Understanding SSD Basics: Principles, Architecture, Risks, and Maintenance

Content

CPU speed and memory bandwidth have increased dramatically, but traditional magnetic hard drives remain a system bottleneck; SSDs have become prevalent in both consumer and enterprise markets.

Several real‑world stories illustrate common SSD pitfalls.

<code>【故事1】上周有个朋友问，他的笔记本电脑配了SSD硬盘，刚买来的时候速度很快，他挺满意的，用了一年多以后发现速度大不如前，以为是OS的问题，结果费了半天劲重装了系统，更新了驱动以及各种软件，折腾完发现IO速度还是没有以前快。

【故事2】还有个朋友，给自己的电脑换了一块240GB的SSD，在一次意外断电后，这块SSD的数据竟然全部丢失，可用容量竟然只能识别到8MB。而厂商技术人员的回复是：这是一个产品缺陷，如果想找回原有数据，则需要人为造成意外断电，会有千分之一的几率恢复数据。-_-!

【故事3】朋友的互联网公司购买了一批PCIE闪存卡作为数据库加速应用。上线以后发现系统资源占用率很高，而且系统很不稳定，导致业务受到很大影响，厂商也束手无策。应用部门相当恼火，选型部门和采购也十分无奈，只好全部撤换。各种协调宕机时间也让各部门头疼不已。

【故事4】有个朋友比较土豪，把自己的移动硬盘也更新成了SSD。我问他：你多久备份一次数据呢？答：每年都会备一次。问：会经常用移动硬盘里的数据么？答：也不经常，几个月可能读一次吧，都是些照片和老电影。我说：那小心吧，因为你的数据可能会消失！会消失！会消失！</code>

SSD Basic Principle

Flash memory stores data by trapping electrons in an insulated gate, representing voltage levels.

Repeated program/erase cycles thin the gate, eventually preventing the cell from holding charge, which leads to failure.

Manufacturers specify two key parameters: P/E cycles (Program/Erase cycles) and DWPD (Drive Writes Per Day) to estimate lifespan.

For example, a 240 GB SSD with 3000 P/E cycles can endure 720 TB of total writes; a 1.8 TB PCIe card with DWPD 10 and a five‑year warranty can write up to 32.8 PB.

SSD Composition and Types

Key components include the controller, NAND chips, interface, DRAM buffer, capacitors or batteries, PCB, and packaging.

Based on NAND technology, there are SLC, eMLC, MLC, TLC, and QLC.

SLC (Single‑Level Cell) stores 1 bit per cell, offering high speed, long life, but higher cost.

MLC (Multi‑Level Cell) stores 2 bits per cell, providing lower cost and higher capacity at the expense of endurance.

TLC and QLC store 3 bits and 4 bits per cell respectively, with further reduced endurance.

eMLC is a higher‑grade MLC selected from the best die and tested with stricter criteria; its typical endurance is around 10 k P/E cycles.

Reference endurance values: SLC ≈ 100 k cycles, eMLC ≈ 10 k, MLC ≈ 3‑5 k, TLC ≈ 500‑1 k.

Form factors include USB flash drives, SATA/PCIe SSDs, M.2 modules, SAS SSDs, and NVMe cards.

Future trends focus on 3D‑NAND and NVMe, which offers parallel I/O processing and lower latency.

NVMe processes I/O across multiple CPU cores, unlike AHCI which relies on a single core.

PCIe flash cards come in Host‑Based and Device‑Based architectures; Device‑Based handles the Flash Translation Layer (FTL) on the card, while Host‑Based does it in the driver.

Controller Functions

The controller is the SSD’s CPU, managing ECC, wear leveling, bad‑block handling, read‑write interference, garbage collection, and over‑provisioning.

ECC corrects bit errors; as bad blocks increase, ECC workload rises, causing performance degradation.

Some vendors implement distributed ECC per NAND die to offload work, though this raises cost.

Over‑provisioning (OP) reserves extra space for replacing bad blocks; consumer SSDs typically have about 7 % OP.

Operational Monitoring and Risks

Use vendor monitoring tools or SMART data to track temperature, remaining life, and error rates.

If performance drops, check for increasing bad‑block counts and ECC activity; replacement may be necessary.

Unexpected power loss can cause data loss; testing with repeated power‑off cycles is advisable before deployment.

JEDEC defines data‑retention limits: consumer SSDs retain data for 1 year at 30 °C, enterprise SSDs for 3 months at 40 °C.

Critical data should still be backed up to magnetic disks for higher reliability.

Despite these risks, SSDs deliver substantial performance gains when properly selected, monitored, and maintained.