“MEMS is also a micro-electromechanical system. We have a more or less understanding of MEMS systems. In order to improve everyone’s understanding of MEMS, this article will introduce the fault management of MEMS storage devices.
MEMS is also a micro-electromechanical system. We have a more or less understanding of MEMS systems. In order to improve everyone’s understanding of MEMS, this article will introduce the fault management of MEMS storage devices.
The full name of MEMS is Micro Electromechanical System. It refers to a high-tech device with a size of a few millimeters or even smaller, and its internal structure is generally on the order of micrometers or even nanometers, which is an independent intelligent system. It is mainly composed of sensors, actuators (actuators) and micro energy. MEMS memory is a new type of storage device, which has the characteristics of high density, low power consumption, non-volatile, multi-probe parallel access, etc., and has obvious advantages over traditional magnetic disks. It can fill the performance gap between RAM and disk, and can take on multiple roles in computer systems, bringing new ideas and new methods to the research on the structure of new high-performance mass storage systems. MEMS is an independent intelligent system that can be mass-produced. Its system size is a few millimeters or even smaller, and its internal structure is generally on the order of micrometers or even nanometers. In the following content, the editor will explain the fault management of MEMS storage devices based on several aspects, as follows.
1. Internal failure
There are two common failures of disks: recoverable failures and unrecoverable failures. MEMS storage devices will also experience similar failures. However, MEMS storage devices can use multiple probes to compensate for component failures, including failures that may cause the device to become unusable.
For MEMS storage devices, effective error correction codes can be calculated from data distributed on multiple probes. In the G2 model, each 512-byte data block and ECC code are distributed among 64 probes. The Ecc code includes a vertical part and a horizontal part. The horizontal part of the ECC code can be recovered from the faulty sector, while the vertical part indicates which sectors can be treated as faulty sectors, and at the same time converts large errors into sector erasure operations. This simple mechanism shows that most internal failures can be recovered.
Like disks, MEMS storage devices also reserve some spare space for storing data that cannot be saved in the default location due to probe and media failure. Multiple probes of the MEMS storage device can access data in parallel on a track, which can avoid the performance and prediction overhead caused by remapping due to failures. Moreover, by setting one or more spare tips on each track, unreadable data is remapped to the same sector as the idle probe.
2. Equipment failure
MEMS storage devices are also susceptible to unrecoverable failures: external mechanical or electrostatic forces can damage the collector brush or break the spring of the actuator, damage the surface of the medium, damage the Electronic device of the device, or destroy the data channel. If these failures occur, they can be dealt with in the same way as disks. For example, use internal redundancy and periodic backup to handle equipment failures.
The mechanical characteristics of MEMS storage devices are more suitable for handling read-modify-write operations in some fault-tolerant mechanisms. Generally, a disk needs a complete revolution to reach the same sector, and MEMS storage devices can quickly reverse the direction, which greatly reduces the delay of read-update-write.
Three, failure recovery
Like disks, file systems and database systems need to maintain the consistency of objects stored on MEMS storage devices. Although the use of synchronous write operations has a certain impact on performance, the low service time of MEMS storage devices can reduce this loss. In addition, the MEMS storage device does not have the time for the shaft to start, so the device starts up fast, about 0.Sms. Even for high-end disks, it takes 15-25s to start the spindle and complete the initialization. At the same time, because there is no need to start the shaft, there is no need to consider the power consumption required to start the shaft, and there is no need to adopt any technology to reduce power consumption. These all make all MEMS storage devices can be started at the same time, and the system startup time is from the second level. Down to the millisecond level.