Data Backup Plan

Tue, 18 Jun 2024 20:22:21 +0800

TL;DR / Geek Summary:

Strategic implementation of the 3-2-1 backup rule to mitigate catastrophic hardware failure risks (e.g., Seagate’s “platter-scratching” firmware bugs).

Critical hardware analysis: Compares high-performance SAS RAID/HBA cards vs. the instability of Windows software-based parity during power loss events.

Multi-layered storage audit: Deep dive into 17+ cloud providers, optical media (organic vs. inorganic dyes), and the volatile charge-leakage issues of SSD/NAND flash.

# Introduction

Data backup is an effective way to minimize losses. Once a hard drive fails, no amount of money can easily restore your data. Many people claim their new hard drives are reliable and don’t need backups, but when the drive suddenly dies and they go to a data recovery company, they find it’s extremely expensive—1000 RMB for a 500GB drive. Regret comes too late. 1000 RMB could have bought several new drives. So, remember: backup your data before the drive fails; otherwise, it’s too late.

# Hardware RAID Cards

A RAID card refers to an HBA card that implements RAID. It typically consists of components like an I/O processor, disk controller, disk connectors, and cache.

HBA (Host Bus Adapter): A physical connection interface between a server’s internal I/O channel and a storage system’s I/O channel (external). Usually, server PC motherboards support the IDE protocol, and IDE disk controllers support it too, so they can connect directly. However, if a disk only supports the SCSI protocol, it cannot connect directly to the server; you need a SCSI card (HBA card) to enable disk support. Thus, the SCSI card becomes the Host Bus Adapter (HBA). The scope of HBA cards is quite broad.

RAID cards can be categorized by disk transmission protocol into SATA RAID cards, SAS RAID cards, and SCSI RAID cards. In the server field, SAS RAID cards are most common. Depending on the SCSI controller on the card, they can be single-channel, dual-channel, triple-channel, or even eight-channel. The number of channels determines how many disks can be carried: Number of channels * disks per single controller.

If there is no integrated SCSI controller and the card uses the motherboard’s SCSI controller to manage disks, it is a zero-channel RAID card.

Additionally, RAID cards are generally equipped with cache and a BBU (Battery Backup Unit). The former is for caching, and the latter is a battery configured for data integrity.

They are relatively expensive, ranging from several hundred to several thousand RMB. They are much more reliable than software RAID, offer higher speeds, and are relatively simple to configure.

# Software RAID

I strongly discourage using Windows’ built-in software RAID. This RAID is intended for servers; if a power outage occurs, it will perform a full copy of the entire data on your disks, which is extremely annoying. If another power outage occurs during this process, it’s absolute “Amennos” (game over).

Article on Enabling Windows Software RAID - This is an article teaching you how to use Windows’ built-in RAID functionality, but I don’t recommend using it. A power outage will trigger a full disk rebuild that takes a long time, and if power fails again during that, your disks might die directly.

# Two-way Synchronization

FreeFileSync is an open-source synchronization software that can sync to both local and remote servers simultaneously. It supports various protocols including FTP, SFTP, and WebDAV. It supports resume, file difference comparison, versioning, permission settings, and encryption. It’s a very powerful tool but requires manual setup. If you’re not tech-savvy, using a cloud drive is recommended.

Tutorial

# Cloud Storage

# Cloud Drive Encyclopedia

Cloud drives can serve as off-site backups, provided you encrypt your files (e.g., using encrypted zip files).

# Free Cloud Drives

1. Baidu Netdisk pan.baidu.com

Baidu Netdisk was established in 2012 (originally Baidu Cloud). It’s a veteran service provider with a massive user base. It covers major PC and mobile OSs and has an official TV client. Free users get 2TB of space.

(Price images as of Feb 3, 2023). Super VIPs enjoy high-speed downloads, video speed control, 5TB space, HD quality, and instant sync.

The base space is large, but free users suffer from severe speed limits, making large file downloads very painful without a membership.

2. Tencent Weiyun www.weiyun.com

Tencent Weiyun was established in 2012 (originally QQ Disk). It’s another veteran tied to QQ, so it has many users. It covers major platforms. Free users get 10GB of space and 1GB of free daily upload traffic.

Members enjoy 3TB storage, 45GB daily upload traffic, playback/transfer acceleration, and offline downloads. Super VIPs get 6TB and online decompression.

Weiyun is more like Google Drive or OneDrive—it charges for space rather than bandwidth. Free users have slow uploads rather than slow downloads. It seems like a neglected project within Tencent now.

3. Alibaba Drive www.aliyundrive.com

Alibaba Drive was launched in 2020/2021. It’s an unthrottled, large-capacity drive that has become very popular. It covers major platforms and has third-party TV clients. Free users get 800GB of permanent space.

Super VIPs enjoy 8TB space, backup drives, photo backup, cloud decompression, etc.

It doesn’t limit speeds (though it’s starting to a bit now). Memberships are mainly for space and video features. Downloads don’t strictly require membership. A downside is that files like .zip or .apk cannot be shared.

4. Tianyi Cloud (e-Cloud) cloud.189.cn

Tianyi Cloud was established in 2011 by China Telecom. Its main advantage is that it’s unlikely to shut down—a very strong selling point. Free users get 30+30GB (30GB cannot be shared) and 2GB daily traffic.

Gold members get 2TB+2TB space and 200GB daily traffic. Platinum members get 4TB+4TB and unlimited photo storage. It’s very reliable and cheaper than Baidu.

5. Lanzou Cloud www.lanzoux.com

Lanzou Cloud was established in 2014. It’s very popular for resource sharing. However, it lacks content moderation, which is a risk. Free users can only upload files up to 100MB; paid users up to 300MB. It’s too small for many needs.

6. Chengtong Netdisk www.ctfile.com

Chengtong was established in 2009. It’s a famous “earn-from-downloads” drive. Personally, I don’t like it because of its annoying mechanics. Free users get 500GB (requires frequent login to keep).

Membership offers high-speed downloads and WebDAV. It’s okay if you disable the “earn” mode, otherwise, it’s very slow.

7. Quark Drive pan.quark.cn

Quark was established in 2019 by Alibaba. It uses Alibaba Cloud infrastructure. It’s affordable and high quality. Free users get 10GB.

It used to have a very cheap “youth plan.” It’s decent and doesn’t restrict sharing as much as Alibaba Drive, though it has started some speed limiting.

8. 123 Cloud Drive www.123pan.com

123 Cloud Drive was established in 2021. It’s a rising star with 2TB free space and no speed limits for downloads. However, it has strict censorship—you cannot share videos, audio, or encrypted archives (though you can still download them yourself).

I currently use this drive for my website, and I’ve been whitelisted to share encrypted archives. It’s great because it doesn’t require login for downloads.

9. Quqi Cloud www.quqi.com

Quqi was established in 2017. It has a nice group feature, good for small teams. Free users get 2TB. It’s a small company without a big backer, but speeds are generally okay.

10. Xunlei Cloud pan.xunlei.com

Xunlei was established in 2003 and returned in 2020. Despite its age, it’s not very popular. It throttles speeds like Baidu. Free users get 500GB.

I have no good feelings for this drive—lots of ads, poor offline downloads, and it negatively affects the domestic BT environment.

11. Nutstore (JianGuoYun) www.jianguoyun.com

Nutstore was established in 2011. It’s focused on enterprise collaboration. It limits traffic (1GB upload/3GB download per month for free users) rather than space. Not recommended for personal use.

12. 115 Drive www.115.com

115 was established in 2009. Famous for its offline download capabilities. Free users get 15GB. It’s expensive and its offline features aren’t what they used to be. For offline downloads, PikPak is better.

13. China Mobile Cloud yun.139.com

China Mobile Cloud (formerly HeCaiYun) was established in 2014. It doesn’t throttle speeds. Free users get 20GB+20GB. Mobile users can get 1TB through tasks. It cannot share APK files.

14. China Unicom Cloud pan.wo.cn

China Unicom Cloud has a history of shutting down and rebranding. I don’t trust it with important files. Free users get 10GB.

15. CowTransfer cowtransfer.com

Established in 2017, focused on quick file transfers rather than long-term storage. Free users get 10GB. It’s unthrottled.

16. Feiji Pan (Small Plane Drive) www.feijipan.com

Established in 2021, allows downloads without login. Free users get 10GB. It’s unthrottled but it’s a new company, so longevity is uncertain. (I’ve stopped using it).

17. BitQiu pan.bitqiu.com

Established in 2006, focused on white-collar office users. Free users get 20GB. It’s a decent alternative to 115 Drive.

# Optical Discs

Ordinary recordable DVDs have dye fading issues because they use organic dyes. Luck lasts 10 years; bad luck lasts only a few.
Blu-ray discs use inorganic dyes, which are relatively better and last longer.
Most importantly, never buy dual-layer (DL) discs; they have a very high failure rate.
Store discs in a cool, dark place. Do not expose them to sunlight, as solar energy is similar to the recording laser and can overwrite your useful data with useless “solar noise.”

# LTO Tape

Not recommended for ordinary people. While tapes are cheap, the drives are expensive (400 to 2000 RMB used) and require many accessories. Also, random access is almost zero because you have to wait for the tape to wind to the specific data location.

# Printing

Printed materials can generally last 50 years. If protected with sealed bags and iron boxes, paper documents are almost impossible to destroy; they will only yellow and age.

# USB Sticks, SSDs, SD/CF Cards

Do not use these for long-term backups. They use flash memory, which stores data as electrical charges. Over time, the charges leak. Eventually, ECC correction fails, and data is lost. This usually happens within 10 years.

# Mental Memory

Useful for passwords, but requires constant use to prevent forgetting.

# Horror Stories of Data Recovery

People who gamble with their data often end up paying a high price for recovery.

After inquiring, I found that recovering data from a used 500GB Seagate drive costs 600-1000 RMB, whereas the drive itself only costs 10-15 RMB shipped.

# Backup Multiple Copies

A backup drive and main drive failed simultaneously—a “minefield” of a DM001 and an MQ03 drive. One had a firmware bug, the other “froze” to death.

Backup and Main Drive Failure Together

# Models that Scratch Platters upon Failure

Certain Seagate models will scratch the platters as soon as they fail. If you have one of these, back up your data immediately.

Seagate 7200.12 1TB

Seagate “Record Player” DM001

# Consequences of No Off-site Backup

Off-site backup protects against robbery, fire, water, earthquakes, or lightning. Recovering data from fire-damaged drives is not only costly but often impossible.

Fire-damaged Drive (Recovery Failed)

Data Protection on LEl_FENG Blog