软件定义存储sds_XCBS云后端存储

User Requirements

Open-source and Interoperability

The open source-based cloud platform has obtained attention since it was born. With healthy ecosystem of standard IaaS constructed by OpenStack and other open-source community, vendors who are in pursuit of project agility, technical force aggregation and long-term self-controlled product leverage such open-source framework as main architecture for “Cloud OS”of scaled IaaS to build enterprise dedicated cloud or cloud platform for application industry.

Full-distributed Architecture

To realize VM block mirror management, mirror warehouse and data backup, major part of users select Ceph as underlying storage module for Cloud scenarios with KVM. Ceph is a mainstream open-source distributed storage platform which is widely used by large-scale internet and HP computing carriers and is now in vigorous development. The community led by Ceph foundation provides good technical ecosystem with deep involvement of various HW and OS leading vendors, ensuring the advantage of Ceph modules technology and outstanding interoperability among vendors.

Production and SLA

However, the adoption of native Ceph by enterprises is limited by the following problems:

Operations and maintenance are conducted by command lines to a large extend, presenting obvious low-efficiency
Deficient integration with enterprise-level HW and application ecosystem, highly depending on development
Lower level of single-point performance and SLA compared to traditional storage

Product Value

XCBS（XSKY Cloud Block Storage）存储是由 XSKY 星辰天合开发的企业级分布式存储产品，对开源云环境做了深度优化。相比社区版上游软件，XCBS 分布式存储系统在稳定性、性能、可靠性、易用性等方面进行了深度优化，实现可视化、自动化运维，为大规模云环境提供了可靠的存储基础架构支撑。

Why XCBS

Deep mutual authentication with open-source cloud platform vendors, deployed as hyperconvergence or storage pool

Dramatically reduce IO latency and linkage jitter compared to community Ceph with the same HW configuration

Visualized DevOps and troubleshooting and fine cluster management at HW-level

Interactions with Open Source Community

Ceph Jewel in 2016: XSKY ranked Top 3 in global source code contribution to Ceph
Ceph Luminous in 2017 : The only Asian team in Top10 contribution to performance optimization
Ceph Kraken in 2017: Haomai Wang,CTO of XSKY, was nominated as the only featured developer
XSKY 星辰天合为 Ceph 上游社区添加 InfiniBand RDMA 互联支持
AS one of Intel's SPDK partners in China, XSKY plays a leading role in integrating SPDK and BlueFS to improve storage efficiency with NVMe
In2018, XSKY participated in the establishment of Ceph foundation and became a top member

Advantages

Seamless connection to open-source solutions with cloud-native product

100% compatible with community Ceph for a long time
Optimized and refactored RBD interface, independently upgraded cloud platform and storage
Rich RESTful API interfaces connected to multiple platforms
Seamless migration of Ceph volumes with RBD interface

System performance optimization with release of hardware potential

Deep optimization of network/hard disk processing efficiency and IO path
Intelligent SSD caching and layered data writing for performance improvement
Patented LibRBD aggregation agent for improvement in single client performance by 20% and reduction in CPU cost per node by 40%
Innovative IO aggregation enables small random blocks to be aggregated into large sequential IO blocks, and provides stable performance under high pressure and high concurrency

Agile delivery with simplified DevOps

Rapid expansion and upgrade with wizard setup deployment and AI-assisted Ops
100% GUI Ops
Alert log available for end-to-end monitoring of all indicators
Support OpenStack online migration across clustered virtual machines and storage
Support self-service inspection tools for intelligent analysis of cluster health and identification of potential system hazards in advance
Support automatic custom rebalancing at idle time for unbalanced OSD due to expansion, shrinkage, etc., and reduce the negative impact

Storage features of enterprise level

Support IO data verification to prevent silent error
Adapting to complex scenarios of storage management with online volume expansion and shrinkage, automatic volume thin provisioning and real-time QoS change
ROW lossless instant snapshot and stretched cluster for high reliability and availability
Treatment of hard disk and network sub-health, and automatic identification of bad disk, slow disk and network failure with automatic isolation