Method and device for testing single-particle effects of large-capacity solid-state hard disks based on high-energy particles

Abstract

The present invention discloses a single particle effect test method and device for a large-capacity solid-state hard disk based on high-energy particles. The test method includes: performing an SSD user data presetting experiment; based on the data presetting experiment, using a device to perform an irradiation test to determine whether the SSD module has single particle locking, functional interruption, and errors under irradiation conditions, determining the judgment result and performing quantitative calculation of the error, and outputting a functional test result; based on the functional test result, performing a silent data destruction test in the post-irradiation test phase; according to the silent data destruction test, recording silent data destruction and fixed bit error data; based on the silent data destruction and fixed bit error data, performing error characteristic quantitative calculation, and outputting the calculation result. The present invention can realize parameterized evaluation of SSD radiation resistance performance.

Description

Technical Field

[0001] The present invention belongs to the field of integrated circuit module testing, and in particular relates to a single particle effect testing method and device for a large-capacity solid-state hard disk based on high-energy particles. Background Art

[0002] With the widespread application of solid state drives (SSDs), users have higher and higher requirements for their reliability. Single event effects (SEEs) can cause abnormal logical functions of storage systems, among which single event functional interruptions (SEFIs) have the greatest impact. The displacement of atoms inside a material caused by elastic collisions of high-energy particles is called displacement damage effect (DDE). DDE can cause the degradation of certain electrical parameters in the device, and even the device will eventually lose its normal function.

[0003] At present, there is still a lack of analysis technology and evaluation methods for the SSD function interruption phenomenon, especially for the positioning of sensitive devices and the determination of error rate evaluation indicators. For example, prior art 1 discloses a NAND flash memory solid state hard disk space environment effect test system and test method, in which the main experimental method for single particle irradiation is to use a laser pulse generator to simulate high-energy particles incident on key devices in the SSD. However, this test method requires these devices to be unpacked to expose the chip surface for testing, which faces two major challenges. First, with the increase in chip packaging density, the current SSD chip mainly uses ball grid array packaging (Ball Grid Array, BGA). Therefore, in order to ensure that the chip can still work after unpacking, it can only be unpacked from the front of the chip. At present, NAND flash memory chips and main controller chips still use wire bonding packaging structure, and there are multi-layer metal wiring structures on the front of the chip. This structure will reflect the incident pulse laser, thereby affecting the effectiveness of the test. At the same time, there are necessary strong acid corrosion and heating steps for chip unpacking, which is easy to cause permanent damage to some chips. Secondly, pulsed laser can only perform irradiation evaluation on a single chip in the SSD at the same time, which makes it difficult to reflect the sensitivity of each type of device under the same irradiation conditions. Prior art 2 discloses a test method for the single particle flip cross section of a solid state drive. A method for evaluating the single particle flip count of the solid state drive as a whole is proposed, which uses the sum of the single particle flip counts of the flash memory unit and the SSD controller. Since this test method does not take into account the impact of the single particle effect of the external dynamic random access memory of the controller on the overall SSD system. Therefore, it is more suitable for evaluating SSDs with a cacheless design. At the same time, the latest SSDs all apply error checking and correction modules in NAND flash memory, static random access memory inside the controller, and dynamic random access memory outside the controller. Therefore, errors caused by single particle effects in SSDs can be divided into correctable errors and uncorrectable errors. Actual engineering applications pay more attention to uncorrectable errors, and this invention patent only tests the original error rate occurring in the SSD. The prior art lacks a method for evaluating the single particle effect of SSDs under actual working conditions and a quantitative analysis method for test results. Summary of the invention

[0004] In order to solve the above technical problems and more comprehensively reflect the radiation resistance performance of SSD, the present invention proposes a single particle effect test method and device for large-capacity solid-state hard disks based on high-energy particles, which can realize parametric evaluation of the radiation resistance performance of SSD.

[0005] To achieve the above object, the present invention provides a single particle effect test method for a large-capacity solid-state hard disk based on high-energy particles, comprising:

[0006] Perform an SSD user data pre-set experiment, send data packets of the specified size to the SSD data to be tested and add verification information to each data packet;

[0007] Based on the data preset experiment, start the irradiation test to determine whether the electronic components in the SSD have functional interruption and single event lockup under the irradiation condition, and determine the judgment result;

[0008] According to the judgment result, a functional test is performed to obtain a functional test result, and based on the functional test result, a silent data destruction test and a fixed bit error test are performed in a post-irradiation test phase;

[0009] According to the silent data corruption test and the fixed bit error test, recording the number of silent data corruption data blocks and the fixed bit error data;

[0010] Based on the silent data destruction quantity and the fixed bit error data, error characteristic quantification calculation is performed and the calculation result is output.

[0011] Optionally, based on the data preset experiment, starting the irradiation test includes:

[0012] In a high-energy particle irradiation environment, the test server performs 4KVerify mixed random read and write tests on the SSDs to be tested in batches to obtain performance information and error information of the SSDs.

[0013] Optionally, determining whether electronic components in the SSD have functional interruption and single event lockup under irradiation conditions, and determining the determination result includes:

[0014] If a single-particle lock occurs, power off and collect the BMC overcurrent warning log of the test server;

[0015] If single event locking does not occur, the functional test is performed.

[0016] Optionally, based on the functional test results, performing a silent data corruption test and a fixed bit error test in a post-irradiation test phase includes:

[0017] If the functional test fails, log data is collected, functional recovery and firmware rewriting are performed, and after success, silent data destruction testing is performed;

[0018] If the functional test is successful, the silent data corruption test and the fixed bit error test are performed according to the collected error log information.

[0019] Optionally, according to the silent data corruption test and the fixed bit error test, recording the fixed bit error data includes:

[0020] Use fio's verify function to perform a read test. If a read data error occurs and the controller does not report the error, it is considered that a silent data corruption error has occurred, and the number of silent data corruption data blocks is recorded. If repeated error correction records at fixed physical addresses are found in the dynamic random access memory during the test, fixed bit error data is recorded.

[0021] Optionally, based on the fixed bit error data, error characteristic quantification calculation is performed, and outputting the calculation result includes:

[0022] The error data generated by the controller and the dynamic random access memory and the fixed bit error data are used to perform quantitative calculation of error characteristics, and the calculation results are statistically calculated and outputted.

[0023] The present invention also provides a high-energy particle-based single-particle effect testing device for a large-capacity solid-state hard disk, comprising: a control center and an irradiation chamber connected via Ethernet;

[0024] The control center is used to control the irradiation chamber;

[0025] The irradiation chamber includes: a high-energy particle beam window, a module under test, a high-energy particle absorber, a test server, a PCIex16 cable, a SlimSAS retiming expansion card, an SFF-8654U.2 interface cable, and an SFF-8088SATA interface cable;

[0026] The high-energy particle beam window is used to isolate the high vacuum environment inside the accelerator beam pipeline from the external normal pressure environment, while allowing the accelerated particle beam to be safely and effectively led out of the accelerator;

[0027] The module to be tested is used to fix the SSD to be tested and ensure that the center point of the high-energy particle beam is located at the expected position;

[0028] The high-energy particle absorber is used to absorb the remaining high-energy particles to prevent interference with other sensitive devices in the test environment;

[0029] The test server is used to run the test scripts and test cases of the SSD, monitor the working status of the SSD and collect the error data that occurs;

[0030] The PCIe x16 cable is used to connect the test server PCIe interface and the SlimSAS retiming expansion card, so that the test server is as far away from the beam window as possible;

[0031] The SlimSAS retiming expansion card is used to convert PCIe signals into SlimSAS standard signals and to perform retiming to compensate for link loss to meet the signal integrity requirements of the PCIe link;

[0032] The SFF-8645U.2 interface cable is used to connect the data of the SSD to be tested using the U.2 interface and the SlimSAS retiming expansion card to meet the needs of simultaneous irradiation testing of multiple SSDs;

[0033] The SFF-8088 SATA interface cable is used to directly connect the SSD to be tested using the SATA interface and the test server.

[0034] Optionally, the control center includes: a high-energy particle collimator controller, a high-energy particle switch and a monitoring computer;

[0035] The high-energy particle collimator controller is used to control the area and flux of high-energy particle irradiation;

[0036] The high-energy particle switch is used to control the high-energy particle irradiation time;

[0037] The monitoring computer is used to remotely log in to the test server through the SSH protocol to ensure that the test personnel can safely and real-time control and observe the test server in the irradiation room.

[0038] Compared with the prior art, the present invention has the following advantages and technical effects:

[0039] 1. Reduce the complexity and damage risk of unpacking operations: The test method in the existing invention technology relies on unpacking the SSD chip. As the packaging density continues to increase, the difficulty of unpacking becomes greater and greater. The present invention improves the experimental method, avoids unpacking operations, and reduces the impact of the test device on chip performance, thereby improving the reliability and effectiveness of the test.

[0040] 2. Complete SSD single-particle effect test under normal working conditions: The technology mentioned in the existing invention can only perform radiation evaluation on a single chip in the SSD and depends on specific write data or engineering development debugging mode. At the same time, the test method of the present invention does not require the SSD to enter the debugging mode, but can complete the single-particle effect test and evaluation in its normal working mode. This feature not only simplifies the test process, but can more truly reflect the performance and stability of the SSD under actual working conditions.

[0041] 3. Comprehensively consider the impact of system-level single-particle effects: The present invention overcomes the limitation of the existing technology that is only specific to NAND flash memory or controllers. By improving the experimental design, the sensitivity of multiple key components inside the SSD can be evaluated simultaneously under the same irradiation conditions, significantly enhancing the analysis capabilities of various types of components. At the same time, the testing technology in the existing invention is not suitable for SSDs with cache. The present invention can test and evaluate single-particle effects in SSDs with cache designs, broadening the scope of technical application and being suitable for the current mainstream design of SSDs.

[0042] 4. Uncorrectable error testing and characterization method: Since modern SSDs widely use error checking and correction modules, the present invention can effectively distinguish between correctable and uncorrectable errors caused by single particle effects and provide a test characterization method, especially focusing on uncorrectable errors that are more important in engineering applications, thereby improving the evaluation accuracy of SSD working reliability.

[0043] 5. In summary, the present invention not only overcomes the deficiencies of the prior art in terms of test operation, device sensitivity analysis, and system-level effect evaluation, but also provides a more comprehensive and accurate SSD single particle effect evaluation method, and better reflects the performance of modern SSDs under actual working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0045] Figure 1 This is a flow chart of a single-particle effect testing method for a large-capacity solid-state hard disk based on high-energy particles according to an embodiment of the present invention;

[0046] Figure 2 This is a diagram illustrating the principle of an uncorrectable error injection threshold in a NAND flash memory in an SSD according to an embodiment of the present invention;

[0047] Figure 3 The solid-state memory module hardware structure diagram of the embodiment of the present invention includes: 1. input / output interface; 2. internal interconnect bus; 3. central processing unit; 4. instruction / data tight coupling matrix; 5. cache; 6. flash memory controller; 7. NAND flash memory; 8. dynamic random access memory physical interface; 9. NOR flash memory; 10. dynamic random access memory; 11. power management module;

[0048] Figure 4 The firmware functional structure diagram of the SSD of an embodiment of the present invention, wherein: (a), input / output interface controller, (b), flash memory conversion layer, (c), wear leveling control program, (d), garbage collection program, (e), bad block management program, (f), error check and correction module, (g), flash memory dedicated error check and correction technology module, (h), flash memory controller, (i), flash memory channel;

[0049] Figure 5This is a structural diagram of a single particle effect test device for a large-capacity solid-state hard disk based on high-energy particles according to an embodiment of the present invention, wherein: A, high-energy particle collimator, B, high-energy particle switch, C, monitoring computer, D, control center, E, Ethernet, F, high-energy particle beam window, G, module under test, H, particle absorber, I, remaining high-energy particles, J, high-energy particles, K, SFF-8654U.2 interface cable, L, test server, M, irradiation chamber, N, SFF-8088SATA interface cable, O, PCIe x16 cable, and P, SlimSAS retiming expansion card;

[0050] Figure 6 This is a structure diagram of a high-energy particle irradiation test program according to an embodiment of the present invention, wherein: ①, rate data statistics module; ②, I / O interruption alarm module; ③, SSD load control program; ④, iostat; ⑤, smartctl; ⑥, telemetry; ⑦, dmesg; ⑧, Linux I / O disk driver;

[0051] Figure 7 This is a diagram showing the effect of a read error injection threshold test in a NAND flash memory according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0053] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0054] like Figure 3As shown, the hardware composition of the solid-state memory module mainly includes: a solid-state memory controller, a NAND flash memory 7 and a dynamic random access memory 10 (Dynamic Random Access Memory, DRAM), etc. The solid-state memory controller contains an input and output interface 1 between the host, an internal interconnection bus 2 for interconnecting various peripherals, a command / data tightly coupled matrix 4 and a cache 5 composed of a central processor 3 and SRAM units, a flash memory controller 6 responsible for communicating with the NAND flash memory and an integrated circuit of a dynamic random access memory physical interface 8. The NOR flash memory 9 is mainly used to store firmware programs executed by the main control. The power management module 11 is usually composed of several DC-DC conversion circuits. In order to prevent the data from being damaged after the SSD is abnormally powered off, the power management module will also include a backup power supply composed of electrolytic capacitors or tantalum capacitors. Existing research shows that in the solid-state memory controller, the command / data tightly coupled matrix and cache are considered to be the most sensitive parts to single particle effects. As the chip manufacturing process continues to decrease, the sensitivity of dynamic random access memory and NAND flash memory is also increasing.

[0055] like Figure 4 The figure shows the firmware structure of the SSD, where the main programs in the firmware include the input / output interface controller a and the error check and correction module f. The flash translation layer is responsible for converting the logical block address (LogicBlockAddress, LBA) for communicating with the host into the physical address in the NAND flash memory. In addition, the flash translation layer b also includes the wear leveling control program c, the garbage collection program d and the bad block management program e for the control and management of the NAND flash memory. The error check and correction module is mainly used to check Figure 3 The error correction code type commonly used is Hamming Code. Figure 4 The flash memory dedicated error check and correction technology module (g) in the back end is different from the above. It is specifically used for error detection and correction control of NAND flash memory. Commonly used error correction code types are Bose-Chaudhuri-Hocquenghem code (BCH) and Low Density Parity Check Code (LDPC). It has stronger error detection and correction capabilities. The firmware structure of SSD also includes flash memory controller h and flash memory channel i.

[0056] The present invention proposes a single particle effect test method for a large-capacity solid-state hard disk based on high-energy particles, such as Figure 1 As shown, the specific steps include:

[0057] The single event effect test method for large-capacity solid-state hard drives based on high-energy particles includes:

[0058] Conduct SSD user data pre-configuration experiments;

[0059] Based on the data preset experiment, the irradiation test is started to determine whether the electronic components in the SSD have single-particle lockout under the irradiation condition, and a determination result is determined, wherein the determination of whether the electronic components in the SSD have single-particle lockout under the irradiation condition is performed based on the BMC information of the test server;

[0060] According to the judgment result, a functional test is performed, a functional test result is obtained, and based on the functional test result, a silent data destruction test is performed in a post-irradiation test phase;

[0061] According to the silent data corruption test, recording fixed bit error data;

[0062] Based on the fixed bit error data, error characteristic quantification calculation is performed and the calculation result is output.

[0063] Specifically, in the presetting stage before irradiation, the SSD user data must be preset first, and the fio program is used to set the test file block size (block size) to 4KB or specify other sizes according to the test requirements. Since there is a wear leveling control module in the SSD, the data written by the user to the SSD will be randomly distributed to increase the service life. The data pattern written by the user is inconsistent with the data stored in the NAND flash memory in the actual SSD, so no specific data pattern is specified during the test process, and the CRC32 check code is written into each 4KB file block for verification. At the same time, considering that the data retention effect (Dataretention) in the NAND flash memory will cause the transistor threshold voltage distribution (Vt distribution) in the flash memory to shift to the erased state. Therefore, the user data preset operation should not be too early before the start of the test to avoid interference with the test results.

[0064] Further, based on the data preset experiment, starting the irradiation test includes:

[0065] The test server performs 4K Verify load-carrying tests on the SSDs to be tested in batches to obtain performance information and error information of the SSDs.

[0066] Specifically, after the test system is fully started during the irradiation phase, the irradiation test begins, and the test server performs a 4KVerify load-carrying test on the SSDs to be tested in batches, including a read and write test of 50%-50% data volume, to ensure that the SSD read and write operations have the same number of input / output operations per second (IOPS) and the amount of test data. The performance information and error information of the SSD are uploaded to the host computer and saved in CSV format. In this process, it is periodically determined whether the electronic components in the SSD have a single event latch-up (SEL) phenomenon under irradiation conditions. In order to prevent the high current event induced by the single event latch-up from damaging the test server or SSD, if the current of the running device increases abnormally, the server baseboard management controller (BMC) will immediately power off to protect the SSD and record the overcurrent event.

[0067] Further, it is determined whether the electronic components in the SSD have single event lockup under the irradiation condition, and the determination result includes:

[0068] If a single-particle lock occurs, power off and collect the BMC logs of the test server;

[0069] If single event latch-up does not occur, the functionality is tested.

[0070] Further, based on the functional test results, a silent data destruction test is performed in a post-irradiation test phase;

[0071] If the functional test fails, log data is collected, functional recovery and firmware rewriting are performed, and after success, a silent data destruction test is performed;

[0072] If the functional test is successful, the silent data destruction test is performed based on the collected error-related log information.

[0073] Specifically, the working status of the SSD is continuously monitored during irradiation. If an unexpected I / O load interruption or abnormal unloading of the SSD occurs, the test program will record key information such as the time of the failure and the amount of data written at that time.

[0074] For SSDs that have completed functional testing, the silent data destruction test program will be run during the post-irradiation test phase. At the same time, error-related log information will be collected. For SSDs that have unrecoverable failures after power-on, various log data will be collected and firmware reset and other operations will be performed to try to restore normal functions. If successful, the silent data destruction test program will continue.

[0075] Further, according to the silent data corruption test, the number of silent data corruption data blocks and fixed bit error data are recorded, including:

[0076] Use fio's verify function to perform a read test. If repeated errors occur at fixed addresses in the memory, the fixed bit error data is recorded.

[0077] Specifically, a silent data corruption test program is run on the SSD to be tested. The program uses the verify function of fio to perform a read test in 4KB file block size sequence. If the data in a 4KB file block fails the CRC32 check, the data block will be reported as a silent data corruption data block.

[0078] If repeated errors at fixed addresses are found in the dynamic random access memory during the silent data test phase, a fixed bit error record is recorded.

[0079]

[0080] Among them, σ SDC is the cross section of silent data destruction, N SDC N is the number of data blocks that failed fio verify after irradiation. Total is the total number of data blocks written in the data pre-setting experiment, F n is the high-energy particle flux in the test environment, t is the working time of the SSD from the time the high-energy particle switch is fully turned on to the time when a functional interruption occurs. If no functional interruption occurs, it is the actual working time of the SSD under irradiation conditions.

[0081] Furthermore, based on the fixed bit error data, error characteristic quantification calculation is performed, and the output calculation result includes:

[0082] The module-level functional interrupt is used to perform quantitative calculation of error characteristics of the controller, dynamic random access memory general error and fixed bit error data, and NAND flash memory, and the calculation results are statistically calculated and output.

[0083] Specifically, the present invention proposes to use a single particle function interruption cross section σ SEFI (Unit: cm 2 / device) is used as a reliability index parameter to characterize SSD under irradiation working environment, and its calculation formula is:

[0084]

[0085] Where N SEFI The number of functional interruptions that occurred. To prevent the potential cumulative effect of large-intensity irradiation, the SSD that experienced functional interruption will not be irradiated again. n is the high-energy particle flux in the test environment (unit: n / cm2 ·s), t is the working time of the SSD from the high-energy particle switch being fully turned on to the functional interruption occurring (unit: s). This parameter is used to reflect the functional interruption of the overall SSD system.

[0086] A method for characterizing errors of SSD internal devices induced by high-energy particle irradiation

[0087] In the radiation test, the main controller, dynamic random access memory and NAND flash memory of the SSD are simultaneously exposed to high-energy particle radiation. The present invention utilizes the data reported by the error checking and correcting modules in these devices to count their error cross sections.

[0088] Among them, the static random access memory inside the main controller and the external dynamic random access memory can be used SBU (Unit: cm 2 / bit) to calculate its error cross section, the calculation formula is:

[0089]

[0090] Where N SBU is the number of single-bit errors that occurred during irradiation, F n is the high-energy particle flux in the test environment (unit: n / cm 2 ·s), t is the working time of the SSD from the time the high-energy particle switch is fully turned on to the time when the function is interrupted (unit: s). C is the capacity of the memory (unit: bit).

[0091] Existing studies on bulk silicon static random access memory (SRAM) exposed to high-energy particles have shown that the high-energy particle displacement damage effect has little effect on its performance. There are few reports on the displacement damage effect of NAND flash memory. The functional failure of dynamic random access memory caused by the displacement damage effect is called the "weak bit effect". The phenomenon is as follows: in the ground irradiation experiment, even if the irradiation is stopped, the error count of the dynamic random access memory cell continues to increase. The principle is that the displacement damage causes the leakage current of the read-write transistor to increase, and the capacitor to discharge in advance, resulting in a significant decrease in the data retention time. If the data retention time is less than the self-refresh time (tref), the data in the DRAM cell will be lost in advance. The method for determining this phenomenon in the SSD system is: when performing a silent data destruction check on the SSD after irradiation, observe the error correction data reported by the SSD dynamic random access memory. If repeated error correction events occur in a certain unit, it is considered that a fixed bit error has occurred. And calculate the fixed bit error cross-section data according to (3).

[0092]

[0093] Among them, σ Stuck is the cross section where fixed position error occurs, N Stuckis the number of single-bit errors in the silent data corruption test after irradiation, F n is the high-energy particle flux in the test environment, t is the working time of the SSD from the time the high-energy particle switch is fully turned on to the time when the function is interrupted. If no function is interrupted, it is the actual working time of the SSD under irradiation conditions. C is the capacity of the memory.

[0094] The calculation method for NAND flash memory is different. According to the principle that data errors occur when it is exposed to high-energy particles, the exposure will cause the following: Figure 2 The transistor threshold voltage distribution shown in the figure has a tailing phenomenon to the left. The tailing phenomenon of the threshold voltage distribution will become more serious with the increase of the irradiation dose, and will eventually exceed the error correction capability of the error check and correction module, inducing uncorrectable read errors. Therefore, the present invention defines the dose at which the SSD first has an uncorrectable error under irradiation conditions as the dose threshold F th It is an important parameter for evaluating the irradiation performance of NAND flash memory in SSD.

[0095] The present invention also provides a high-energy particle-based single-particle effect test device for a large-capacity solid-state hard disk, comprising: a control center and an irradiation chamber connected via Ethernet;

[0096] The control center is used to control the irradiation chamber;

[0097] The irradiation chamber includes: a high-energy particle beam window, a module under test, a high-energy particle absorber, a test server, a PCIex16 cable, a SlimSAS retiming expansion card, an SFF-8654U.2 interface cable, and an SFF-8088SATA interface cable;

[0098] The high-energy particle beam window is used to isolate the high vacuum environment inside the accelerator beam pipeline from the external normal pressure environment, while allowing the accelerated particle beam to be safely and effectively led out of the accelerator;

[0099] The module under test is used to fix the SSD to be tested and ensure that the center point of the high-energy particle beam is located at the expected position;

[0100] The high-energy particle absorber is used to absorb the remaining high-energy particles to prevent interference with other sensitive devices in the test environment;

[0101] The test server is used to run the test scripts and test cases of the SSD, monitor the working status of the SSD and collect the error data that occurs;

[0102] The PCIe x16 cable is used to connect the test server PCIe interface and the SlimSAS retiming expansion card, so that the test server is as far away from the beam window as possible.

[0103] The SlimSAS retiming expansion card is used to convert PCIe signals into SlimSAS standard signals, and simultaneously perform retiming to compensate for link loss to meet the signal integrity requirements of the PCIe link.

[0104] The SFF-8645U.2 interface cable is used to connect the data of the SSD to be tested using the U.2 interface and the SlimSAS retiming expansion card to meet the needs of simultaneous irradiation testing of multiple SSDs.

[0105] The SFF-8088 SATA interface cable is used to directly connect the SSD to be tested using the SATA interface and the test server.

[0106] Further, the control center includes: a high-energy particle collimator controller, a high-energy particle switch and a monitoring computer;

[0107] The high-energy particle collimator controller is used to control the area and flux of high-energy particle irradiation;

[0108] The high-energy particle switch is used to control the high-energy particle irradiation time;

[0109] The monitoring computer is used to remotely log in to the test server through the SSH protocol to ensure that the test personnel can safely and real-time control and observe the test server in the irradiation room.

[0110] Specifically, we built Figure 5 Hardware test environment. A monitoring computer C is set up in the control center D, and connected to E through Ethernet. It uses the SSH protocol to communicate with the test server L located in the irradiation room M in full-duplex mode. The test server needs to be equipped with an SFF-8088 standard interface and a PCIe x16 interface that supports channel splitting for batch testing of SSDs. The test server is connected to the SATA interface SSD in the tested module G through the SFF-8088SATA interface cable N. At the same time, the test server is connected to the SlimSAS retiming expansion card through the PCIe x16 cable O, and then converted to the SFF-8654U.2 interface cable K to connect to the U.2 interface SSD in the tested module G. In order to prevent the scattered high-energy particles from affecting the electronic system outside the tested module, the above-mentioned cable must be used to connect the SSD. At the same time, it is necessary to consider that the high-frequency signal of the SSD transmission line is sensitive to the transmission channel conditions. The parasitic capacitance and parasitic inductance in the cable will affect the transmission of high-frequency signals. Therefore, the length design and material design of the cable need to pay extra attention to prevent the induction of errors in the transmission link and affect the accuracy of the test results. This ensures that the module under test is as far away from the test server as possible while the link is working normally.

[0111] The module under test consists of an SSD, a fixing fixture, and a cooling system to ensure normal working conditions of the SSD.

[0112] The irradiation position calibration needs to be performed between the high-energy particle beam window F and the test module to ensure that the high-energy particle beam can cover all devices or some specific devices in the module under test. After the incident high-energy particles J pass through the module under test, the remaining high-energy particles I are absorbed by the particle absorber H to prevent pollution of the test environment and adverse effects on the test server. The high-energy particle beam is controlled by the high-energy particle collimator A controller in the control center to adjust parameters such as high-energy particle flux and irradiation area. The high-energy particle switch B is used to control the high-energy particle irradiation time to facilitate precise control of the high-energy particle injection.

[0113] The software structure diagram of high energy particle irradiation test is as follows: Figure 6 As shown in the figure, a shell test program is written. The program structure is as follows Figure 6 As shown in the figure, the program is divided into a rate data statistics module ①. During the test, the performance data of each SSD running the test load is recorded once per second through the iostat④ command tool in the Linux general command. The smartctl⑤ and telemetry⑥ commands capture the diagnostic data in the SSD, such as the Self-Monitoring Analysis and Report Technology (SMART) data, and display the error type and number of errors that occur in the current SSD. If a functional interruption occurs, the dmesg command tool ⑦ saves the system-level error information and reports it to the I / O interruption alarm module ②. These functions rely on the Linux I / O disk driver ⑧ to implement. The SSD load control program ③ can use the multiple physical central processor NUMA nodes of the test server to allocate CPU resources for the test task load, increase the number of SSDs tested in a single irradiation process, and improve the test efficiency. When executing the test task, the information feedback from the iostat④ command tool will be used to determine whether the SSD has a functional interruption.

[0114] This implementation case provides some experimental results of high-energy particle irradiation obtained by using the test device proposed by the present invention:

[0115] According to the above test system and test method, an irradiation test was carried out, using Φ50mm white light neutrons as the above high-energy particles for irradiation, with a flux of 1.1×10 7 n / cm 2Table 1 shows the cross-sectional results of the SSD functional interruption test. It can be seen that the cross-sectional results of single-particle functional interruption vary significantly between different controller processes and NAND Flash types. For example, there are large cross-sectional differences between samples VIII and IX and the rest of the samples. This difference reflects the reliability level of SSDs in irradiation environments with different process technologies and error correction capabilities.

[0116] Table 1

[0117]

[0118] Table 2 shows the error data of the static random access memory inside the SSD controller obtained in the experiment. It can be seen that this test method has good consistency in different SSDs.

[0119] Table 2

[0120]

[0121] Table 3 shows the error data of the dynamic random access memory inside the SSD controller obtained in the experiment, which shows that the test method has good consistency in different SSDs and has significant differences from the static random access memory inside the SSD controller.

[0122] Table 3

[0123]

[0124] Figure 7 The figure shows the test results of read errors in NAND flash memory. It can be seen that uncorrectable read errors appear in large numbers after a certain point in time. After calculation, the F th The parameter is 4.48×10 9 n / cm 2 .

[0125] The above results show that the technical solution of the present invention can effectively quantify the key components in SSD, and the differences are small in the same hardware and software SSD architecture. This parameter can be used to evaluate the radiation resistance index of sensitive components inside SSD, verifying the feasibility of this method.

[0126] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A single-particle effect test method for a large-capacity solid-state hard disk based on high-energy particles, characterized in that: include: Perform an SSD user data pre-set experiment, send data packets of the specified size to the SSD data to be tested and add verification information to each data packet; Based on the SSD user data preset experiment, start the irradiation test to determine whether the electronic components in the SSD have functional interruption and single event lockup under irradiation conditions, and determine the judgment result; According to the judgment result, a functional test is performed to obtain a functional test result, and based on the functional test result, a silent data destruction test and a fixed bit error test are performed in a post-irradiation test phase; According to the silent data corruption test and the fixed bit error test, recording the number of silent data corruption data blocks and the fixed bit error data; Based on the number of silent data corrupted data blocks and fixed bit error data, error characteristic quantification calculation is performed and the calculation result is output.

2. The high-energy particle based single-particle effect test method for large-capacity solid-state hard disks according to claim 1, characterized in that: Presetting the experiment based on the data and starting the irradiation test include: In a high-energy particle irradiation environment, the test server performs 4KVerify mixed random read and write tests on the SSDs to be tested in batches to obtain performance information and error information of the SSDs.

3. The single-event effect testing method for a large-capacity solid-state hard disk based on high-energy particles according to claim 1 is characterized in that: Determine whether the electronic components in the SSD have functional interruption and single event lockup under irradiation conditions, and determine the judgment results including: If a single-particle lock occurs, power off and collect the BMC overcurrent warning log of the test server; If single event locking does not occur, the functional test is performed.

4. The high-energy particle based single-particle effect test method for a large-capacity solid-state hard disk according to claim 3, characterized in that: Based on the functional test results, the silent data corruption test and the fixed bit error test are performed in the post-irradiation test phase, including: If the functional test fails, log data is collected, functional recovery and firmware rewriting are performed, and after success, silent data destruction testing is performed; If the functional test is successful, the silent data corruption test and the fixed bit error test are performed according to the collected error log information.

5. The high-energy particle based single-event effect testing method for large-capacity solid-state hard disks according to claim 1, characterized in that: According to the silent data corruption test and the fixed bit error test, recording the number of silent data corruption data blocks and the fixed bit error data includes: Use fio's verify function to perform a read test. If a read data error occurs and the controller does not report the error, it is considered that a silent data corruption error has occurred, and the number of silent data corruption data blocks is recorded. If repeated error correction records at fixed physical addresses are found in the dynamic random access memory during the test, fixed bit error data is recorded.

6. The high-energy particle based single-event effect testing method for large-capacity solid-state hard disks according to claim 1, characterized in that: Based on the number of silent data corrupted data blocks and fixed bit error data, error characteristic quantification calculation is performed, and the output calculation results include: The error data generated by the controller and the dynamic random access memory and the fixed bit error data are used to perform quantitative calculation of error characteristics, and the calculation results are statistically calculated and outputted.

Images