data, resulting in extremely high read and write performance. In addition, they have
built-in fault handling of failing chips and automatic wear leveling (see next section), which is welcome in this context.
However, the high implementation complexity of SSD disks (complex controller
and firmware) which caused several manufacturers to release new firmware versions to eliminate bugs, sets a question mark behind the reliability of these disks.
Compact Flash Compact flashcards [58] are also flash based, exist for a long time,
and are often used in industry as direct replacement for hard disks as they use the
P-ATA protocol which is a predecessor of S-ATA. They are physically quite small
and they have built-in support for wear leveling.
SD Card Secure Digital Cards (SD Cards) are widely used in consumer electronics
such as photo cameras, smartphones, etc. They are available in different sizes, some
of them very small. Transfer rates, however, are quite low, currently available SD
card do not exceed write speeds of 20 Mbytes/second.
Flash Chips Flash chips are available in different technologies, sizes, and speeds.
Solid-state disks use flash chips that support the Open NAND Flash Interface [59,
60] (ONFI) specification. These are the currently fastest chips on the markets
featuring write speeds up to 300 Mbytes per flash chip. Recently, Intel flash chips
were measured with a write speed of 156 Mbytes/s, which is faster than a hard disk.
In contrast to the other options, flash chips are not removable. In the next section,
we discuss the properties and limitations of flash chip in more detail.
Table 8.2 shows a more detailed comparison of the different storage options we
consider. All options except hard disks use flash chips to store the content. The
S-ATA-based options, i.e., hard disks and S-ATA, require a high complexity,
high-speed controller (1.5 Ghz) on the host side which is not desirable in this
context. SD cards and compact flash have too low write speed as storing the
recovery points must be performed as fast as possible.
ONFI-based flash disks, which are also used in solid-state disks seem to be the
currently best option to store recovery points as they have large capacity, high write
speed, short access time, and a low complexity controller (100 MHz). In addition,
they need low power, i.e., the max power consumption is according to the ONFI
specification only 0.165 W. The downside, however, is that flash has write limitations that have to be taken into account when designing the stable storage
architecture.
8.7.2 Flash Properties and Limitations
Flash memory is organized in blocks (usually 512 bytes) and in erase units consisting of a number of adjacent blocks (usually 32 or 64) [61]. Reads and writes are
performed block-wise, and writes must be made only to previously erased blocks.
Blocks can either be erased automatically or manually, where the former option is
more comfortable while the latter is faster.
132
8 Recovery Preparation
built-in fault handling of failing chips and automatic wear leveling (see next section), which is welcome in this context.
However, the high implementation complexity of SSD disks (complex controller
and firmware) which caused several manufacturers to release new firmware versions to eliminate bugs, sets a question mark behind the reliability of these disks.
Compact Flash Compact flashcards [58] are also flash based, exist for a long time,
and are often used in industry as direct replacement for hard disks as they use the
P-ATA protocol which is a predecessor of S-ATA. They are physically quite small
and they have built-in support for wear leveling.
SD Card Secure Digital Cards (SD Cards) are widely used in consumer electronics
such as photo cameras, smartphones, etc. They are available in different sizes, some
of them very small. Transfer rates, however, are quite low, currently available SD
card do not exceed write speeds of 20 Mbytes/second.
Flash Chips Flash chips are available in different technologies, sizes, and speeds.
Solid-state disks use flash chips that support the Open NAND Flash Interface [59,
60] (ONFI) specification. These are the currently fastest chips on the markets
featuring write speeds up to 300 Mbytes per flash chip. Recently, Intel flash chips
were measured with a write speed of 156 Mbytes/s, which is faster than a hard disk.
In contrast to the other options, flash chips are not removable. In the next section,
we discuss the properties and limitations of flash chip in more detail.
Table 8.2 shows a more detailed comparison of the different storage options we
consider. All options except hard disks use flash chips to store the content. The
S-ATA-based options, i.e., hard disks and S-ATA, require a high complexity,
high-speed controller (1.5 Ghz) on the host side which is not desirable in this
context. SD cards and compact flash have too low write speed as storing the
recovery points must be performed as fast as possible.
ONFI-based flash disks, which are also used in solid-state disks seem to be the
currently best option to store recovery points as they have large capacity, high write
speed, short access time, and a low complexity controller (100 MHz). In addition,
they need low power, i.e., the max power consumption is according to the ONFI
specification only 0.165 W. The downside, however, is that flash has write limitations that have to be taken into account when designing the stable storage
architecture.
8.7.2 Flash Properties and Limitations
Flash memory is organized in blocks (usually 512 bytes) and in erase units consisting of a number of adjacent blocks (usually 32 or 64) [61]. Reads and writes are
performed block-wise, and writes must be made only to previously erased blocks.
Blocks can either be erased automatically or manually, where the former option is
more comfortable while the latter is faster.
132
8 Recovery Preparation
