共 8 篇文章 |
|
几个 Ceph 性能优化的新方法和思路(2015 SH Ceph Day 参后感)在 Ceph 集群中,往往使用 SSD 来作为 Journal(日志)和 Caching(缓存)介质,来提高集群的性能。为 PCI-e 驱动器定制的标准化的软件接口为 SSD 定制(别的是为 PCIe 所做的)SSD Journal : HDD OSD 比例可以从常规的 1:5 提高到 1:20对全 SSD 集群来说,全 NVMe SSD 磁盘Ceph ... 阅159 转1 评0 公众公开 16-08-12 15:24 |
Pool, PG and CRUSH Config ReferencePOOL, placement 和CRUSH 配置手册Pool, PG and CRUSH Config Reference?# PG 数量计算方法:每个OSD推荐配置100个pg,每个pool写4副本,有10个osd,则pg数量 = 100 * 4 / 4.So for# 10 OSDs and osd pool default size = 4, we''d recommend approximately# (100 * 10) / 4 = 250.osd pool defaul... 阅62 转0 评0 公众公开 16-08-12 10:54 |
Journal Config ReferenceJournal配置手册Journal Config Reference?Ceph OSDs use a journal for two reasons: speed and consistency.Speed: The journal enables the Ceph OSD Daemon to commit small writesquickly.Ceph OSD Daemons support the following journal settings:journal max write bytes 每次写Journal 最大写10M.The maximu... 阅38 转0 评0 公众公开 16-08-12 10:03 |
OSD Config ReferenceOSD 配置手册OSD Config Reference ?osd.0osd.1osd.2.[osd] osd journal size = 1024[osd.0] host = osd-host-a[osd.1] host = osd-host-bGeneral Settings?The delay is a randomvalue less than osd scrub min interval *osd scrub interval randomized ratio.# 当集群刚启动,或者osd进程重启,在接收读写请求之前,osd... 阅154 转0 评0 公众公开 16-08-11 21:16 |
However, you may override the defaults.Configuration Settings?When modifying heartbeat settings, you should include them in the [global]section of your configuration file.Monitor Settings?Description:Type:Default:The number of seconds laggy estimates will decay.60*60.The weight for new samples in laggy estimation deca... 阅70 转0 评0 公众公开 16-08-11 20:53 |
Type:This means that a failure of some OSDs could resultin a temporary service disruption if one or more OSDs fails.Consider addingmore OSDs to increase storage capacity.Repeat the foregoingprocess with a higher number of OSD failures (e.g., a rack of OSDs) to arrive ata reasonable number for a near full ratio.[global... 阅114 转0 评0 公众公开 16-08-11 20:30 |
海量小文件存储与Ceph实践。我们在Filestore存储引擎基础上对小文件存储设计了优化方案并进行实现,方案主要思路如下:将若干小文件合并存储在RADOS系统的一个对象(object)中,<小文件的名字、小文件在对象中的offset及小文件size>组成kv对,作为相应对象的扩展属性(或者omap,本文以扩展属性表述,ceph都使用kv数据库实现,如leveldb)进... 阅1188 转4 评0 公众公开 16-05-13 12:44 |