Jianzhi core taocloud full flash SDS helps build high-performance cloud services

FASS& Bare metal clouds

Compared with the virtualization mode , Bare metal servers provide users with physical server rental services , It continues the unique advantage of non-destructive performance of physical machine , It can meet the strict requirements of core application scenarios for high performance and stability at the same time . Use bare metal server to monopolize client resources , Don't share with other customers CPU、 Memory, etc. , Performance can be fully exploited . Because of the complete isolation at the physical machine level , Can provide more security 、 Reliable computing resources . Bare metal servers can achieve performance that is no worse than physical servers , At the same time, it can realize the elastic expansion and use of resources , Fully automated operation and maintenance management , More and more attention is paid to financial securities 、HPC And so on 、 Users with high data security requirements are welcome .

Storage system as the underlying of cloud platform IT resources , Efficiency of the platform 、 Stable operation plays a key supporting role , Therefore, we need to build security 、 reliable 、 Efficient storage platform , Lay a solid foundation for comprehensively improving the user's use effect .

Full face cloud ecosystem

Storage systems need to work with mainstream clouds 、 Containers 、 The virtualization platform is fully connected , Cooperate with cloud platform , Provide dynamic and elastic expansion capability , With the fast-growing tenant distribution of cloud platform 、 Business system deployment 、 Business data and applications such as the proliferation of shared files match .

High performance

Ensure that the underlying storage system of bare metal cloud IO、 bandwidth 、 Delay does not become a bottleneck , It can meet the business needs of different applications on the upper layer .

Adapt to multiple services 、 Elastic expansion

Bare metal cloud platform needs to carry a variety of user applications , Access to the storage resource pool covers file sharing 、 Block storage 、 High bandwidth transmission, etc , And the business demand for capacity is also increasing year by year , Therefore, the storage system must provide a variety of interface modes , And convenient capacity expansion , Simplify the operation process of system personnel for storage resource pool expansion .

Highly reliable

Provide mature 、 Stable storage resource pool , Have diversified data protection capabilities , In terms of software and hardware matching 、 Equipment operation reliability 、 The system fault recovery ability and other aspects should be strictly tested and verified , To ensure that the storage resource pool provides secure and reliable data storage support for the cloud platform , Minimize the risk of data loss .

·

Solution

△ The overall network architecture of the system

This scheme adopts FASS The distributed full flash storage system serves as the data storage base . According to business requirements and data characteristics , Build two storage clusters for users , A configuration SATA SSD Build capacity storage cluster with hard disk , Another configuration NVme SSD Hard disks build performance-based storage clusters .

Both storage clusters pass 100GbE High speed network connection , Provide storage services in block storage .FASS Use cross node SSD Resource to create a storage pool , And divide into multiple LUN volume , among SATA SSD The storage cluster uses iSER agreement ,NVme SSD The storage cluster uses NVMeoF Protocol mount LUN resources .

To satisfy AI、HPC And other high-performance file storage requirements , The scheme is also configured FASS High performance file gateway ,FASS Provide storage space for file gateway in the way of block storage , Then provide high-performance file storage services through the file gateway .

Bare metal ECS clusters pass 25GbE network connections , While ensuring performance , Make full use of resources , Reduce input costs .

FASS Every LUN Thin provisioned volumes , Space that can be allocated far more than actual physical resources , Meet the changing business needs of each host . The background can be based on the actual capacity of each host , Dynamically allocate physical resources .

Both storage clusters are redundant through double copies , And ensure that multiple copies of the same data will not be stored in the same fault domain by configuring the fault domain , The failure of any number of nodes in the fault domain will not make all data copies inaccessible , Greatly meet the user's demand for system reliability . in addition , The storage system protects the core business data volumes through snapshots , Fully ensure the safety of users' important data .

As a standard distributed storage system ,FASS Support 256 Node Scale-out Expand , A single volume can reach 2PB, It can change with business needs , Easily increase capacity and performance by scaling horizontally , Meet the huge needs of cloud business .

·

Program advantages

Excellent performance

FASS Metadata services are memory based K/V Database implementation , File data and metadata are stored and managed separately , With excellent performance expansion capabilities . adopt NVMe The disc provides the ultimate I/O performance , Combined with the software layer XPE Engine design 、 Microcontroller technology 、 Lightweight metadata indexing and other mechanisms , The block storage side can reach the level of one million per volume , Single cluster （4 node ） Tens of millions IOPS、 Hundred microseconds delay , Single cluster （4 node ） Bandwidth up to 50GB.

Unified storage 、 Easy to share 、 Elastic expansion

File gateway +FASS The distributed full flash storage scheme provides both high-performance block and file access , It can meet the storage and access requirements of various applications .

As the business grows , The scheme can configure the corresponding cluster size as required , Meet business performance requirements with minimal investment . Compared with traditional full flash storage ,FASS The storage system supports horizontal expansion , A single cluster can be expanded to 256 Nodes , A single volume can reach 2PB, With the expansion of cluster size , The overall performance of the system increases linearly , Fully meet users' growing capacity and performance requirements .

Ultra high availability architecture design

FASS Clusters provide a variety of redundancy mechanisms ： Replicas can provide higher performance ; Erasure code （EC） It can provide a comprehensive cost-effective choice of capacity and cost , So as to meet the different performance of users 、 Different capacities 、 Project requirements with different costs . Whether copy or EC, When some cluster nodes fail （1～n/2, Depending on the redundancy type ）, Still provide continuous data access support .

The file gateway also adopts the high availability mechanism of multi node cluster , As long as one file gateway node survives , It can continuously provide file access support for front-end applications , Realize ultra-high availability guarantee .

Rich advanced functions

QoS、 snapshot 、 Fault domain 、 Rich advanced functions such as thin provisioning , Perfect solution for fault recovery 、 Data security 、 System reliability 、 Capacity utilization and other issues , Comprehensively improve the user experience .

Fully integrate the ecological industry

FASS The storage system can be connected with OpenStack、ZStack、K8S、VMware、KVM Wait for the mainstream cloud 、 virtualization 、 Perfect docking of container platform , And support Restful API Docking with various platforms .

In response to the national information innovation strategy ,FASS The storage system is fully adapted to mainstream localization CPU And the operating system , Like flying 、 Kunpeng series server , Kirin operating system, etc .

·

Applicable scenario

In new information technology 、 In the context of mobile Internet , The bare metal server based on distributed full flash block storage can meet the needs of Internet 、 Artificial intelligence 、 big data 、 Genetic engineering and other technology companies , With its ultra-high cost performance 、 High performance 、 Customizable and other advantages , Solve the problem of resource shortage of customers in the expansion period .

01
Finance 、 Securities industry scene with high requirements for safety and supervision ： Enterprises that have strict requirements for data security .

02

High performance computing scenarios ： Supercomputing Center 、 High performance computing scenarios such as gene sequencing have a large amount of data processing , The computing performance of the server 、 stability 、 High real-time requirements .

03

Core database scenario ： Some key database services cannot be deployed on virtual machines , It has to be exclusive through resources 、 Network isolation 、 Physical servers with guaranteed performance .

04

Mobile application scenarios ： Around mobile apps , Especially the development of mobile games 、 test 、 go online 、 operating .