How SSDs Supercharged a Telecom Billing System and What You Need to Know About Their Longevity

Background

On 2014-11-17 a provincial telecom operator upgraded its billing system to Oracle Database 11g Release 2 (11.2) and introduced solid‑state drives (SSDs) into the existing storage array. Instead of a full‑flash solution such as Exadata X5, about 5 TB of bare‑metal SSDs were added to a conventional disk array.

SSD Deployment Strategy

After evaluation, the team decided to place the SSDs in a dedicated tablespace that hosts the high‑frequency billing tables (e.g., daily transaction, invoice, and settlement tables). The steps were:

Create a new tablespace on the SSD LUNs:

CREATE TABLESPACE SSD_TS DATAFILE '/ssd1/ssd_ts01.dbf' SIZE 500G AUTOEXTEND ON NEXT 100M MAXSIZE UNLIMITED;

Move target tables to the new tablespace:

ALTER TABLE billing_daily MOVE TABLESPACE SSD_TS;

Rebuild indexes on the SSD tablespace to benefit from low latency I/O.

Performance Results

Multi‑round benchmark tests simulated a full‑day billing run. Compared with the original disk‑only configuration, the SSD‑backed tablespace reduced the total processing time by roughly 40 % (e.g., from 120 minutes to 72 minutes). The production rollout showed an even larger gain, likely because real‑world I/O patterns were more random than the synthetic tests.

One‑Year Operational Experience

After twelve months of continuous operation the operations team reported:

No SSD failures.

No measurable performance degradation.

Stable I/O latency and throughput.

These observations support the feasibility of deploying SSDs at scale in traditional enterprise core systems.

SSD Endurance Considerations

SSD endurance is expressed in program/erase (P/E) cycles. Typical NAND process nodes have the following ratings:

34 nm NAND ≈ 5 000 P/E cycles.

25 nm NAND ≈ 3 000 P/E cycles.

For a 600 GB SSD rated at 3 000 P/E cycles, the theoretical total write capacity is:

3 000 cycles × 600 GB ≈ 1 800 TB of data.

Assuming the billing workload writes 200 GB per day, the drive would last about 9 years, well beyond the typical hardware refresh cycle.

NAND Type Selection

Enterprise SSDs are available with different NAND cell technologies:

SLC – single‑level cell, highest endurance, highest cost.

eMLC – enterprise‑grade multi‑level cell, balanced endurance (≈ 3 000 P/E) and cost.

MLC – consumer‑grade, lower endurance.

For production environments the eMLC variant is recommended because it provides sufficient durability while keeping the price reasonable.

Best Practices and Caveats

Even with acceptable endurance, SSDs can fail unexpectedly. Recommended safeguards include:

Implement regular backup of the SSD tablespace and associated redo logs.

Configure Oracle Flash Recovery Area on separate storage.

Monitor SSD health via SMART attributes or vendor‑provided management tools.

Maintain a disaster‑recovery plan that can fall back to the traditional disk array if needed.