Home NAS server (3) | SSD cache acceleration mechanical hard disk

household NAS The server （3）| SSD Cache acceleration mechanical hard disk

Original address ：CSDN post

Summary of the previous article ： household NAS The server （2）| HyperV Of Winserver 2022 and Ubuntu 22.04 Dual system

---

---

1. Preface

Yahoo , Readers, we meet again o(￣▽￣)ブ. Remember that in the last chapter, we have already installed Windows Server, And the Ubuntu Server Success lies in Hyper-V Install in . After successful installation , The author gave the mechanical hard disk directly to Ubuntu virtual machine .

But everybody knows that , The random read and write performance of mechanical hard disk is very low , So in order to improve HDD The ability to read and write at random , In this chapter , I will take you , By setting bcache Tools , Will be with SSD SSD as cache , To improve the random reading and writing ability of mechanical hard disk .

Besides , I will also briefly pass my previous articles , To reduce the power consumption of mechanical hard disk . My blog is here ：SeaChest Utilities Tools , Make your hard disk low power consumption and long life .

---

2. SSD Cache acceleration

SSD Caching can effectively alleviate the problem of low random read and write performance of mechanical hard disk , Therefore, it will be the first thing to set up before the server is put into use .

Before starting the demonstration , The author first explains my Ubuntu Equipment structure . If you find that your equipment structure is different , You need to adjust the order yourself Or in the Hyper-V Modify the device location in . The author's equipment structure is as follows ：

Device name	describe
/dev/sda	System disk . The entity is external Winserver Of virtual disks .
/dev/sdb	Solid state disk , Will be used as SSD cache .
/dev/sdc	first 16T Mechanical drive .
/dev/sdd	the second 16T Mechanical drive .

2.1 Hard disk partition

Although the hard disk is not forced to partition , The author still likes to divide each storage device into a zone , Use here fdisk Tool creation GPT Table and partition ：

#  Enter the mechanical hard disk  fdisk  Interface 
> sudo fdisk /dev/sdc
#  establish  gpt  Partition table 
> g
#  Create a new partition , All default settings , That is, create only one partition , Occupy all equipment space 
> n

If readers have their own zoning needs , You can adjust the partition structure by yourself . here SSD Do not participate in partition .

After partition , The author's current zoning equipment is /dev/sdc1 And /dev/sdd1.

2.2 bcache

2.2.1 bcache install

bcache It's a Linux Tools under , It allows the use of solid state drives as read and write caches （writeback Pattern ） Or read cache （writethrough perhaps writearound Pattern ） Come for another block equipment （ Usually mechanical hard disk or hard disk array ） Speed up .

Besides ,bcache The team also developed a software based on bcache Of bcachefs, This file system is providing SSD Cache acceleration , It also provides COW(Copy On Write Copy while writing ）、 Data encryption 、 Data backup, disk array and other functions . But here we have set xfs file system , You don't need to use , The following sections refer to Bcache - ArchWiki.

stay Ubuntu If you want to enable , Need to download first bcache Tools . Fortunately, ,apt The source already contains bcache, Just use the following command to download and install ：

sudo apt install bcache-tools

Be careful ： hinder bcache Initialization operation , All data on the hard disk will be cleared , therefore bcache It should be the step before file system initialization .

then , Start Initialize solid state disk and mechanical disk ：

#  Create a back-end device （ Generally speaking, this is your mechanical hard disk ）.
#  The back-end device can be the whole device 、 A partition or anything else  block  equipment .
make-bcache -B /dev/sdc1
make-bcache -B /dev/sdd1

#  Create a cache device （ This is usually your SSD ）.
#  The cache device can be an entire device 、 A partition or anything else  block  equipment 
make-bcache -C /dev/sdb

If the above command is successfully executed , So now /dev There will be two more devices in the directory ：/dev/bcache0 and /dev/bcache1, Corresponding to /dev/sdc and /dev/sdd. At present, the corresponding relationship between the cache and the back-end device is ：

SSD Cache related devices	The original device （ Partition ）	describe
nothing	/dev/sdb	SSD Cache disk
/dev/bcache0	/dev/sdc1	Mechanical drive 1
/dev/bcache1	/dev/sdd1	Mechanical drive 1

after , We use solid state drives as cache devices , Add to mechanical hard disk （ Back end devices ） In the cache list of . First View the current SSD cache device UUID：

ls /sys/fs/bcache/

Through this line of command, you can see that there is a UUID, That's the cache device UUID 了 . without , Then use the following command Register caching device ：

echo /dev/sdb > /sys/fs/bcache/register 

The above two commands can see the cache device UUID, Put this UUID remember , And add it to the cache list of back-end devices , that will do Enable SSD Cache acceleration ：

echo UUID__from_previous_command > /sys/block/bcache0/bcache/attach
echo UUID__from_previous_command > /sys/block/bcache1/bcache/attach

2.2.2 bcache operation

Show bcached Device information ：

bcache-super-show /dev/sdXY

Stop the backend device ：

echo 1 > /sys/block/sdX/sdX[Y]/bcache/stop

Take the cache device offline ：

echo 1 > /sys/block/sdX/sdX[Y]/bcache/detach

2.2.3 bcache Pattern

bcache There are altogether 4 Kind of , Namely ：

• none： No caching at all .
• writeback： The best cache mode , All read and write requests are in SSD When the request is completed, it is regarded as the request is completed . Be careful , Although this mode can greatly improve the performance , But because of SSD The data on the will be synchronized asynchronously with the back-end device , Therefore, if power failure occurs during synchronization , May lead to SSD Data loss on . For all that ,bcache After the next restart , The asynchronous synchronization process that was not completed last time will continue to complete , Minimize the risk of data loss .
• writethrough： Default cache mode , The performance is lower than the above one . In this mode , All read operations will pass through the cache layer . The difference is , All write requests will be made to SSD And back-end devices , And only after the backend device writes , Consider the request completed . The advantage of this is , The performance of the cache is improved for read operations in a period of time after writing .
• writearound： Low performance mode , But it can prolong SSD Life span of . In this mode , Only cache read operations .

Back end device runtime , You can view the caching mode adopted by the backend device ：

cat /sys/block/bcache0/bcache/cache_mode
cat /sys/block/bcache1/bcache/cache_mode

The author considers SSD The problem of the life span of , And I'm not very interested in the cache of write operations , Therefore, choose to modify the cache mode to writearound Pattern ：

echo writearound > /sys/block/bcache0/bcache/cache_mode
echo writearound > /sys/block/bcache1/bcache/cache_mode

Of course , If readers have their own needs , You can choose your own cache mode .

2.3 xfs file system

The author chose xfs file system , Mainly valued its actions RHEL The stability of the default file system of the distribution , At the same time, the performance is not backward ext4 How many? . Then there's a chance , The author will issue a separate issue on Linux Under the comparison of major file systems , We will discuss it in detail at that time .

On the back-end device with a good cache mechanism , We will start to format the file system . This file system is optional , It doesn't have to be xfs file system , Only the author chose xfs.

First , install xfs Related tools of ：

sudo apt install xfsprogs

after , Set up a file system for our backend device partition , Pay attention to... Directly after that bcache* To operate the equipment ：

sudo mkfs.xfs /dev/bcache0
sudo mkfs.xfs /dev/bcache1

thus ,xfs The file system already contains SSD Successfully created on the cached hard disk , You can mount .

2.4 fstab Power on auto mount

After the file system is established , It's best to set it to mount automatically .

here , I plan to mount the two disks on /mnt/sdc and /mnt/snapraid_backup. First, create a mount point ：

sudo mkdir -p /mnt/sdc /mnt/snapraid_backup

And then by looking at bcache* The equipment UUID And write it down ：

sudo lsblk -f

Use the editor to open /etc/fstab, Add the following code ：

UUID=<bcache0  Of  UUID> /mnt/sdc xfs defaults 0 0
UUID=<bcache1  Of  UUID> /mnt/snapraid_backup xfs defaults 0 0

Save and restart the system , If it works , With SSD The cached mechanical hard disk is established .

---

3. summary

This issue , The author finally put SSD The cache acceleration mechanism is set up , It's also a good adjustment bcache The function of . stay bcache blessings , Random access speed over a period of time From the original 2MB/s To the present 30Mb/s, Compared with the original mechanical hard disk , Already there. A huge improvement .

Besides , In this issue, the file system will also be created , For the next issue mergerfs and snapraid The appearance of . Only when mergerfs + snapraid Set up , It is really a safe harbor for data .

Comrades, if you think my content is useful , Please leave me a compliment and comment , It's best to pay attention , I will be very happy ！

Recently, I saw many blog posts with great moisture , The content is plain and homogeneous , I still don't want to become such a marketing number after all , May the spirit of open source survive ！