Ibm ds Series-Cache

First, we will summarize the previous series of articles on IBM DS3K/4 K/5 K:

========================================================== ======================================

Ibm ds Storage ISCSI Configuration -- http://taotao1240.blog.51cto.com/731446/1003338

IBM Copy Service -- Flashcopy Introduction -- http://taotao1240.blog.51cto.com/731446/1002206

IBM copy service -- flashcopy lab -- http://taotao1240.blog.51cto.com/731446/1007957

IBM Copy Service -- VolumeCopy -- http://taotao1240.blog.51cto.com/731446/1009501

IBM DS3500 with Windows 2008 Multipathing-http://taotao1240.blog.51cto.com/731446/999875

Ibm ds Storage -- Replace drive option -- http://taotao1240.blog.51cto.com/731446/978740

IBM copy service -- ERM -- http://taotao1240.blog.51cto.com/731446/1011445

Ibm ds Storage Upgrade Controller Firmware -- http://taotao1240.blog.51cto.com/731446/1124072

Three Modes of ibm ds storage Failover-http://taotao1240.blog.51cto.com/731446/1135679

Several common problems with ibm ds storage (1) -- http://taotao1240.blog.51cto.com/731446/1130424

IBM Dynamic Disk Pool details-http://taotao1240.blog.51cto.com/731446/1261875

========================================================== ======================================

Cache, a common term in computers, is also available on the Host, storage, and even vswitch routers. Why is there a cache? One sentence is to solve the difference between the CPU speed and the hard disk speed. What are precautions for cache in the DS series?

Through SM, we can see the cache attribute of Lun, for example, we can modify each attribute by ourselves, but we must understand what each attribute means before modification, otherwise it will cause unnecessary trouble.

650) this. width = 650; "style =" background-image: none; border-right-0px; padding-left: 0px; border-top-width: 0px; border-bottom-width: 0px; border-left-width: 0px; padding-right: 0px; padding-top: 0px; "title =" image "border =" 0 "alt =" image "src =" http://www.bkjia.com/uploads/allimg/131228/003Z11038-0.png "width =" 446 "height =" 484 "/>

1. What is the difference between read caching and dynamic cache read prefetch?

Read caching is the read cache, that is, when the host requests to read A data block A from the disk, it will first check whether there is any in the cache, and if so, it will directly return to the host, in this way, you do not need to read data from the disk. If not, you can send A read request to the disk. the disk will first put Block A in the cache, it guessed that the next host may need to use data block. This is the read cache.

Dynamic cache read prefetch, that is, read-ahead, also known as the pre-read mechanism, has a set of internal algorithms to determine the amount of data pre-read. Similarly, if the host requests to read A data block A from the disk, it will first check whether there is any data block in the cache, and if there is one, it will directly return it to the host, thus saving the need to read data from the disk; if no, send A read request to the disk. However, the disk caches data blocks B, C, and D near, it guesses that the host may use data blocks B, C, and D. This is the pre-read mechanism. One thing to emphasize is that read-ahead is not applicable to all scenarios. It is better in the case of sequential IO and is not applicable to scenarios with random IO.

2. What does "enable write cache with Caching" mean? We often encounter the need to change the ctrl, because the cache is stored on the controller. After changing the ctrl, will the cache on the ctrl be lost? None! There will be mirroring, that is, the image. The two ctrl caches will mirror each other. When the host sends an IO to control A and writes the IO to control B's cache, when control A receives the write acknowledged message from control B, control A will return A successful write operation to the host. This ensures the cache consistency between the two ctrl S. Of course, this one-step copy operation will slightly affect the performance, but it can ensure data consistency. For details, see

650) this. width = 650; "style =" background-image: none; border-right-0px; padding-left: 0px; border-top-width: 0px; border-bottom-width: 0px; border-left-width: 0px; padding-right: 0px; padding-top: 0px; "title =" image "border =" 0 "alt =" image "src =" http://www.bkjia.com/uploads/allimg/131228/003Z1A96-1.png "width =" 644 "height =" 303 "/>

3. "enable write cache without batteries"

The Cache reduces the response time of the application, but there is a problem. We know that the cache has NVSRAM, and the data in the cache is lost once the power is down. Therefore, ensure that the data in the cache is available, there must be a battery to ensure that. There are two Battery pairs in the DS series to ensure that the cache data is safe. Each controller has its own battery. One problem is that the cache in memory will flush when the power is down? Disk? No. It is a Flash Backup Device located on the motherboard. Only when the data in the cache is fully flushed to the flash backup device can the system truly turn off power. Therefore, the maintenance guide emphasizes that when both power supplies are switched off in the power off system, you must wait until the battery lights are completely destroyed to unplug the controller. This is because power off is not the real power off, and there is data in the cache!

What factors will affect the cache? Will battery charging be affected? The answer is no under normal circumstances. The cache is disable only when the battery capacity is lower than a threshold, because it does not have enough power to flush the cache to USB after it feels that the power is lower than the threshold, therefore, it will disable the cache to avoid data loss.

When talking about battery, I have to mention learn cycle. Will it always cache force disable? Learn cycle is divided into three stages: discharge, rest, and charge. disable is only possible in the discharge stage. Because different machines preset whether the discharge is lower than that threshold value. For example, if the threshold is 60%, some machines will be placed in 61% during discharge, and then charge will not affect the cache, but some will also be placed in 55%, this is 60% ~ In the 55% process, the cache will be disable. Is an example of learn cycle.

650) this. width = 650; "style =" background-image: none; border-right-0px; padding-left: 0px; border-top-width: 0px; border-bottom-width: 0px; border-left-width: 0px; padding-right: 0px; padding-top: 0px; "title =" image "border =" 0 "alt =" image "src =" http://www.bkjia.com/uploads/allimg/131228/003Z1EQ-2.png "width =" 1560 "height =" 153 "/>

Therefore, the "enable write cache without batteries" attribute means that the write cache will enable no matter whether there is a battery or not. In fact, we do not recommend this because data in the cache will be lost once the power is down, data Consistency is affected.

4. Cache Flushing

-- Data times out 10 S by default). If the cache blocks is not rewritten within 10 S, flush to disk.

-- The cache space reaches the threshold. When the value of start is reached, flush cache is started. When the value of stop is reached, flush cache is stopped.

Flushing is a common "first-in-first-out" Least recently used], that is, first refreshing the oldest cache data

650) this. width = 650; "style =" background-image: none; border-right-0px; padding-left: 0px; border-top-width: 0px; border-bottom-width: 0px; border-left-width: 0px; padding-right: 0px; padding-top: 0px; "title =" image "border =" 0 "alt =" image "src =" http://www.bkjia.com/uploads/allimg/131228/003Z14442-3.png "width =" 610 "height =" 484 "/>

How to modify Flush cache? "Storage Subsystem" -- "Change" -- "Cache Settings"

650) this. width = 650; "style =" background-image: none; border-right-0px; padding-left: 0px; border-top-width: 0px; border-bottom-width: 0px; border-left-width: 0px; padding-right: 0px; padding-top: 0px; "title =" image "border =" 0 "alt =" image "src =" http://www.bkjia.com/uploads/allimg/131228/003Z13226-4.png "width =" 364 "height =" 484 "/>

There is a cache block size. What does this mean? Cache is composed of Cache Blocks. A 4 kb cache block can hold 8 Logical Block AddressLBAs data. Each IO is divided into at least one Full Cache Block. For example, if the cache block size of a 12 kb IO is 4 k, it is divided into three cache blocks. If the cache block size is 8 k, there are two cache blocks, but one of which only stores 4 kb and 4 kb is wasted. If the cache block size is 16 KB, It is a cache block, 4 kb is wasted.

650) this. width = 650; "style =" background-image: none; border-right-0px; padding-left: 0px; border-top-width: 0px; border-bottom-width: 0px; border-left-width: 0px; padding-right: 0px; padding-top: 0px; "title =" image "border =" 0 "alt =" image "src =" http://www.bkjia.com/uploads/allimg/131228/003Z11206-5.png "width =" 644 "height =" 215 "/>

Bytes ----------------------------------------------------------------------------------------------------

The cache is written so much before it is written...

This article is from the "star & storage" blog, please be sure to keep this source http://taotao1240.blog.51cto.com/731446/1287115