Fujitsu PRIMERGY RX2540 M4 server review and testing

Published on June 27, 2018

Fujitsu PRIMERGY RX2540 M4 server review and testing

    The rapid development of new technologies often entails situations when the equipment used is no longer able to cope with the tasks of companies. We have to either upgrade the existing fleet of server systems, or change them for more productive ones. At the same time, it is necessary to adequately evaluate the options for such a replacement, to know all the pros and cons of the equipment. We decided to test the Fujitsu PRIMERGY RX2540 M4 server and tell what features it provides to users. In this article we will talk about what the server consists of, about its architecture, we will describe the course and results of testing.


    So, the subject of our research is a 2-unit server (approx. 1 unit = 4.445 cm or 1.75 inches) installed in a 19 inch rack.

    Let's start with the front.


    Fig.1. Layout of HDD / SSD baskets

    1) There are 16 compartments for hard drives or solid state drives with a size of 2.5 inches. Hard disk drives (HDD) with a memory capacity of up to 2 TB and solid state drives (SSD) with a memory capacity of up to 7.68 TB are supported. The design of the compartment itself is shown in Figure 2.


    Fig.2. HDD / SSD

    Compartment This compartment is removed and inserted into the basket by pressing the green locking tab on the right.

    2) On the right side is the control panel.



    It includes two USB 3.0 connectors, a compartment for outputting the VGA connector, a Reset button (for emergency rebooting), disk operation indicators and a power on / off button.

    3) Under the control panel is an optical drive.


    Fig.4. The location of the optical drive

    It is an ultra-thin DVD drive connected via a SATA connector. The design of the optical drive is shown in Figure 5.

    Fig.5. Optical drive

    4) Under the optical drive there is a compartment for connecting a module with backup disks (or Backup).


    Fig. 6. Backup bay location

    Either RDX or LTO modules can be inserted into this bay. They look like this:


    Fig.7. LTO streamer and RDX cartridge.

    It is worthwhile to tell more about each type of module.

    RDX (from the English. Removable Disk Exchange) is a technology that improves the resiliency and durability of the drive. This module is a hard disk, fixed in a special case, providing a weak fixation of the disk. This technology allows you to avoid premature hard drive failure due to vibrations or falls from a small height.

    LTO (from the English. Linear Tape-Open) is a technology for recording data on magnetic tape. It may seem strange in our time to record information on a magnetic tape. However, LTO shows good durability results. According to the manufacturers, the magnetic tape can store information for up to 30 years (for railways the maximum value is 10 years). In addition, it is easier to recover data from a torn tape than from crumpled HD plates. LTO-7 is currently supported. It allows the use of cartridges with a memory capacity of up to 6 TB with a write speed of 300 MB / s and 750 MB / s with data compression. Although plans announced standard LTO-10. It will allow installing cartridges with a memory capacity of up to 48 TB and a write speed of 2750 MB / s.

    Go to the back of the server.
    1) The first thing you should pay attention to is nutrition.

    Fig.8. Location of connectors for power supplies

    The server supports up to 2 power supplies and provides “hot” redundancy. Accordingly, it is necessary to connect the power supply units to different phases so that in case of a failure at least one of them continues to work. Supports power supplies from 450 to 1300 watts. The selection of power supply units should be made on the basis of the load (the amount of RAM, disks, processors, etc.).


    Fig.9. Power supplies

    2) There are six expansion slots. Three of them are PCI-Express 3.0 x8 and three are PCI-Express 3.0 x 16.


    Fig.10. Location of PCI slots

    For all six slots to work, you need to install a second processor (the motherboard allows it). Expansion slots allow you to connect graphics, network cards and various controllers.

    3) On the back of the server there are connectors built into the motherboard.

    Fig.11. Location of integrated connectors

    This is a VGA connector, two USB 3.0 connectors, two gigabit ports and one gigabit port for management (or Management). It was mentioned above that VGA and USB connectors are present on the front panel. It is worth noting that there are several front panels for Fujitsu for this server. And some do not have these connectors. On the back side, they are required, and on the front panel are optional.

    It remains to examine the server from the inside and see what it is assembled from. Some information will repeat the above if you specify any connectors or modules. This is necessary in order to orient where and what is inside the system. Let's start with the location of the main elements.

    Fig.12. The inner part of the server

    1) The basis of any server is the motherboard.

    Fig.13. Motherboard Location The

    D3384 motherboard is installed on this server (only revisions can vary). The design of the board is as follows:


    Fig.14. Motherboard D3384

    It contains 2 processor sockets for Intel E5 processor family, 24 slots for DDR4 RAM (with a total memory capacity of up to 3 TB), an Intel C620 chipset, six PCIe connectors and the other inputs and outputs mentioned above.

    2) Now consider how the slots for RAM are located.

    Fig.15. Location of memory slots

    They are located in 3 places. In the upper and lower parts are six slots, and in the middle of the remaining twelve. These slots support the installation of memory strips up to 128 GB at a speed of 2666 million transmissions / s.

    3) And most importantly - processors.

    Fig.16. The location of connectors for processors

    They are located between the rows of RAM. Installation of processors of the entire E5 series is supported (up to the Intel Xeon E5-8180M-V5 processor, 2.5 GHz). The processor looks like this:

    Fig.17. Intel Xeon Processor E5-8180M-V5 processor

    As a cooling system for processors, passive radiators are installed. One of them is located in Figure 18.

    Fig.18. Passive radiator Heat Sink B1016-V1

    60X38 modules, which include 2 fans, are responsible for active cooling. They are placed in the following order:

    Fig.19. The location of the active cooling system

    And look like this:

    Fig.20. Coolers 60X38 A

    maximum of three such modules can be installed (that is, six active fans). With this configuration, the server will be cooled as much as possible.

    Now it’s worth telling with which configuration the tests were performed and what they showed.
    When testing, we used Microsoft Windows Server 2012 R2 Standard. The list of supported operating systems is quite impressive (including hypervisors). More information on the list can be found at the link .

    Server components are presented in the table below.

    The D3384-A1 motherboard (with 38400 MB / s bandwidth), two Intel Xeon E5-4112-V5 processors, four Fujitsu 32 GB DDR4-2666 RDIMM RAM boards, 8-port FTS disk controller PRAID CP400i (12 Gb / s for SAS and 6 Gb / s for SATA drives) with a 500 GB hard drive connected, a 2-port Intel X722 network controller (working under the 10GBASE-T standard) and a 2-port network controller, working on a 1000BASE-T standard.

    Testing


    So that the results do not seem abstract values, for comparison, we present the results of tests of other middle-class servers. Let's start with the processor test:

    1) CPU Queen. The essence of the test is to calculate the integers. This allows you to test the ability of the processor to branch and predict errors. To solve the problems, the “Queens Problem” algorithm is used. More details about this algorithm can be found on the link .


    The test showed 62251 values ​​per second. The result shows that the processor copes well with several tasks at the same time, which gives a big gain when used in virtualization.

    2) CPU PhotoWorxx - the performance of performing arithmetic operations and working with RGB images is tested. Its essence is as follows:

    • Filling the image with pixels of random colors;
    • Rotate the image by 90 and 180 degrees;
    • Image differentiation;
    • Color space transform (used when converting to JPEG).



    The test showed a result of 36443 million pixels per second. This suggests that the processor copes well when working with images.
    3) CPU ZLib - test the performance of the processor and memory subsystem using integer operations. Testing is done by creating ZIP archives. To do this, use the open library zlib ( more about the library ). It is often used to test multi-core systems.


    The indicator keeps at around 525.2 MB / s.
    4) FPU VP8 - processor testing by compressing the video with the Google VP8 codec or WebM 1.1.0. Encoding is carried out for 1 pass of the video stream, which has an extension of 1280x720 and comes with a speed of 8192 kbit / s (taking into account the maximum configured quality). The frames themselves are generated using a set of Julia fractals ( more). What is the difference between CPU tests called CPU and FPU? The fact is that the first processors could perform a strictly defined set of instructions. To expand the capabilities of the processor, co-processors were developed that performed specific mathematical operations and returned the result to the main processor. This allowed to increase the overall speed of the computer. Today, coprocessors are integrated into the central processor, and the boundary has begun to blur. FPU tests are the testing of “coprocessors” in the processor.


    The test showed 6493 units. Despite the fact that many coprocessors are integrated into the CPU, they do not show the highest result. To work with a video stream, you should use a separate video card. In our example, the CPU gives part of its processing power to the processing of video graphics.

    5) Memory Read - test the speed of sending data from RAM to the processor. That is, how fast the processor reads data from the RAM.


    The test shows that the processor reads data from the memory at a speed of 71320 MB / s.

    6) Memory Write - test the speed of sending data from the processor to the RAM. In other words, at what speed the processor writes data to the memory cells.


    The test showed that the processor writes data to RAM at a speed of 51299 MB / s.

    7) Memory Copy - test the speed of transferring data from one memory cell to another via the processor cache. That is, at what speed the processor will copy data from one memory cell to another.


    The test showed that the processor copies data from one cell to another at a speed of 61469 MB / s.

    8) Memory Latency - tests the average time for a processor to read data from RAM. If the previous reading test from memory showed a reading speed, then this test shows the average delay time.


    The test demonstrates that the average latency, when read, is 85.7 ns.

    Weigh the pros and cons


    Note the advantages:

    1) Compact server size

    2) The presence of 16 bays for HDD / SSD, which allows you to build disk arrays fault-tolerant subsystems

    3) The optical drive is a significant advantage. Often there are cases that it is necessary to restore the system from optical discs. The reason is that the optical drive is always determined by the system, and for various buses (for example, USB 3.0) installation of drivers is required or initialization may fail at system startup.

    4) Connecting RDX and LTO modules. They allow you to store information much longer than the usual HDD and SSD drives.

    5) The presence of 2 power supplies, which increases the server fault tolerance in the event of a power failure or failure of one of the units. This is a very important advantage, as many servers release with one power supply.

    6) Motherboard with wide connectivity:

    • Support 2 processors.
    • 24 slots for RAM with a total volume of up to 3 TB.
    • 6 PCIe connectors, which allows you to add a video adapter, connect additional network adapters and so on.

    7) Availability of a network controller operating at a speed of 10 Gbit / s. It should be noted that the controller contains 2 ports, which increases its fault tolerance and allows the use of traffic balancing.

    8) Impressive processor power when performing complex tasks and high speed memory access, as shown in tests CPU Queen, Memory Read / Write / Copy.

    Now let's talk about the disadvantages:

    1) Built-in RAID-controller. Its minus is manifested in a small number of ports. If you have compartments with 16 disks, you will have to purchase an additional controller to use all available disks at the same time.

    2) Weak performance in the test video stream and video acceleration. The entire burden falls on the central processor.

    Summing up, it is worth noting the good characteristics in terms of installed processor power and RAM, as well as high speed access to it. A large amount of disk space is also an advantage. All this allows you to use the server as a hosting for services with a large amount of data (databases, corporate mail, cloud services), virtualization, and more.

    The disadvantages include the built-in RAID controller with 8 ports (with 16 drive bays) and the lack of a video card for working with modern graphic applications using 3D models and video acceleration. In general, if necessary, the motherboard allows you to install additional RAID-controllers, video cards, network controllers and other devices. The server out of the box shows very good results, and when used in specific tasks it can be upgraded and upgraded.

    Thank you for your attention, we are ready to answer your questions.