Oracle RAC部署到分布式存储上,可以解决数据集中存储的IO瓶颈,同事还能支持横向扩展,有了真正的分布式存储,才能真正意义上的云架构。再数据读写速度上面有几倍的提升。
1:实现方式
a:Turn standard NVMe PCIe SSDs inside database servers into scalable shared storage
b:0.4 to 50 TB per node; 2 to 100 nodes
c:Leverage proven Oracle ASM for high availability and data mirroring
d:Maximize database performance with ReadLocal™ Technology
结构图
architecture highlights
‥shared storage based on standard NVMe PCIe SSDs
‥use of high-capacity SAS HDDs for bulk storage needs
‥storage located inside the database nodes (converged nodes) or in separate storage nodes
‥x86 servers used as database and storage nodes
‥FlashGrid software manages SSD devices and connectivity, integrates with Oracle ASM
‥ASM manages data, volumes, mirroring, snapshots
‥2-way or 3-way mirroring of data across separate nodes
‥Choice of 10/40/100 GbE or InfiniBand/RDMA for network connectivity
‥FlashGrid Read-Local Technology minimizes network overhead by serving reads from local SSDs at the speed of PCIe
服务器架构方式1
服务器架构方式2
2:Minimize Costs with Commodity Hardware and Free Software
Reduce storage costs by up to 90% compared to proprietary storage hardware. Commodity
servers and SSDs have lower costs and allow second-sourcing for better pricing and parts
availability. Pay only for the flash capacity that you actually need for a particular cluster,
even if required capacity is as small as 1TB. Fully eliminate or cut the amount of the high-cost
Fibre Channel infrastructure.
3:Standardize High Bandwidth and Low Latencies
with NVMe PCIe SSDs
NVMe, the industry standard for PCIe SSDs, delivers outstanding performance of up to 3 GB/s
and up to 750,000 random IOPS per SSD. Multiple NVMe SSDs can be installed per server,
up to 18 SSDs in some server models. The hot-plug capability makes adding or replacing
an NVMe SSD as easy as adding or replacing a HDD.
4:Prevent Network Congestion with FlashGrid ReadLocal™ Technology
A congested storage network can limit the performance advantages of flash storage. The FlashGrid ReadLocal™
Technology reduces network traffic by up to 90% or more by serving read requests from local SSDs instead of moving data
over the network. In 2-node or 3-node clusters 100% of the read traffic can be served locally. Unlike caching, the FlashGrid
Read-Local™ Technology accelerates even large table scans and backup operations.
如图: