ye yuqiang has posted comments on this change. ( http://gerrit.cloudera.org:8080/15313 )
Change subject: KUDU-3063: Set a ratio to reserve some nvm memory ...................................................................... Patch Set 2: (3 comments) http://gerrit.cloudera.org:8080/#/c/15313/1/src/kudu/util/nvm_cache.cc File src/kudu/util/nvm_cache.cc: http://gerrit.cloudera.org:8080/#/c/15313/1/src/kudu/util/nvm_cache.cc@76 PS1, Line 76: DEFINE_double(nvm_cache_usage_ratio, 1.25, > Should also tag as advanced; most users won't need to modify this. tag it as advanced. http://gerrit.cloudera.org:8080/#/c/15313/1/src/kudu/util/nvm_cache.cc@76 PS1, Line 76: 1.2 > I'm curious what's the worst case when using memkind with NVM? Could it ha There's a micro benchmark under memkind, https://github.com/memkind/memkind/blob/master/test/fragmentation_benchmark_pmem.cpp I changed the block size to allocate from 10kB, 20kb, 30kb, 40kb, 80kb, 100Kb, 200Kb, 300Kb, 400Kb, 800Kb, 1mb, 2mb, 3mb, 4mb, 8mb randomly. The result shows usage is 0.7x. Suppose it should be a very bad case. If the block size range from 4mb to 8mb, the usage is 0.9x. http://gerrit.cloudera.org:8080/#/c/15313/1/src/kudu/util/nvm_cache.cc@80 PS1, Line 80: > Nit:Is a validator needed to ensure the value is not above 1.0? Is that ok to validate it before memkind_create_pmem and also log the reduced capacity here? -- To view, visit http://gerrit.cloudera.org:8080/15313 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-Project: kudu Gerrit-Branch: master Gerrit-MessageType: comment Gerrit-Change-Id: Iceb1931f5addbf76774854ca1613b6a085b577e3 Gerrit-Change-Number: 15313 Gerrit-PatchSet: 2 Gerrit-Owner: ye yuqiang <[email protected]> Gerrit-Reviewer: Adar Dembo <[email protected]> Gerrit-Reviewer: Alexey Serbin <[email protected]> Gerrit-Reviewer: Bankim Bhavsar <[email protected]> Gerrit-Reviewer: Kudu Jenkins (120) Gerrit-Reviewer: ye yuqiang <[email protected]> Gerrit-Comment-Date: Mon, 02 Mar 2020 06:55:14 +0000 Gerrit-HasComments: Yes
