Cloud storage of memory advanced features blog directory
- Cloud computing storage
- 1. Storage Advanced Features
- 1.1 virtual disk allocation
- 1.2 Snapshot
- 1.3 copies
- 1.3.1 a copy of the distribution volume
- 1.3.2 a copy of replicated volumes
- 1.3.3 a copy of the volume and distribution of replicated volumes
- 1.3.4 multiple copies Disk Management
- (1) the number of hosts, the number of copies of the disk can be divisible case
- (2) the number of hosts, the number of copies of the disk can not be divisible case
- 1.3.5 multi-copy master reading and writing data
- 1.4 Data balance
Cloud computing storage
1. Storage Advanced Features
1.1 virtual disk allocation
When Sangfor virtualization platform to create a new disk for the virtual machine, there are two ways Optional disk allocation a pre-assigned a streamlined distribution to
(1) pre-assigned
Set pre-allocated disk in the disk initialization takes up all the space configuration, which can significantly enhance the IOPS and throughput of the virtual machine, virtual machine disk space but not in use can not be used by other virtual machines
(2) to streamline distribution
Virtual machine disk write-demand, virtual unused space can also be used for other virtual machine, you can save storage space, but must first apply for virtual machine storage space when writing to disk, this process has a certain impact on write performance, compared to it is better to pre-allocate disk performance
1. What is a streamlined distribution
A storage allocation is characteristic streamlined management, the core idea is to "cheat" operating system, using a host-by-byte check will not go all the space and storage space allocated to the host will not be instantly filled with this feature, the traditional supply-demand supply becomes fully
specified when thin volumes can be written as the dynamic expansion of host data capacity, but the capacity does not extend above the final volume to create a volume capacity
2. streamline distribution details
(1) LUN virtualization
Virtual LUN assigned to the host, to achieve the storage capacity on demand
For example: host sees space to store 500GB of actual end occupies only 200GB
(2) LUN dynamic expansion
LUN space by automatic growth strategy
expansion process hosts with no downtime
to maximize space utilization, reduce management complexity
(3) customer value
Improve resource utilization, for the first time to avoid wasting storage resource planning
3. streamline the advantages and problems of distribution function
Advantage
When the thin volumes between storage mirroring replication, cloning and storage, data synchronization only actually used, to save CPU, disk, and bandwidth overhead
problem
When the storage pool capacity is used up, but there is no time to add new physical resources. Resulting in thin volumes can not be extended space properly, the data can not be written to the underlying data block, have a significant impact on the application
so the system is running pay close attention to the storage pool to run state
4. Compact storage pool allocated state detecting
Streamlining the volume expansion from the storage capacity of the unallocated capacity pool
When the unallocated space of less than a certain percentage of the storage pool, the user is prompted
when a storage pool of unallocated space is less than the threshold percentage, an alarm is triggered, until adding new resources to the physical storage pool
after adding new physical resources, storage pool status to normal
1.2 Snapshot
1.2.1 Introduction Snapshot
Defined snapshot
SNIA (Storage Networking Industry Association) definition of a snapshot (Snapshot) is: a fully usable copy on the specified data set, which includes a copy of the corresponding image data at a point in time (the beginning point in time copy) of. Snapshots can be a copy of the data it represents, it can be a replica of the data
Value snapshot
Fast backup / restore
save multiple recovery point objectives
redefine the data use
Classification snapshot
When you copy write (COW: Copy On Write)
Write redirect (ROW: Redirect On Write)
1.2.2 snapshot technology
(1) COW(1)
(2) COW(2)
(3) COW(3)
(4) ROW(1)
(5) ROW(2)
(6) ROW(3)
(7) COW and ROW Compare
COW | ROW | |
---|---|---|
State of the source volume | Source volume to date | Source volume status freeze |
Write procedure | :当一个新的写操作执行 时:1.首先读出写操作将要覆盖地址的当前数 据 2.将读出数据保存至专用空间并建立索引3.新的写操作执行(写入目标地址) | 读写操作步骤 1.到源盘的新的写操作被存入日志(并索引)2. 读源卷时,先检索日志3.读快照时,源卷需要引用4.当快照取消时,写日志必须全部执行以与源卷保证 数据状态同步更新 |
读写路径影响 | 1.源卷的读路径基本无影响 2.源卷的写操作受拷贝影响3.对快照(卷)的读写路径都有影响 | 1.源卷的写操作基本无影响2.源卷的读路径潜在受影响3.快照(卷)的读写路径最优化 |
1.3 副本
副本是对原始数据的完全拷贝。
通过为系统中的文件增加副本,保存冗余的文件数据,可以有效地提高数据可靠性。aSAN可以选择配置文件副本数为2副本或3副本。副本数量、副本保存的位置对虚拟存储上层的应用不可见
1.3.1 副本分布卷
虚拟存储对上层应用可见的逻辑卷,aCloud/aDesk读写文件是对分布卷进行操作的 ,由分布卷组成分布式存储。
数据写入时先进行分布,分散到各个主机。
1.3.2 副本复制卷
1.复制卷将文件写入具体磁盘,通过副本间的同步复制实现数据冗余。
2.每个复制卷包含的物理磁盘个数跟副本数量相同,一个复制卷下每个物理磁盘的数据内容完全一致。
3.复制卷的最小组成单位是物理磁盘,即同一文件的不同副本只能保存在不同的物理磁盘上。
1.3.3 副本分布卷和复制卷
1.3.4 多副本磁盘管理
(1)主机个数、磁盘个数可以被副本数整除情况
(2)主机个数、磁盘个数不能被副本数整除情况
1.3.5 多副本主机读写数据
在没有故障等异常情况下,文件副本数据是始终一致的,不存在所谓主副本和备副本之分。
如果对文件X进行修改,如写入一段数据,这段数据会被同时写到2个副本文件。
如果是从文件X读取一段数据,则只会从其中一个副本读取。
1.4 数据平衡
1.4.1 数据平衡介绍
After the virtual storage capacity expansion, the old data still exists on the original disk, disk capacity utilization for different replication volume may be larger difference, this time by enabling data balanced plan that will balance the data to a different disk, you can make hosts rational use of resources.
1.4.2 Data balanced configuration
Balanced configuration data execution times the recommended configuration for the service available time slot, so as not to affect the normal business of virtual machines, long time if the data exceeds the configured to perform balance will automatically stop, so next time data to perform data migration balance again