factor. It is allowed to “shove” up to eight single-width modules or four double-width modules. In addition, each computing module is equivalent to a standard server equipped with a processor, memory, network interface, Baseboard Management Controller and local hard disk. The main highlight of the solution lies precisely in the variations of the modules mentioned.
Computing module PowerEdge C8220. It is a fairly large computing power, placed in a compact design. In the C8000 chassis, thus, you can put up to eight compute nodes, thereby placing up to 16 processors of the new generation in 4U.
In the case when computational power for a company shares the “palm of superiority” with visualization capabilities (for example, when working with research results), it makes sense to pay attention to the “hybrid” computing / GPU module PowerEdge C8220X , which allows to further increase productivity and density placement of processors and memory in the rack, as well as the use of GPU and other accelerators.
All this, of course, is good, but it is necessary to store data somewhere. And here,just like Chip and Dale, the PowerEdge C8000XD is in a hurry - a storage module that packs up to 1.4 times the capacity in a 4U rack. At the same time, depending on the task, the module can be equipped with both 2.5 "hard drives and 3.5" (including hot-swappable)
And now is the time to answer the logical question of some readers, who, most likely, did not leave them from the very beginning of the topic: “why is this for us?”. Imagine the situation: you have a cloud (not that with “white-maned horses”, but a normal iron cloud). A large client appears - a pharmaceutical company - that leases all the power and starts working on the visualization of its developments, etc. Obviously, for the tasks of the client "in honor" will be the GPU. But what to do to the “usual” cloud when this client leaves, and another one comes in its place, which will first of all need disk space (and the more, the better)? The data center will have to get out, incur costs of conversion, etc. But when using the C8000, the need for this will disappear once and for all. To install the new modules, it is not necessary to replace the chassis itself, which will significantly save money. In fact - a great designer with hot-swappable components. As a result, in the same room, without any re-planning and rearrangement, we can organize a system not only for high-performance computing (HPC), but also for processing / analyzing large amounts of data (Big Data), for hosting services, supplies cloud services and applications, for financial computing and even for automating electronic device design (EDA) processes.
• High-performance computing - research organizations, oil and gas companies, biomedical organizations, health organizations and other industries that need high-performance computing systems and parallel processing of large amounts of data from which clusters of high-density graphics processors can be formed.
• Large amounts of data — Web 2.0, retail stores, media, telecommunications companies, government and other organizations that require large-scale data analysis and scalability at lower operating costs.
• Hosting - providers of cloud services and Internet services, for example, large-scale hosting that require maximum performance and income per square meter of occupied space.
• Build cloud systems — vendors who need to quickly deploy servers according to changing workloads can, for example, quickly convert a four-node PowerEdge C8000 client configuration into a two-node database system with large disk space.
• Finance - banks, financial and investment firms that require high-performance parallel computing systems for processing large amounts of data with low latency, scalable capabilities and maximum performance per rack-space.
• CAD electronic devices - electronics manufacturers who need parallel computing systems for large amounts of data with a scalable combination of computing resources, graphics processor and data storage system.
Finally, separately I would like to say a few words about the “appetites” of the decision. Improved energy efficiency is achieved through power distribution within the rack. Internal power supplies are single-width slide rails (two for each housing) with one or two power supplies for each power module. It is also possible to refuse to install power supplies inside the chassis and use power from external sources.
At the same time, we did not leave the ideology of the “designer” here either. You can configure what is called, for every taste:
As you can see, the work was done qualitative and considerable, so it was especially interesting for us to see the first results of the implementations. Thus, the Texas Advanced Computing Center (TACC) Computing Center uses the PowerEdge C8000 to build its new supercomputer Stampede , which is expected to become a model for developing systems that provide petaflop computing performance. TASS Deputy Director, Dr. Dan Stanzione (Dan Stanzione) commented on the choice of center: “Dell solutions are indispensable for demanding computing and visualization in complex computer and engineering studies, such as weather forecasting, climate modeling, energy research and production, drug development, design and the production of new materials, as well as the construction of more economical and safer cars and aircraft. "
Well, time will tell. And for a snack, as a postscript, a pivot table of C8000 specifications.
It is no secret that for the average businessman-owner in the first place there will always be a dual goal: how to spend less and earn more. The goal, of course, is remarkable, but only if you try to link it with the wishes of the flexibility of the infrastructure for productive and hyper-scalable applications and tasks, there is a high probability that it will fail. But, as is known, there are no hopeless situations, and DELL have found their own way.
Following the needs of the market, the company introduced, including in Russia, new server chassis and PowerEdge C8000 modules.
')
Below we will try to figure out what kind of animal it is and what it is eaten with.
Initial inspection
Perhaps, each of the respected harabrachiteli in childhood was some kind of designer. Someone built spaceships from LEGO, someone twisted a domestic "immortal" metal constructor, and someone even had transformers. In fact, the PowerEdge C8000 is nothing more than a constructor for “adult sysadmins,” allowing you to combine compute modules, storage modules, modules with a GPU, and power supplies in one compact 4U form factor. It is allowed to “shove” up to eight single-width modules or four double-width modules. In addition, each computing module is equivalent to a standard server equipped with a processor, memory, network interface, Baseboard Management Controller and local hard disk. The main highlight of the solution lies precisely in the variations of the modules mentioned.
Computing module PowerEdge C8220. It is a fairly large computing power, placed in a compact design. In the C8000 chassis, thus, you can put up to eight compute nodes, thereby placing up to 16 processors of the new generation in 4U.
In the case when computational power for a company shares the “palm of superiority” with visualization capabilities (for example, when working with research results), it makes sense to pay attention to the “hybrid” computing / GPU module PowerEdge C8220X , which allows to further increase productivity and density placement of processors and memory in the rack, as well as the use of GPU and other accelerators.
All this, of course, is good, but it is necessary to store data somewhere. And here,
Time "H"
And now is the time to answer the logical question of some readers, who, most likely, did not leave them from the very beginning of the topic: “why is this for us?”. Imagine the situation: you have a cloud (not that with “white-maned horses”, but a normal iron cloud). A large client appears - a pharmaceutical company - that leases all the power and starts working on the visualization of its developments, etc. Obviously, for the tasks of the client "in honor" will be the GPU. But what to do to the “usual” cloud when this client leaves, and another one comes in its place, which will first of all need disk space (and the more, the better)? The data center will have to get out, incur costs of conversion, etc. But when using the C8000, the need for this will disappear once and for all. To install the new modules, it is not necessary to replace the chassis itself, which will significantly save money. In fact - a great designer with hot-swappable components. As a result, in the same room, without any re-planning and rearrangement, we can organize a system not only for high-performance computing (HPC), but also for processing / analyzing large amounts of data (Big Data), for hosting services, supplies cloud services and applications, for financial computing and even for automating electronic device design (EDA) processes.
Let's expand the topic a bit with examples:
• High-performance computing - research organizations, oil and gas companies, biomedical organizations, health organizations and other industries that need high-performance computing systems and parallel processing of large amounts of data from which clusters of high-density graphics processors can be formed.
• Large amounts of data — Web 2.0, retail stores, media, telecommunications companies, government and other organizations that require large-scale data analysis and scalability at lower operating costs.
• Hosting - providers of cloud services and Internet services, for example, large-scale hosting that require maximum performance and income per square meter of occupied space.
• Build cloud systems — vendors who need to quickly deploy servers according to changing workloads can, for example, quickly convert a four-node PowerEdge C8000 client configuration into a two-node database system with large disk space.
• Finance - banks, financial and investment firms that require high-performance parallel computing systems for processing large amounts of data with low latency, scalable capabilities and maximum performance per rack-space.
• CAD electronic devices - electronics manufacturers who need parallel computing systems for large amounts of data with a scalable combination of computing resources, graphics processor and data storage system.
Finally, separately I would like to say a few words about the “appetites” of the decision. Improved energy efficiency is achieved through power distribution within the rack. Internal power supplies are single-width slide rails (two for each housing) with one or two power supplies for each power module. It is also possible to refuse to install power supplies inside the chassis and use power from external sources.
At the same time, we did not leave the ideology of the “designer” here either. You can configure what is called, for every taste:
- 4 + 0 to 5600 W without reservation
- 2 + 0 to 2600 W without reservation
- 3 + 1 to 4200 W with redundancy
- 2 + 2 to 2800 W with redundancy
- 2 + 1 up to 2800 W with redundancy
- 1 + 1 to 1400 watts with redundancy
As you can see, the work was done qualitative and considerable, so it was especially interesting for us to see the first results of the implementations. Thus, the Texas Advanced Computing Center (TACC) Computing Center uses the PowerEdge C8000 to build its new supercomputer Stampede , which is expected to become a model for developing systems that provide petaflop computing performance. TASS Deputy Director, Dr. Dan Stanzione (Dan Stanzione) commented on the choice of center: “Dell solutions are indispensable for demanding computing and visualization in complex computer and engineering studies, such as weather forecasting, climate modeling, energy research and production, drug development, design and the production of new materials, as well as the construction of more economical and safer cars and aircraft. "
Well, time will tell. And for a snack, as a postscript, a pivot table of C8000 specifications.
| Chassis | 4U rack height - installation up to 8 pcs. modules of single width, or up to 4 pcs. double width modules (or their combination), with 2 power modules installed. - installation up to 10 pcs. modules of single width, or up to 4 pcs. double width modules (or their combination) using external power modules. |
| Types of modules | 1) C8220 single-width computing module: - 2 x Intel Xeon E5-2600, 16 x DDR3, 2 x 2.5 ”HDD (without hot plug), 2 x 1GbE, 2 x PCIE x8, 1 x PCIE slot for mezzanine cards; 2) C8220X double-width computing module: - 2 x Intel Xeon E5-2600, 16 x DDR3, 2 x 2.5 ”HDD (without hot plug) + 8 x 2.5” HDD or 4 x 3.5 ”HDD (hot plug), 2 x 1GbE, 1 x PCIE x16, 1 x PCIE mezzanine card slot; 3) C8220X double-width computing module with GPU: - 2 x Intel Xeon E5-2600, 16 x DDR3, 2 x 2.5 ”HDD (without hot plug), 2 x 1GbE, 2 x GPGPU (300W), 1 x PCIE slot for mezzanine cards; 4) C8000XD double width storage module - 12 x 2.5 ”(15mm) (hot plug) or - 12 x 3.5 ”(hot plug) or - 24 x 2.5 ”(9mm) (without hot swapping) |
| Power supplies | Up to 4 pcs. 1400W (efficiency 94+ Platinum) with 2 single-width modules. Configurable: 4 + 0 to 5600 W without reservation 3 + 1 to 4200 W with redundancy 2 + 2 to 2800 W with redundancy 2 + 1 up to 2800 W with redundancy 2 + 0 to 2600 W without reservation 1 + 1 to 1400 watts with redundancy |
| Fans | 6 pieces 120mm non-redundant, fixed, highly efficient with PWM control |
Source: https://habr.com/ru/post/155963/
All Articles