8.7 Ultrareliable Storage
If no additional hardware support for storing the recovery point data is available on
the target platform and cannot be added, the system has to fall back to a
software-implemented stable storage, i.e., the data are not copied by the hardware to
the stable storage, but the process is implemented in software and the data copied
by the processor.
The stable storage is also exposed to external radiation and must therefore
feature-built fault detection and correction mechanism.
A software implementation needs state variables to track the progress of the
operation. In case of a full system backup, the state of the software recovery point
mechanism is also stored on stable storage. In case of a recovery, the full state is
overwritten; therefore, it’s necessary to keep an inventory of the stored recovery
points on stable storage.
If only part of the system is included in the recovery point, the state might also
be kept in memory in addition to the inventory on stable storage.
However, code and data must be protected by some validation mechanism
(checksum) to guarantee the correctness of the code and data in memory. Corrupted
code can be reloaded from ROM, corrupted data restored from the stable storage
inventory.
For more details about the recovery process, refer to Chap. 9.
Here, we present the following requirements for software-based stable storage:
– Simple design,
– Very fast data storage and retrieval,
– Optimized for linear access,
– Transparent device recovery in case of faulty read or write operation,
– Small memory footprint, and
– Low complexity controller.
8.7.1 Physical Storage Media
We consider five options for the storage of recovery point information and discuss
shortly their advantages and disadvantages.
Hard disks are often used for stable storage in commercial systems. They are
cheap, provide huge capacity, and their linear read and write performance is relatively fast. Their disadvantage is, however, that they are quite big, power hungry,
need a complex and fast controller on the host side (S-ATA), and are prone to
vibration and shock which disqualifies them for use in an embedded system.
Solid-State Disk Solid-state disks are designed as plug-in replacements for hard
disks and use the (S-ATA) protocol with its already explained disadvantage. In
contrast to hard disks, solid-state disks use an array of flash chips to store that actual
8.7 Ultrareliable Storage
131
If no additional hardware support for storing the recovery point data is available on
the target platform and cannot be added, the system has to fall back to a
software-implemented stable storage, i.e., the data are not copied by the hardware to
the stable storage, but the process is implemented in software and the data copied
by the processor.
The stable storage is also exposed to external radiation and must therefore
feature-built fault detection and correction mechanism.
A software implementation needs state variables to track the progress of the
operation. In case of a full system backup, the state of the software recovery point
mechanism is also stored on stable storage. In case of a recovery, the full state is
overwritten; therefore, it’s necessary to keep an inventory of the stored recovery
points on stable storage.
If only part of the system is included in the recovery point, the state might also
be kept in memory in addition to the inventory on stable storage.
However, code and data must be protected by some validation mechanism
(checksum) to guarantee the correctness of the code and data in memory. Corrupted
code can be reloaded from ROM, corrupted data restored from the stable storage
inventory.
For more details about the recovery process, refer to Chap. 9.
Here, we present the following requirements for software-based stable storage:
– Simple design,
– Very fast data storage and retrieval,
– Optimized for linear access,
– Transparent device recovery in case of faulty read or write operation,
– Small memory footprint, and
– Low complexity controller.
8.7.1 Physical Storage Media
We consider five options for the storage of recovery point information and discuss
shortly their advantages and disadvantages.
Hard disks are often used for stable storage in commercial systems. They are
cheap, provide huge capacity, and their linear read and write performance is relatively fast. Their disadvantage is, however, that they are quite big, power hungry,
need a complex and fast controller on the host side (S-ATA), and are prone to
vibration and shock which disqualifies them for use in an embedded system.
Solid-State Disk Solid-state disks are designed as plug-in replacements for hard
disks and use the (S-ATA) protocol with its already explained disadvantage. In
contrast to hard disks, solid-state disks use an array of flash chips to store that actual
8.7 Ultrareliable Storage
131