Elastic Block Storage
Carrier light computing storage (elastic block storage)
Elastic block storage service benchmarking AWS EBS
Project background
LUISUAN Technology has in-depth cooperation with an operator to create a customized light computing storage disk frame based on the GP7000 all-flash storage platform, benchmarking AWS EBS elastic block storage, and providing a high-performance, elastically scalable, and highly reliable persistent block storage base for the operator's cloud host.
Core Highlights
| Highlights | key data |
|---|---|
| Ultimate storage performance | Up to 70 million IOPS · 300GB/s bandwidth |
| Light computing integration | Built-in GP7000-DPU, protocol offloading and nearby processing |
| Elastic block storage | Benchmark AWS EBS, create/mount/expand on demand |
| Green energy saving | The overall energy consumption of the cluster is reduced by more than 30% |
Why you need elastic block storage
In the cloud computing system, elastic block storage (such as AWS EBS) provides mountable persistent disks for cloud servers: creating volumes on demand, mounting/unmounting at any time, flexible adjustment of capacity and performance, and supporting snapshot backup, encryption, and multi-tenant isolation. With the expansion of operators' AI training, big data and high-concurrency enterprise applications, the adaptability of the original storage facilities is insufficient, and a set of EBS-like high-performance block storage base is needed.
| AWS EBS capabilities | This plan is realized |
|---|---|
| Block Volume | Elastic block volume based on GP7000, which can be mounted to cloud hosts |
| Performance tiering (gp3/io2, etc.) | Multi-level IOPS/throughput configuration, up to 70 million IOPS, 300GB/s |
| Snapshots and backups | Snapshot/backup mechanism and dual-port NVMe high availability |
| Flexible expansion and scheduling | Connect to cloud management and control platform for automated management and flexible scheduling |
| Encryption and tenant isolation | Multi-tenant volume isolation and encrypted storage |
core design
Storage module: end-to-end NVMe all-flash
- Dual-port NVMe U.2 SSD, single disk capacity up to 248TB
- The whole machine can be configured with up to 24 hard drives and 300GB/s bandwidth
- Up to 70 million IOPS, end-to-end NVMe architecture
Light computing module: GP7000-DPU
- Integrated customized lightweight computing unit with storage protocol processing and data offload computing capabilities
- Avoid redundant computing power of traditional computing nodes and achieve the integration of "light computing + high-performance storage"
Storage and calculation collaboration and management and control docking
- Diskless storage and computing separation architecture, independent storage expansion, storage tasks are processed nearby
- Open up the operator's cloud management and control platform to automate management, flexible scheduling and status monitoring
Benchmarking the core capabilities of AWS EBS
- Elastic block-level storage: Create/mount/unmount block volumes on demand, elastically adjust capacity and performance
- High performance layering: Multi-level IOPS/throughput configuration to meet AI training, high-concurrency database and other scenarios
- Data is reliable: Snapshot and backup, dual-port high availability, data integrity assurance
- safety and isolation: Multi-tenant volume isolation and encrypted storage, adapting to the multi-tenant architecture of the cloud platform
Solution value
| value | illustrate |
|---|---|
| Excellent performance and energy efficiency | Extreme storage performance supports AI training and high-concurrency databases, reducing cluster energy consumption by more than 30% |
| Lower the replacement threshold | Accurately match resource allocation, smoothly connect with existing facilities, and significantly reduce replacement costs and risks. |
| Can be copied and upgraded | The "universal platform + customized transformation" methodology can be extended to data centers in different regions |
Related products
otherCase
Intelligent Computing Center·AI Inference Acceleration
KV Cache offloading · GPU utilization increased to 94%
View Case
Scientific research institutions · High performance computing
High-throughput base for supercomputing and scientific research big data
View Case
Futures trading backtesting system
Backtest reporting cycle shortened from 30 days to 2.5 hours
View Case
Smart ranch solution
From experience farming to data farming, operational efficiency increases by 40%
View CaseWant to know how we implement your scenario?
Contact us and the LUISUAN technical team will provide targeted solution suggestions based on your business.
Inquire Now