Hugepage

From HPCWIKI
Revision as of 17:17, 16 June 2023 by Admin (talk | contribs)
Jump to navigation Jump to search

Computer memory is allocated to processes as pages. Usually these pages are rather small (4K), meaning that a process consuming a lot of memory will also be consuming a lot of pages. Searching through a multitude of pages can result in system slow downs, which is why some servers can benefit from enabling huge pages.


In Database, Huge pages is especially useful on systems like database servers. Processes like MySQL and PostgreSQL can make use of huge pages if they are enabled, and will put less strain on your RAM cache


In Network stack, Hugepage support is required for the large memory pool allocation used for packet buffers (the HUGETLBFS option must be enabled in the running kernel as indicated the previous section). By using hugepage allocations, performance is increased since fewer pages are needed, and therefore less Translation Lookaside Buffers (TLBs, high speed translation caches), which reduce the time it takes to translate a virtual page address to a physical page address. Without hugepages, high TLB miss rates would occur with the standard 4k page size, slowing performance.[1]


most standard on modern systems would be 2 MB

$ grep -i huge /proc/meminfo
AnonHugePages:         0 kB
ShmemHugePages:        0 kB
FileHugePages:         0 kB
HugePages_Total:       0
HugePages_Free:        0
HugePages_Rsvd:        0
HugePages_Surp:        0
Hugepagesize:       2048 kB
Hugetlb:               0 kB

we can change Hugepagesize

$ sudo sysctl -w vm.nr_hugepages=102400  # for example 100MB
[sudo] password for hpcmate: 
vm.nr_hugepages = 102400
hpcmate@VPNGW:~/setup/vpn (master)$ grep -i huge /proc/meminfo           
AnonHugePages:         0 kB
ShmemHugePages:        0 kB
FileHugePages:         0 kB
HugePages_Total:   102400
HugePages_Free:    102400
HugePages_Rsvd:        0
HugePages_Surp:        0
Hugepagesize:       2048 kB
Hugetlb:        209715200 kB

To make it persistent across futher reboot, add following line in /etc/sysctl.conf and reboot

vm.nr_hugepages = 102400
#Check kernel support HugeTLB
$grep -i huge /boot/config-5.4.0-150-generic
CONFIG_CGROUP_HUGETLB=y
CONFIG_ARCH_WANT_GENERAL_HUGETLB=y
CONFIG_ARCH_ENABLE_HUGEPAGE_MIGRATION=y
CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE=y
CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD=y
CONFIG_HAVE_ARCH_HUGE_VMAP=y
CONFIG_ARCH_WANT_HUGE_PMD_SHARE=y
CONFIG_TRANSPARENT_HUGEPAGE=y
# CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS is not set
CONFIG_TRANSPARENT_HUGEPAGE_MADVISE=y
CONFIG_TRANSPARENT_HUGE_PAGECACHE=y
CONFIG_HUGETLBFS=y
CONFIG_HUGETLB_PAGE=y



#HugeTBL can be setup through /etc/sysctl.conf by adding
vm.nr_hugepages=256
# also need to edit the /etc/fstab file to mount the hugepages at startup
$ sudo vi /etc/fstab by adding
huge /mnt/huge hugetlbfs defaults 0 0

$ sudo mkdir /mnt/huge
$ sudo chmod 777 /mnt/huge

#also reboot your machine as the huge pages must be setup just after boot to make
#sure there is enough contiguous memory for the 2MB pages
#For a single-node system, the command to use is as follows (assuming that 1024 of 2MB pages are required)
echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages

# On a NUMA node system
echo 1024 > /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages
echo 1024 > /sys/devices/system/node/node1/hugepages/hugepages-2048kB/nr_hugepages


Some kernel versions may not allow reserving 1 GB hugepages at run time, so reserving them at boot time may be the only option by pass the hugepages option to the kernel.

default_hugepagesz=1G hugepagesz=1G hugepages=4


For 64-bit applications, it is recommended to use 1 GB hugepages if the platform supports them.

Reference