Hi,chris: why would 1GB be huge and wasteful? where will vpp get memory when it support 10k,100k or more flow tables? Malloc from system? Memory fragmentation? Why not use dpdk mempool or memzone that will use hugepage to improve the performance of looking flow table?
发件人: Luke, Chris [mailto:chris_l...@comcast.com] 发送时间: 2017年11月21日 23:04 收件人: WANGHONGBO [王洪波] 抄送: vpp-dev@lists.fd.io 主题: RE: [vpp-dev] hi vpp-dev, another question about hugepage As I recall, VPP only looks for 2MB pages; 1GB would be huge and wasteful. A quick search reminded me of this thread, which may be instructive for you: https://lists.fd.io/pipermail/vpp-dev/2017-September/006364.html Chris. From: wanghongbo [mailto:wanghon...@kingsoft.com] Sent: Tuesday, November 21, 2017 7:41 To: Luke, Chris <chris_l...@cable.comcast.com> Cc: vpp-dev@lists.fd.io Subject: 答复: [vpp-dev] hi vpp-dev, another question about hugepage Hi, chris: Can you spare a few minutes to help me? I have config the 1GB hugepages . But vpp still used the 2M hugepage. I want to know configurations How VPP run with 1GB huge pages. [root@localhost ~]# cat /proc/cmdline BOOT_IMAGE=/vmlinuz-3.10.0-693.el7.x86_64 root=/dev/mapper/centos-root ro crashkernel=auto rd.lvm.lv=centos/root rd.lvm.lv=centos/swap default_hugepagesz=1G hugepagesz=1G hugepages=64 iommu=pt intel_iommu=on isolcpus=0-11 rhgb quiet /usr/bin/vpp[1815]: clib_sysfs_read: open `/sys/devices/system/node/node0/hugepages/hugepages-2048kB/free_hugepages': No such file or directory /usr/bin/vpp[1815]: clib_sysfs_read: open `/sys/devices/system/node/node1/hugepages/hugepages-2048kB/free_hugepages': No such file or directory Thanks wanghongbo 发件人: Luke, Chris [mailto:chris_l...@comcast.com] 发送时间: 2017年11月21日 10:31 收件人: WANGHONGBO [王洪波] 主题: RE: [vpp-dev] hi vpp-dev, another question about hugepage I understand too, I am just trying to make it easier for people to answer your questions and to make it useful for others in future who can search 😊 Chris. From: wanghongbo [mailto:wanghon...@kingsoft.com] Sent: Monday, November 20, 2017 21:14 To: Luke, Chris <chris_l...@cable.comcast.com> Subject: 答复: [vpp-dev] hi vpp-dev, another question about hugepage OK, understand. I just want to list more information for question. Thank you. 发件人: Luke, Chris [ <mailto:chris_l...@comcast.com> mailto:chris_l...@comcast.com] 发送时间: 2017年11月20日 22:14 收件人: WANGHONGBO [王洪波]; <mailto:vpp-dev@lists.fd.io> vpp-dev@lists.fd.io 主题: RE: [vpp-dev] hi vpp-dev, another question about hugepage Hello, Apologies for not answering your query, however is there any chance you could stop pasting images all the time? Cut-and-paste will suffice. Including images will routinely trip the list-max-message-size limits. Besides, text is searchable and images are not, thus limiting the usefulness to future generations. Thanks, Chris. From: <mailto:vpp-dev-boun...@lists.fd.io> vpp-dev-boun...@lists.fd.io [ <mailto:vpp-dev-boun...@lists.fd.io> mailto:vpp-dev-boun...@lists.fd.io] On Behalf Of wanghongbo Sent: Monday, November 20, 2017 9:11 AM To: <mailto:vpp-dev@lists.fd.io> vpp-dev@lists.fd.io Subject: [vpp-dev] hi vpp-dev, another question about hugepage Hi vpp-dev: I configure the 1G hugepage in /boot/grub2/grub.cfg and split cpu core. Then I configure the /etc/sysctl.d/80-vpp.conf. After reboot the kernel was crash. How can I configure the vpp hugepage? Thank you very much.
_______________________________________________ vpp-dev mailing list vpp-dev@lists.fd.io https://lists.fd.io/mailman/listinfo/vpp-dev