From e96068511528e8d7a4362a4a1cc306c8f11f5c7c Mon Sep 17 00:00:00 2001 From: Krzysztof Majzerowicz-Jaszcz Date: Wed, 18 Jan 2023 16:10:22 +0100 Subject: [PATCH] Preliminary seq-cutoff documentation Signed-off-by: Krzysztof Majzerowicz-Jaszcz --- _data/sidebars/opencas_sidebar.yml | 4 +++ pages/design_documentation/seq_cutoff.md | 32 ++++++++++++++++++++++++ 2 files changed, 36 insertions(+) create mode 100644 pages/design_documentation/seq_cutoff.md diff --git a/_data/sidebars/opencas_sidebar.yml b/_data/sidebars/opencas_sidebar.yml index 11c0969..f3d13ae 100644 --- a/_data/sidebars/opencas_sidebar.yml +++ b/_data/sidebars/opencas_sidebar.yml @@ -73,6 +73,10 @@ entries: url: /cleaner.html output: web + - title: Sequential Cutoff + url: /seq_cutoff.html + output: web + - title: Getting Started output: web folderitems: diff --git a/pages/design_documentation/seq_cutoff.md b/pages/design_documentation/seq_cutoff.md new file mode 100644 index 0000000..aeacddb --- /dev/null +++ b/pages/design_documentation/seq_cutoff.md @@ -0,0 +1,32 @@ +--- +title: Sequential Cutoff +last_updated: January, 2023 +toc: false +permalink: seq_cutoff.html +--- + +In some usage scenarios, it is desirable not to cache sequential workloads. This +kind of workloads is handled quite well by the storage devices, both traditional and +solid state. More often than not, sequential data are cold (not subsequently used) +as well as not being critical to the system performance, e.g. maintenance processes. + +Open CAS can be configured not to cache sequential workloads by using a feature +called Sequential Cutoff. + +# Sequential Cutoff - general description +Sequential Cutoff allows the user to specify the following parameters + * **cache-id** - identifier of a cache instance. <1-16384> + * **core-id** - identifier of a core instance. If not specified, Sequential Cutoff parameters +are set for all cores in a cache instance. <0-4095> + * **seq-threshold** - the amount of data after which an I/O request is handled by [**Pass Through cache mode**](/cache_configuration#pass-through). +[KiB] + * **seq-policy** - sequential cutoff policy to be used with a given core. Possible options are + * **always** - always bypass the cache if the amount of data is over threshold + * **full** - only bypass the cache if the cache is full and the amount of data is over threshold (default), + * **never** - disable Sequential Cutoff entirely + + + + + +