https://github.com/johnramsden/gczns
https://github.com/johnramsden/gczns
Last synced: 8 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/johnramsden/gczns
- Owner: johnramsden
- License: mit
- Created: 2024-10-31T05:33:18.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-11-08T23:18:34.000Z (about 1 year ago)
- Last Synced: 2025-01-12T14:12:26.896Z (12 months ago)
- Language: C++
- Size: 10.7 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# GCZNS
## Hardware
ZNS:
```
sudo ./zbd report -i /dev/nvme0n2
```
* Vendor ID: WZS4C8T1TDSP303
* Zone model: host-managed
* Capacity: 1941.325 GB (3791650816 512-bytes sectors)
* Logical blocks: 473956352 blocks of 4096 B
* Physical blocks: 473956352 blocks of 4096 B
* Zones: 904 zones of 2048.0 MB
* Maximum number of open zones: 14
* Maximum number of active zones: 14
SSD:
* SN540
* WUS4B7696DSP303
* 894.3GiB
## Design
**ZNS**:
* **Storage Structure**: 904 zones, each sized at 1.0518 GiB, totaling 950.79 GiB.
* **Concurrency**: Max of 14 zones active or open simultaneously.
* **Write Strategy**: Use 14 concurrent writers, each writing to a separate zone. When a zone completes, move on to a new one.
**SSD**:
* **Storage Structure**: 894.3 GiB total capacity, split into 904 equal regions (size equivalent to ZNS zones).
* **Concurrency**: Write to 14 regions concurrently.
* **Write Strategy**: As each region completes, transition to a new region, maintaining the 14 concurrent writers.
---
This setup provides parity between ZNS and SSD experiments by aligning the storage and write processes to allow for consistent comparisons.