€•SYŒsphinx.addnodes”Œdocument”“”)”}”(Œ rawsource”Œ”Œchildren”]”(Œ translations”Œ LanguagesNode”“”)”}”(hhh]”(hŒ pending_xref”“”)”}”(hhh]”Œdocutils.nodes”ŒText”“”ŒChinese (Simplified)”…””}”Œparent”hsbaŒ attributes”}”(Œids”]”Œclasses”]”Œnames”]”Œdupnames”]”Œbackrefs”]”Œ refdomain”Œstd”Œreftype”Œdoc”Œ reftarget”Œ"/translations/zh_CN/accounting/psi”Œmodname”NŒ classname”NŒ refexplicit”ˆuŒtagname”hhh ubh)”}”(hhh]”hŒChinese (Traditional)”…””}”hh2sbah}”(h]”h ]”h"]”h$]”h&]”Œ refdomain”h)Œreftype”h+Œ reftarget”Œ"/translations/zh_TW/accounting/psi”Œmodname”NŒ classname”NŒ refexplicit”ˆuh1hhh ubh)”}”(hhh]”hŒItalian”…””}”hhFsbah}”(h]”h ]”h"]”h$]”h&]”Œ refdomain”h)Œreftype”h+Œ reftarget”Œ"/translations/it_IT/accounting/psi”Œmodname”NŒ classname”NŒ refexplicit”ˆuh1hhh ubh)”}”(hhh]”hŒJapanese”…””}”hhZsbah}”(h]”h ]”h"]”h$]”h&]”Œ refdomain”h)Œreftype”h+Œ reftarget”Œ"/translations/ja_JP/accounting/psi”Œmodname”NŒ classname”NŒ refexplicit”ˆuh1hhh ubh)”}”(hhh]”hŒKorean”…””}”hhnsbah}”(h]”h ]”h"]”h$]”h&]”Œ refdomain”h)Œreftype”h+Œ reftarget”Œ"/translations/ko_KR/accounting/psi”Œmodname”NŒ classname”NŒ refexplicit”ˆuh1hhh ubh)”}”(hhh]”hŒSpanish”…””}”hh‚sbah}”(h]”h ]”h"]”h$]”h&]”Œ refdomain”h)Œreftype”h+Œ reftarget”Œ"/translations/sp_SP/accounting/psi”Œmodname”NŒ classname”NŒ refexplicit”ˆuh1hhh ubeh}”(h]”h ]”h"]”h$]”h&]”Œcurrent_language”ŒEnglish”uh1h hhŒ _document”hŒsource”NŒline”NubhŒtarget”“”)”}”(hŒ.. _psi:”h]”h}”(h]”h ]”h"]”h$]”h&]”Œrefid”Œpsi”uh1h¡h KhhhžhhŸŒ ”h]”hæ)”}”(hŒ$Johannes Weiner ”h]”(hŒJohannes Weiner <”…””}”(hjhžhhŸNh NubhŒ reference”“”)”}”(hŒhannes@cmpxchg.org”h]”hŒhannes@cmpxchg.org”…””}”(hjhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”Œrefuri”Œmailto:hannes@cmpxchg.org”uh1jhjubhŒ>”…””}”(hjhžhhŸNh Nubeh}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h Khjubah}”(h]”h ]”h"]”h$]”h&]”uh1hßhjubeh}”(h]”h ]”h"]”h$]”h&]”uh1hÊhŸh¯h KhhÇhžhubeh}”(h]”h ]”h"]”h$]”h&]”uh1hÅhh²hžhhŸh¯h Kubhæ)”}”(hŒ„When CPU, memory or IO devices are contended, workloads experience latency spikes, throughput losses, and run the risk of OOM kills.”h]”hŒ„When CPU, memory or IO devices are contended, workloads experience latency spikes, throughput losses, and run the risk of OOM kills.”…””}”(hjKhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h K hh²hžhubhæ)”}”(hŒßWithout an accurate measure of such contention, users are forced to either play it safe and under-utilize their hardware resources, or roll the dice and frequently suffer the disruptions resulting from excessive overcommit.”h]”hŒßWithout an accurate measure of such contention, users are forced to either play it safe and under-utilize their hardware resources, or roll the dice and frequently suffer the disruptions resulting from excessive overcommit.”…””}”(hjYhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h K hh²hžhubhæ)”}”(hŒ¢The psi feature identifies and quantifies the disruptions caused by such resource crunches and the time impact it has on complex workloads or even entire systems.”h]”hŒ¢The psi feature identifies and quantifies the disruptions caused by such resource crunches and the time impact it has on complex workloads or even entire systems.”…””}”(hjghžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h Khh²hžhubhæ)”}”(hŒ°Having an accurate measure of productivity losses caused by resource scarcity aids users in sizing workloads to hardware--or provisioning hardware according to workload demand.”h]”hŒ°Having an accurate measure of productivity losses caused by resource scarcity aids users in sizing workloads to hardware--or provisioning hardware according to workload demand.”…””}”(hjuhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h Khh²hžhubhæ)”}”(hŒ÷As psi aggregates this information in realtime, systems can be managed dynamically using techniques such as load shedding, migrating jobs to other systems or data centers, or strategically pausing or killing low priority or restartable batch jobs.”h]”hŒ÷As psi aggregates this information in realtime, systems can be managed dynamically using techniques such as load shedding, migrating jobs to other systems or data centers, or strategically pausing or killing low priority or restartable batch jobs.”…””}”(hjƒhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h Khh²hžhubhæ)”}”(hŒThis allows maximizing hardware utilization without sacrificing workload health or risking major disruptions such as OOM kills.”h]”hŒThis allows maximizing hardware utilization without sacrificing workload health or risking major disruptions such as OOM kills.”…””}”(hj‘hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h Khh²hžhubh±)”}”(hhh]”(h¶)”}”(hŒPressure interface”h]”hŒPressure interface”…””}”(hj¢hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1hµhjŸhžhhŸh¯h K#ubhæ)”}”(hŒyPressure information for each resource is exported through the respective file in /proc/pressure/ -- cpu, memory, and io.”h]”hŒyPressure information for each resource is exported through the respective file in /proc/pressure/ -- cpu, memory, and io.”…””}”(hj°hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h K%hjŸhžhubhæ)”}”(hŒThe format is as such::”h]”hŒThe format is as such:”…””}”(hj¾hžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h K(hjŸhžhubhŒ literal_block”“”)”}”(hŒ]some avg10=0.00 avg60=0.00 avg300=0.00 total=0 full avg10=0.00 avg60=0.00 avg300=0.00 total=0”h]”hŒ]some avg10=0.00 avg60=0.00 avg300=0.00 total=0 full avg10=0.00 avg60=0.00 avg300=0.00 total=0”…””}”hjÎsbah}”(h]”h ]”h"]”h$]”h&]”Œ xml:space”Œpreserve”uh1jÌhŸh¯h K*hjŸhžhubhæ)”}”(hŒiThe "some" line indicates the share of time in which at least some tasks are stalled on a given resource.”h]”hŒmThe “some†line indicates the share of time in which at least some tasks are stalled on a given resource.”…””}”(hjÞhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h K-hjŸhžhubhæ)”}”(hXThe "full" line indicates the share of time in which all non-idle tasks are stalled on a given resource simultaneously. In this state actual CPU cycles are going to waste, and a workload that spends extended time in this state is considered to be thrashing. This has severe impact on performance, and it's useful to distinguish this situation from a state where some tasks are stalled but the CPU is still doing productive work. As such, time spent in this subset of the stall state is tracked separately and exported in the "full" averages.”h]”hX'The “full†line indicates the share of time in which all non-idle tasks are stalled on a given resource simultaneously. In this state actual CPU cycles are going to waste, and a workload that spends extended time in this state is considered to be thrashing. This has severe impact on performance, and it’s useful to distinguish this situation from a state where some tasks are stalled but the CPU is still doing productive work. As such, time spent in this subset of the stall state is tracked separately and exported in the “full†averages.”…””}”(hjìhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h K0hjŸhžhubhæ)”}”(hŒ}CPU full is undefined at the system level, but has been reported since 5.13, so it is set to zero for backward compatibility.”h]”hŒ}CPU full is undefined at the system level, but has been reported since 5.13, so it is set to zero for backward compatibility.”…””}”(hjúhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h K9hjŸhžhubhæ)”}”(hX‹The ratios (in %) are tracked as recent trends over ten, sixty, and three hundred second windows, which gives insight into short term events as well as medium and long term trends. The total absolute stall time (in us) is tracked and exported as well, to allow detection of latency spikes which wouldn't necessarily make a dent in the time averages, or to average trends over custom time frames.”h]”hXThe ratios (in %) are tracked as recent trends over ten, sixty, and three hundred second windows, which gives insight into short term events as well as medium and long term trends. The total absolute stall time (in us) is tracked and exported as well, to allow detection of latency spikes which wouldn’t necessarily make a dent in the time averages, or to average trends over custom time frames.”…””}”(hjhžhhŸNh Nubah}”(h]”h ]”h"]”h$]”h&]”uh1håhŸh¯h K