Download as PDF, PPTX
Uploaded byLukas Pirl
Linux numa evolution
The document outlines notes and slides from a seminar given by Fredrik Teschke and Lukas Pirl on NUMA (Non-Uniform Memory Access) architecture at the Hasso Plattner Institute in Potsdam. The seminar covered topics including NUMA scheduling and load balancing challenges, memory allocation policies, and the Linux kernel APIs and mechanisms for managing NUMA systems. A number of links are provided to Linux kernel documentation and LWN articles discussing the history and ongoing development of NUMA support in Linux.
Linux numa evolution
- 1. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam
- 2. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 2
- 3. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://storage.pardot.com/6342/95370/lf_pub_top500report.pdf UNIX Linux 3
- 4. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 4
- 5. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam `diff -up` https://www.kernel.org/doc/Documentation/SubmittingPatches (20.11.2014) 5
- 6. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://thread.gmane.org/gmane.linux.kernel/1392753 6
- 7. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://upload.wikimedia.org/wikipedia/commons/e/e1/Linus_Torvalds,_2002,_Australian_Linux_conference.jpg …
- 8. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam https://www.kernel.org/doc/Documentation/development-process/2.Process, http://www.linuxfoundation.org/sites/main/files/publications/whowriteslinux.pdf 8
- 9. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam https://www.kernel.org/doc/Documentation/development-process/2.Process “ 9
- 10. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 10
- 11. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 11
- 12. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 12
- 13. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://en.wikipedia.org/wiki/Scheduling_%28computing%29 13
- 14. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/254445/ ? 14
- 15. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 15
- 16. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam ✔ ✘ https://www.cs.sfu.ca/~fedorova/papers/usenix-numa.pdf 16
- 17. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam ✔ ✔ ✘ https://www.cs.sfu.ca/~fedorova/papers/asplos284-dashti.pdf Traffic Imbalance 17
- 18. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam → 18
- 19. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 19
- 20. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … 20
- 21. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lse.sourceforge.net/numa/topology_api/in-kernel/ 21
- 22. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam asm/topology.h int __cpu_to_node(int cpu); int __memblk_to_node(int memblk); unsigned long __node_to_cpu_mask(int node); int __parent_node(int node); # /! supports hierarchies http://lse.sourceforge.net/numa/topology_api/in-kernel/ 22
- 23. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 23
- 24. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam int __cpu_to_node(int cpu); http://home.arcor.de/efocht/sched/ keep task & mem on same node 24
- 25. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam ⇒ L = local_node(); # regular load balancing as for multicore # (O(1) scheduler): balance_node(L); N = most_loaded_node(); C = most_loaded_cpu(N); if load(L) <= system_load() steal_tasks_from_cpu(C); http://home.arcor.de/efocht/sched/ 25
- 26. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 26
- 27. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/67005/ 27
- 28. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam → http://lwn.net/Articles/80911/ Node Physical CPU CPU0 CPU1 Physical CPU CPU2 CPU3 Node Node Node minimize cost of moving task (& mem) 28
- 29. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam node node node node 29
- 30. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … node node node node node node 30
- 31. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … node node node node node node 31
- 32. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … node node node node node node 32
- 33. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … 33
- 34. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam ? 34
- 35. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … mem node mem node 35
- 36. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … mem node mem node 36
- 37. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … mem node mem node 37
- 38. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … mem mem nodenode 38
- 39. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … mem node mem node 39
- 40. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … mem node mem node 40
- 41. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … mem node mem node 41
- 42. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam ⇒ http://lwn.net/Articles/486858/ 42
- 43. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/486858/ mem follows task 43
- 44. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/486858/ task follows mem 44
- 45. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/486858/ tasks w/ shared mem on same node 45
- 46. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/488709/ 46
- 47. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/488709/ memmem nodenode ⚡ ⚡ ⚡ 47
- 48. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam Unless you're going to listen to feedback I give you, I'm going to completely stop reading your patches, I don't give a rats arse you work for the same company anymore. You're impossible to work with. http://lwn.net/Articles/522093/ “ 48
- 49. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/522093/, http://lwn.net/Articles/524535/ 49
- 50. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 50
- 51. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/524977/, http://thread.gmane.org/gmane.linux.kernel/1392753 51
- 52. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 52
- 53. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam autonuma 53
- 54. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam ⚡ http://lwn.net/Articles/568870/ scheduling: NUMA and load balancing and groups 54
- 55. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://thread.gmane.org/gmane.linux.kernel/1631332 mem node mem node 55
- 56. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://thread.gmane.org/gmane.linux.kernel/1631332 mem node mem node 56
- 57. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://thread.gmane.org/gmane.linux.kernel/1631332 57
- 58. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 58
- 59. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 59
- 60. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://thread.gmane.org/gmane.linux.kernel/1808344 60
- 61. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam ⇔ ∃ http://thread.gmane.org/gmane.linux.kernel/1808344 node node node node 61
- 62. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam ⇔ http://thread.gmane.org/gmane.linux.kernel/1808344 backplane controller node backplane controller backplane controller backplane node node node node 62
- 63. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam … http://lwn.net/Articles/591995/ 63
- 64. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/591995/ 64
- 65. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/591995/ 65
- 66. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam http://lwn.net/Articles/591995/ node node mem NIC DMA? IO adapter mem 66
- 67. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 67
- 68. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam 68
- 69. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam
- 70. Fredrik Teschke, LukasPirl seminar on NUMA, Hasso Plattner Institue, Potsdam → ≈ → CPU CPU CPU CPU CPU node node node
Editor's Notes
- #2 today Linux has some understanding on how to handle non-uniform mem access (Tux gnawing on mem modules) get most out of hardware 10 years ago: very different picture what we want to show: where are we today and how did we get there how did Kernel evolve: making it easier for developers we got our information from lwn.net: linux weekly news -> articles, comments etc. lkml.org: linux kernel mailing list: lots of special sub-lists discussion of design/implementation of features include patches (source code) git.kernel.org find out what got merged when but for really old stuff that was not possible so also change logs of kernels before 2005
- #3 Why Linux anyways? isn’t Windows usually supported best? not for typical NUMA hardware
- #4 Linux market share is rising (Top 500) top 500 supercomputers (http://top500.org/) first Linux system: 1998 first basic NUMA support in Linux: 2002 from 2002: skyrocketed not economical to develop custom OS for every project no licensing cost! important if large cluster major vendors contribute
- #5 Linux is popular for NUMA systems hardware in supercomputing: very specific develop OS support prior to hardware release applications very specific fine tuning required OSS desired easily adapt knowledge base exists
- #6 kernel development process depicted design implement diff -up: list changes describe changes email to maintainer, CC mailing list discuss dotted arrow: Kernel Doc design often done without involving the community but better in the open if at all possible save a lot of time redesigning things later if there are review complaints: fix/redesign
- #7 development process example at top: see that this is a patch set each patch contains description of changes diff and then replies via email so basically: all a bunch of mails this just happens to be Linus favourite form of communication
- #8 step 7: send pull request to Linus … mostly Kernel Doc 2.6.38 kernel: only 1.3% patches were directly chosen by Linus but top-level maintainers ask Linus to pull the patches they selected getting patches into kernel depends on finding the right maintainer sending patches directly to Linus is not normally the right way to go chain of trust subsystem maintainer may trust others from whom he pulls changes into his tree
- #9 some other facts major release: every 2–3 months 2-week merge window at beginning of cycle linux-next tree as staging area git since 2005 before that: patch from email was applied manually made it difficult to stay up to date for developers and for us: a lot harder to track what got patched into mainstream kernel linux-kernel mailing list: 700 mails/day
- #10 paragraph taken from Kernel documentation on dev process There is [...] a somewhat involved (if somewhat informal) process designed to ensure that each patch is reviewed for quality and that each patch implements a change which is desirable to have in the mainline. This process can happen quickly for minor fixes, or, in the case of large and controversial changes, go on for years. recent NUMA efforts: lots of discussion
- #11 people short look at kernel hackers working on NUMA there are many more, just the most important early days: Paul McKenny (IBM) beginning of last decade nowadays Peter Zijlstra redhat, Intel sched Mel Gorman IBM, Suse mm Rik van Riel redhat mm/sched/virt finding pictures quite difficult - just regular guys work on kernel full-time for companies providing linux distributions also listed: parts of kernel the devs focus on mm: memory management sched: scheduling can see two core areas scheduling: which thread runs when and where and mem mgmt: where is mem allocated, paging both relevant for NUMA
- #12 now recap of some areas first: NUMA hardware this slide: very basic - you probably know it by heart left: UMA right: NUMA multiple memory controllers access times may differ (non-uniform) direct consequence: several interconnects
- #13 caution: terminology in the community Linux does some things different than others this influences terminology node: as in NUMA node highlighted area: one node != node (computer) in cluster may have several processors now three terms you have to be very careful with task, process and thread in Linux world: task is not a work package instead: scheduling entity that used to mean: task == process then threads came along Linux is different: processes and threads are pretty much the same threads are just configured to share resources pthreads_create() -> new task spawned via clone() we’ll just talk about tasks means both processes and threads --------------------- http://www.makelinux.net/books/lkd2/ch03lev1sec3 https://en.wikipedia.org/wiki/Native_POSIX_Thread_Library man pthreads Both of these are so-called 1:1 implementations, meaning that each thread maps to a kernel scheduling entity. Both threading implementations employ the Linux clone(2) system call.
- #14 recap: scheduling goals fairness each process gets its fair share no process can suffer indefinite postponement equal time != fair (safety control and payroll at a nuclear plant) load no idle times when there is work throughput maximize tasks/time latency time until first response/completion
- #15 recap: the problem when talking about NUMA still observe scheduling goals e.g. in supercomputing: high throughput in other presentations: already heard about possible approaches preserve memory locality: keep task close because takes longer to access remote memory two ways to do this: scheduling (task placement) vs. mm keep related tasks close: if they share memory avoid congestion of mem controllers, interconnects that would then be bottleneck for application few things you should keep in mind overhead: if we want to make more complex decisions have to arrive there somehow: probably also gathering data / calculating heuristics scheduling invoked very frequently: is it worth the overhead? short vs. long-running tasks applications where NUMA makes sense normally don’t run for 50ms short-running task: probably not worth rescheduling to different node also not worth overhead gathering statistics, and making decisions empirical observation that we found in multiple places shared memory tasks not always isolated share memory: global level (C lib) / task groups aka threads latter ideally placed on same node
- #16 kernel development and academic science how do the two mix? no references to academic work mails discussions articles instead: mailing list discussions serve as theoretical considerations we know such work exists (see Fabian Eckert’s presentation)
- #17 2001 DINO avoid NUMA-agnostic migrations thread placement scheduling: predefined thread classes based on cache misses / time keep classes on one node memory migration migrate a fixed number K of pages different strategies (pattern detection etc.) empirically determined K which seems optimal migrate memory too often interconnect stress migrate memory not often enough memory controller stress
- #18 some overlap in authors same basic assumption: remote access cost not the problem 2013 -> worked on Kernel 3.6 (released end of 2012) main concern: congestion of mem controller / interconnect mechanisms: page co-location, interleaving, replication thread clustering “So even without improving locality (we even reduce it for PCA), we are able to substantially improve performance”
- #19 if no patch submitted to kernel mailing list chances of receiving attention are low again: formal requirements are very high plain-text only no attachments only include text you are specifically replying to patches directly pasted into an email ignorance is high if violated
- #20 2002 → today gap 2006 – 2011 dating of changes where available: kernel release dates otherwise: date of main article referring to patch set kernel version: contains merged code = above timeline below the timeline = not merged into mainstream
- #21 no understanding of nodes unaware of memory locations/latencies no memory migration between nodes no affinity processing, memory allocations ⇒ performance of application may vary system load where is the process scheduled may be all allocations remote … basically, everything can happen! if system ends up unbalanced, no chance to fix this
- #22 rudimentary “discovery” of topology by McKenney, IBM obtained from firmware supposed to map to any kind of system elements processor (physical) memory block memory block: physically contiguous block of mem node node: container for any elements not necessarily 1-1 mapping to hardware does not represent attached hardware NIC IO controller … interconnects how to pin process close to hardware? manual? symbol at bottom right: this was merged into the Kernel!
- #23 brief API overview __cpu_to_node(int cpu); returns node the CPU belongs to __memblk_to_node(int memblk); returns node the memory belongs to __node_to_cpu_mask(int node); useful for pinning/affinity __parent_node(int node); supports hierarchies! no distances/latencies
- #24 manually discover nodes and their CPUs/RAM derive placement approach manually pin tasks to CPUs provoke less migrations over nodes
- #25 scheduler pools CPUs by node 1st time active consideration of nodes __cpu_to_node(int cpu); assigns static home node per task run & allocate memory here initial load balancing node with minimum number of tasks policies same node new node if own memory mgmt. always new node system might get unbalanced over time
- #26 dynamic load balancing invoked frequently per CPU idle CPUs: every tick loaded CPUs: every 200ms ⇒ “multi-level balance”: inside node across nodes
- #27 compute load probably balanced well still, main problem: memory spreads out CPU affinity might help “no return”
- #28 libnuma by Andi Kleen (Suse) syscalls library command-line utility mem alloc policies BIND set specific node PREFERRED prefers a specific node DEFAULT prefers current node INTERLEAVE only on nodes with decent-sized memory home node == “preferred” adds flexibility
- #29 levels hyperthreading share all caches cores have own caches node: own memory balancing intervals HT CPU: every 1-2ms even small differences physical CPU: less often rarely if whole system busy process loses cache affinity after few ms node: rarely longer cache affinity enhanced scheduling approach traverse hierarchy bottom → top at each level: balance groups? domain policy influences decision prefer balancing at lower level
- #30 distance between nodes obtainable from ACPI SLIT - System Locality Information Table apparently not used for node balancing à la “if another node required, take a closer one” why? track access patterns better? DINO, Carrefour highly app-specific? assumption same parent == same data might be wrong ex. Linux’ “init” process even though: knowing the distance is not enough
- #31 app needs threads on two nodes (concurrency > CPUs/node)
- #32 another app needs 4 nodes scheduled on idle nodes bad: 4-node load separated by 2-node load swap 2 to relaxe interconnects
- #33 resulting, better placement ⇒ placement complex esp. for not fully connected a lot of work ahead
- #34 memory allocation policies should be set for long-running allocation-intense tasks
- #35 timeline: gap of 7 years groundwork is laid API calls to read topology memory policies: NUMA-aware allocation scheduler knows balancing between NUMA nodes is more expensive will try to avoid that sounds good? apparently thats what most people thought 7 year gap but as we will see: still plenty that is missing and continue in 2012 sched/numa
- #36 a typical long-running computation… process starts main controlling thread
- #37 process loads its data for computation allocations done where it runs (DEFAULT)
- #38 process starts worker threads due to load: some scheduled on other node
- #39 lets say some workers finish early e.g. input sanitizers: finished cleaning up input what happens: spread out after all unnecessary load on interconnects
- #40 what possibilities do we have? remember basic approaches: mm vs. sched so we could migrate the memory
- #41 or reschedule the threads
- #42 or maybe do a combination of both
- #43 sched/numa Feb 2012 the challenge this was just one scenario but represents what may happen: memory spread out over nodes especially if tasks run for long time and are memory intense
- #44 sched/numa Feb 2012 first possibility: migrating memory tackles two questions when how to do that efficiently when: on page fault page still in page table but marked as not present (concept we will see again later) this bit is set: when task is migrated to different node or when task explicitly requests migration of all its memory how to do it efficiently so page is only migrated to node when requested by fault handler this spreads load out over time e.g. no dedicated kernel thread that does batch-migrations and only migrates pages that are actually used effectively: mem follows task
- #45 that was mm now scheduling part so far: static home node scheduler tried to keep task there (and allocate mem there) but as seen: situation may change e.g. lots of remote memory then assign new home node for task and request lazy migration for mem on other nodes this is the task follows memory part
- #46 also something novel NUMA groups declar group of tasks as NUMA group via system call effect they share the same home ndoe if one is migrated, all are you can actually bind memory to the group what is this good for? tasks w/ shared mem (e.g. threads) run on one node (hopefully)
- #47 autonuma Mar 2012 new player: Andrea Arcangeli also redhat employee at the time things spread out anyways remember example just now e.g. tasks that do not fit on one node basically says: forget the home node/ preferred node different approach: clean up two possibilities: sched vs mm migrate task or page decide based on page access statistics gather using page faults k-thread periodically marks anonymous pages as “not present” upon access: fault generated in fault handler update statistics for each task record: how many pages on each node for each page record: what was last node to access it
- #48 migrate task? if mostly remote page accesses and other tasks currently running on that node are not that well suited migrate memory? b/c memory may be spread out: can only migrate task to largest part heuristic: if on 2 subsequent faults -> access from same remote node then add to migration queue problems (pointed out by Peter) kernel worker threads used to scan address space -> force page faults and to migrate queued pages e.g. if system slow: now direct accountability -> why is it slow? for-each-CPU loop in scheduler in schedulers hot path doesn’t scale with # CPUs
- #49 timeline discussion btw Peter and Andrea two grew a bit out of hand Unless you're going to listen to feedback I give you, I'm going to completely stop reading your patches, I don't give a rats arse you work for the same company anymore. You're impossible to work with. apart from that: short comparison of sched/numa and autonuma sched/numa avoid separation in first place -> home node move mem with task (lazy) possibly change home node of task dev can explicitly define NUMA group -> share home node autonuma scleanup afterwards statistics gathering via page faults next step: combination into numa/core maybe redhat stepped in Peter tried to combine the best of both
- #50 numa/core Oct 2012 combine existing ideas lazy page migration (sched/numa) benefit: less performance impact when task is migrated page faults to track access patterns (autonuma) determine ideal placement dynamically: no static home node modify some things scan address space: proportional to task runtime problem before: task w/ little work but lots of mem -> large impact only if task gathered >1s runtime ignore short-running (theory: don’t benefit from NUMA aware placement) add some new stuff identify shared pages from CPU access patterns add last_cpu to page struct -> auto-detect NUMA groups assume task remains on CPU for some time page fault: accessed by other CPU == other task? instead of manually defining them as before try to move memory-related tasks to same node actually made it into linux-next staging tree for next kernel release
- #51 timeline while Peter and Andrea were arguing other devs had noticed (Mel Gorman, IBM) while Peter worked on numa/core Gorman worked on balancenuma
- #52 balancenuma Mel Gorman: objections to implementation of both approaches but also: objections to approaches themselves both specific solutions on how to schedule / move memory tested, but not widely tested on lots of NUMA hardware and not compared to many different approaches his vision: compare more policies more of an academic approach first step: make it easier to build & evaluate such policies basic mechanisms can be shared btw policies add basic infrastructure page fault mechanism lazy migration vmstats (virtual memory statistics) approximate cost of policy on top of this: implemented baseline policy MORON mem follows task migrates memory on page fault && remote access his suggestion for going forward test other policies e.g. rebase sched/numa and autonuma onto this foundation finally merged after 1 year of back and forth ----------------- pte: page table entry sched/numa obscures costs hard-codes PROT_NONE as hinting fault even (should be architecture-specific decision) well integrated, work in context of process that benefits autonuma kernel threads: mark pages to capture statistics obscures costs some costs: in paths that sched programmers are weary of blowing up performance tests: best performing solution.
- #53 now you – as a kernel hacker – can build NUMA-aware policies consider both scheduling memory management now made easier reuse basic mechanisms (e.g. lazy page migration) evaluate your policies compare them to existing
- #54 timeline next step on top of balancenuma not only mem mgmt but also scheduling
- #55 basic scheduler support Peter started pitching in again reuse of more existing stuff a little bit of autonuma per task counters: detect where task mem lives problem: NUMA scheduling possibly in conflict with scheduling goal of max. load only handle special case for now: swap w/ other task that also benefits a dash of numa/core agreed that identifying groups was a good thing but less heuristic: remember which task accesses page not enough space in page_struct for full task id use bottom 8 bits: collisions possible and a pinch of tweaks ignore shared libraries would pull everything together by ignoring read-only pages and shared executable pages mostly in CPU cache anyway summary NUMA-aware scheduling (not just mm) try to uphold load balancing goal and auto detect NUMA groups also in Kernel!
- #56 pseudo-interleaving also already in kernel basically yet another tweak for special case: workload (e.g. group) > 1 node if that happens, then so far mem distribution btw nodes is random example begin with one task (purple) starts allocating mem among that mem also some that will be shared by other task (green) e.g. threads in same process maybe at some point mem spills over into other node then other task that shares some of the mem comes (orange) scheduled on other node (e.g. b/c of load) starts allocating memory now not ideal distribution goals keep private mem local to each thread avoid excessive NUMA migration of pages distribute shared mem across nodes (max. mem bandwidth) how-to identify active nodes for workload balance mem lazily btw these
- #57 what would really be ideal private pages local for each task shared pages distributed evenly reduce congestion of interconnect
- #58 these are exactly the goals of this patch keep private mem local to each thread avoid excessive NUMA migration of pages (back and forth) distribute shared mem across nodes (max. mem bandwidth) how to achieve that? identify active nodes for workload balance shared mem lazily btw these
- #59 now you – as a developer – can lean further back kernel will try to optimize also for NUMA groups (e.g. threads) and even if workload > 1 node ...or you can still manually tune
- #61 scheduling domains and complex topologies elements in one hierarchy level (node level) might not be equally expensive to migrate to
- #62 mesh topology topology does not really matter there is always a neighbor with distance = 1 distance straight forward ⇒ no new domain hierarchy
- #63 ex: backplane toplogy controllers: nodes w/o memory cannot run tasks problems controllers add 1 to distance controllers in same domain as nodes but cannot run tasks distances for all combinations of nodes new scheduling domain groups of nodes nodes with same distance to all other nodes
- #64 notes from Storage, Filesystem, and Memory Management Summit 2014 feedback needed! performance, problems, enhancements, … devs are willing to improve for others problems! this is rare!
- #65 4-node system close to optimal 4+ nodes bad performance page access tracking too expensive? need more awareness of topology? not fully meshed performance test highly individual a benchmark needed possible? highly app specific
- #66 IO cache location unaware force free of memory for page cache swapping vs. uncached IO page aging swap out unused pages page cache is interleaved
- #67 IO / device awareness group network processes group IO-heavy processes multi-level swap to other nodes to disk
- #68 much to to many possible ways again: test feedback develop
- #69 Questions? … or per email.
- #71 SMP <-> NUMA caches should be warm <-> memory should be close HT <-> same node migration costs largely done by kernel (from the beginning?) <-> needs manual optimization