zswap.rst 7.2 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152
  1. .. _zswap:
  2. =====
  3. zswap
  4. =====
  5. Overview
  6. ========
  7. Zswap is a lightweight compressed cache for swap pages. It takes pages that are
  8. in the process of being swapped out and attempts to compress them into a
  9. dynamically allocated RAM-based memory pool. zswap basically trades CPU cycles
  10. for potentially reduced swap I/O.  This trade-off can also result in a
  11. significant performance improvement if reads from the compressed cache are
  12. faster than reads from a swap device.
  13. .. note::
  14. Zswap is a new feature as of v3.11 and interacts heavily with memory
  15. reclaim. This interaction has not been fully explored on the large set of
  16. potential configurations and workloads that exist. For this reason, zswap
  17. is a work in progress and should be considered experimental.
  18. Some potential benefits:
  19. * Desktop/laptop users with limited RAM capacities can mitigate the
  20. performance impact of swapping.
  21. * Overcommitted guests that share a common I/O resource can
  22. dramatically reduce their swap I/O pressure, avoiding heavy handed I/O
  23. throttling by the hypervisor. This allows more work to get done with less
  24. impact to the guest workload and guests sharing the I/O subsystem
  25. * Users with SSDs as swap devices can extend the life of the device by
  26. drastically reducing life-shortening writes.
  27. Zswap evicts pages from compressed cache on an LRU basis to the backing swap
  28. device when the compressed pool reaches its size limit. This requirement had
  29. been identified in prior community discussions.
  30. Whether Zswap is enabled at the boot time depends on whether
  31. the ``CONFIG_ZSWAP_DEFAULT_ON`` Kconfig option is enabled or not.
  32. This setting can then be overridden by providing the kernel command line
  33. ``zswap.enabled=`` option, for example ``zswap.enabled=0``.
  34. Zswap can also be enabled and disabled at runtime using the sysfs interface.
  35. An example command to enable zswap at runtime, assuming sysfs is mounted
  36. at ``/sys``, is::
  37. echo 1 > /sys/module/zswap/parameters/enabled
  38. When zswap is disabled at runtime it will stop storing pages that are
  39. being swapped out. However, it will _not_ immediately write out or fault
  40. back into memory all of the pages stored in the compressed pool. The
  41. pages stored in zswap will remain in the compressed pool until they are
  42. either invalidated or faulted back into memory. In order to force all
  43. pages out of the compressed pool, a swapoff on the swap device(s) will
  44. fault back into memory all swapped out pages, including those in the
  45. compressed pool.
  46. Design
  47. ======
  48. Zswap receives pages for compression through the Frontswap API and is able to
  49. evict pages from its own compressed pool on an LRU basis and write them back to
  50. the backing swap device in the case that the compressed pool is full.
  51. Zswap makes use of zpool for the managing the compressed memory pool. Each
  52. allocation in zpool is not directly accessible by address. Rather, a handle is
  53. returned by the allocation routine and that handle must be mapped before being
  54. accessed. The compressed memory pool grows on demand and shrinks as compressed
  55. pages are freed. The pool is not preallocated. By default, a zpool
  56. of type selected in ``CONFIG_ZSWAP_ZPOOL_DEFAULT`` Kconfig option is created,
  57. but it can be overridden at boot time by setting the ``zpool`` attribute,
  58. e.g. ``zswap.zpool=zbud``. It can also be changed at runtime using the sysfs
  59. ``zpool`` attribute, e.g.::
  60. echo zbud > /sys/module/zswap/parameters/zpool
  61. The zbud type zpool allocates exactly 1 page to store 2 compressed pages, which
  62. means the compression ratio will always be 2:1 or worse (because of half-full
  63. zbud pages). The zsmalloc type zpool has a more complex compressed page
  64. storage method, and it can achieve greater storage densities. However,
  65. zsmalloc does not implement compressed page eviction, so once zswap fills it
  66. cannot evict the oldest page, it can only reject new pages.
  67. When a swap page is passed from frontswap to zswap, zswap maintains a mapping
  68. of the swap entry, a combination of the swap type and swap offset, to the zpool
  69. handle that references that compressed swap page. This mapping is achieved
  70. with a red-black tree per swap type. The swap offset is the search key for the
  71. tree nodes.
  72. During a page fault on a PTE that is a swap entry, frontswap calls the zswap
  73. load function to decompress the page into the page allocated by the page fault
  74. handler.
  75. Once there are no PTEs referencing a swap page stored in zswap (i.e. the count
  76. in the swap_map goes to 0) the swap code calls the zswap invalidate function,
  77. via frontswap, to free the compressed entry.
  78. Zswap seeks to be simple in its policies. Sysfs attributes allow for one user
  79. controlled policy:
  80. * max_pool_percent - The maximum percentage of memory that the compressed
  81. pool can occupy.
  82. The default compressor is selected in ``CONFIG_ZSWAP_COMPRESSOR_DEFAULT``
  83. Kconfig option, but it can be overridden at boot time by setting the
  84. ``compressor`` attribute, e.g. ``zswap.compressor=lzo``.
  85. It can also be changed at runtime using the sysfs "compressor"
  86. attribute, e.g.::
  87. echo lzo > /sys/module/zswap/parameters/compressor
  88. When the zpool and/or compressor parameter is changed at runtime, any existing
  89. compressed pages are not modified; they are left in their own zpool. When a
  90. request is made for a page in an old zpool, it is uncompressed using its
  91. original compressor. Once all pages are removed from an old zpool, the zpool
  92. and its compressor are freed.
  93. Some of the pages in zswap are same-value filled pages (i.e. contents of the
  94. page have same value or repetitive pattern). These pages include zero-filled
  95. pages and they are handled differently. During store operation, a page is
  96. checked if it is a same-value filled page before compressing it. If true, the
  97. compressed length of the page is set to zero and the pattern or same-filled
  98. value is stored.
  99. Same-value filled pages identification feature is enabled by default and can be
  100. disabled at boot time by setting the ``same_filled_pages_enabled`` attribute
  101. to 0, e.g. ``zswap.same_filled_pages_enabled=0``. It can also be enabled and
  102. disabled at runtime using the sysfs ``same_filled_pages_enabled``
  103. attribute, e.g.::
  104. echo 1 > /sys/module/zswap/parameters/same_filled_pages_enabled
  105. When zswap same-filled page identification is disabled at runtime, it will stop
  106. checking for the same-value filled pages during store operation. However, the
  107. existing pages which are marked as same-value filled pages remain stored
  108. unchanged in zswap until they are either loaded or invalidated.
  109. To prevent zswap from shrinking pool when zswap is full and there's a high
  110. pressure on swap (this will result in flipping pages in and out zswap pool
  111. without any real benefit but with a performance drop for the system), a
  112. special parameter has been introduced to implement a sort of hysteresis to
  113. refuse taking pages into zswap pool until it has sufficient space if the limit
  114. has been hit. To set the threshold at which zswap would start accepting pages
  115. again after it became full, use the sysfs ``accept_threshold_percent``
  116. attribute, e. g.::
  117. echo 80 > /sys/module/zswap/parameters/accept_threshold_percent
  118. Setting this parameter to 100 will disable the hysteresis.
  119. A debugfs interface is provided for various statistic about pool size, number
  120. of pages stored, same-value filled pages and various counters for the reasons
  121. pages are rejected.