CN107135277A - A method and device for implementing an all-in-one machine - Google Patents

Abstract

The present invention provides a method and device for implementing an all-in-one machine. The above method includes the following steps: using at least one database server to calculate data; managing the data through at least one storage module, integrating data calculation, transmission and storage functions, and faster Deployment and application launch time, which facilitates the subsequent business expansion and upgrade of enterprises.

Description

A method and device for implementing an all-in-one machine

technical field

The invention belongs to the field of computers, and in particular relates to a method and device for realizing an all-in-one computer.

Background technique

As Hyper-Scale applications are widely deployed in large IT enterprises, the database server system based on the IOE architecture needs to cope with the concurrent access of a large number of users. Components such as networks and storage devices can meet customer needs.

When applications require high storage I/O instruction response speed (IOPS) and low storage access latency (Latency), large-scale high-performance storage and high-end hosts are usually purchased. These devices are not only expensive, but also poor in deployment and scalability. , and requires professional maintenance. The environment of the entire architecture is complex, requiring host personnel, S-fibre switches, and storage technicians to participate in deployment and maintenance.

Database server cluster systems based on host and centralized storage architecture (so-called IOE architecture) are widely used in finance, telecommunications, social security, and energy industries, and undertake important core business systems of customers.

At present, related technologies mainly adopt the IOE architecture design scheme. Two key application hosts are responsible for database computing servers, and high-end storage is mapped to key application hosts to store data through the FC protocol. When a large number of IO accesses are made, the two computing nodes can simultaneously support front-end application access. The traditional IOE architecture has high costs, poor scalability, high maintenance costs, and performance that cannot meet big data queries.

At present, most of the existing database cluster systems use key application hosts and disk array architectures. Disk arrays basically use traditional mechanical hard disks. Traditional mechanical hard disks have been developed for decades and can still cope with data throughput. However, when dealing with high I/O In terms of command response speed (IOPS), it is difficult to meet the needs of enterprises for high-performance data access. Traditional mechanical hard disks can usually provide hundreds of IOPS. If the application requires higher IOPS, a large number of mechanical hard disks are required to build a large disk array and arrange data reasonably. Although large-scale disk arrays have also used SSD disks in recent years, they are still unable to meet high-throughput application scenarios due to the limitation of fiber bandwidth of 16Gb; in addition, the technical complexity of traditional infrastructure limits the performance of flash memory. A huge burden of operation and maintenance has come, which makes enterprises passively invest a lot of infrastructure maintenance costs while maintaining important business decisions.

Therefore, there is a need for an architecture solution that provides higher performance with lightweight equipment to solve the problems of poor scalability, high maintenance costs, and performance that cannot meet large data queries.

Contents of the invention

The present invention provides a method and device for implementing an all-in-one machine to solve the above problems.

The invention provides a method for realizing an all-in-one machine. The above method comprises the following steps:

use at least one database server to perform calculations on the data;

The data is managed by at least one storage module.

The present invention also provides an all-in-one machine, including: at least one database server and at least one storage module, wherein the at least one database server is connected to the at least one storage module:

at least one database server for computing data;

At least one storage module is used to manage the data.

Adopt the following solution: use at least one database server to calculate data, manage data through at least one storage module, and design a converged architecture, use standard X86 servers to integrate computing and storage into a common converged architecture, and use InfiniBand high-speed between nodes Interconnected to form a lightweight load balancing cluster, precisely integrated into a one-stop cabinet. Pre-integrated distributed clusters, databases, and cloud platform management systems, all software and hardware components are pre-configured and tuned to achieve integrated and fast delivery, and provide higher performance with lightweight equipment while minimizing TCO and maximizing ROI. The all-flash configuration eliminates the processing bottleneck caused by the I/O delay of the traditional architecture, fully releases the CPU processing power, has excellent performance on OLTP databases, and provides unparalleled processing power for database services at a higher cost performance.

Description of drawings

The accompanying drawings described here are used to provide a further understanding of the present invention and constitute a part of the application. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations to the present invention. In the attached picture:

Fig. 1 shows the processing flowchart of the all-in-one machine implementation method in Embodiment 1 of the present invention;

FIG. 2 is a structural diagram of an all-in-one machine according to Embodiment 2 of the present invention.

detailed description

Hereinafter, the present invention will be described in detail with reference to the drawings and examples. It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

Fig. 1 shows the processing flow chart of the all-in-one machine implementation method in Embodiment 1 of the present invention, including the following steps:

Step 102: Use at least one database server to perform calculations on the data.

Step 104: Manage the data through at least one storage module.

Further, at least two flash memory cards are configured for the storage module.

Further, using Share-Nothing distributed technology to fuse the at least one storage module into a clustered logical resource pool;

The data is migrated from the storage module to the database server using RDMA zero-copy technology.

Further, the at least one database server communicates with the at least one storage module through an IB network.

Wherein, the database server is a general X86 server.

Therefore, high IOPS flash disks are configured on general-purpose X86 servers, and the Share-Nothing distributed architecture is adopted to integrate the local SSD storage of each node into a unified clustered logical resource pool for the entire site. The LUNs allocated from the storage resource pool pass through the IB network Mapping the front-end computing nodes, the main features of the Infiniband protocol are high bandwidth, low latency, and good system scalability.

Network transmission uses RDMA zero-copy technology to quickly move data from storage nodes to computing nodes without any impact on the operating system, and does not require much computer processing capabilities. It eliminates external memory copy and text exchange operations, This frees up memory bandwidth and CPU cycles for improved application performance.

Two DB servers (database servers) are installed with operating system Redhat6.5, and the system is installed with Infiniband driver and storage management software. The database can be deployed with Inspur K-DB KAC; three Storage Servers need to be installed with operating system Redhat 6.5, Infiniband driver and Sdata storage management Software, each Storage Server is configured with two PCIE-SSD flash memory cards; two DB servers and three Storage Servers are interconnected through the IB network. For specific implementation parameters, refer to the following table:

The installation parameters can refer to the following table:

Flash memory technology: Compared with disk technology, flash memory can provide higher IOPS, usually up to hundreds of thousands at the current stage, which largely solves the bottleneck of enterprise applications' demand for IOPS and brings more value to customers . The innovative non-mechanical structure of flash memory technology eliminates the magnetic track structure of traditional disks, which can transmit data faster, greatly improve the data reading speed, and significantly reduce time delay. The high IOPS of flash memory can effectively meet the IOPS requirements of applications with high virtual machine density, and linearly upgrade IOPS to meet the expansion requirements of applications.

Distributed storage: With software definition as the core, using the Share-Nothing distributed architecture, the local SSD storage of each node is integrated into a unified clustered logical resource pool for the whole site, which realizes the dynamic deployment of storage resources and improves the Scale- Up and Scale-out multi-dimensional elastic expansion capabilities. Allow users to expand capacity on demand and flexibly balance resources according to business load.

RDMA zero-copy technology: RDMA's zero-copy network technology enables data to be quickly moved from one system to remote system memory without any impact on the operating system and without the need for much computer processing power. It eliminates external memory copy and text swap operations, thereby freeing memory bandwidth and CPU cycles for improved application performance.

InfiniBand: InfiniBand is a unified interconnect structure that can handle storage I/O, network I/0, and inter-process communication (IPC). The main features of the nfiniband protocol are high bandwidth, low latency, and good system scalability (it can easily realize an Infiniband network with tens of thousands of devices without congestion), and it supports redundant I/O in single or multiple interconnected networks channels, thus keeping the data center operational in the event of a partial failure.

FIG. 2 is a structural diagram of an all-in-one machine according to Embodiment 2 of the present invention.

As shown in FIG. 2, an all-in-one machine according to an embodiment of the present invention includes: at least one database server 202 and at least one storage module 204, wherein the at least one database server is connected to the at least one storage module:

At least one database server 202, used for calculating the data;

At least one storage module 204, configured to manage the data.

Further, the storage module 204 includes at least two flash memory cards.

Further, the storage module 204 uses the Share-Nothing distributed technology to fuse the at least one storage module into a clustered logical resource pool;

The database server 202 uses RDMA zero-copy technology to read the data from the storage module.

Further, the at least one database server 202 communicates with the at least one storage module 204 through an IB network.

Further, the database server 202 is a general X86 server.

Based on the general X86 server development platform, it integrates data computing, transmission and storage functions, faster deployment and application launch time, and facilitates subsequent business expansion and upgrading of enterprises; provides millions of IOPS, subtle-level delay, and improves business processing capabilities , shorten the business waiting time, improve the work efficiency of the enterprise; the whole set of technical solutions has a fully redundant architecture, and there is no single point of failure, which greatly ensures the high availability of the business system.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (10)

1. a kind of all-in-one machine realization method, it is characterised in that comprise the following steps：

Data are calculated using at least one database server；

The data are managed by least one memory module.

2. according to the method described in claim 1, it is characterised in that configure at least two flash cards for the memory module.

3. according to the method described in claim 1, it is characterised in that will be described using Share-Nothing distributed computing technology At least one memory module is fused into concentrating type logical resource pond；

The data are migrated to the database server from the memory module using RDMA zero duplication technologies.

4. according to the method described in claim 1, it is characterised in that at least one described database server and described at least one Pass through IB network services between individual memory module.

5. method according to any one of claim 1 to 4, it is characterised in that the database server is general X86 Server.

6. a kind of all-in-one, it is characterised in that including：At least one database server and at least one memory module, wherein, At least one described database server is connected with least one described memory module：

At least one database server, for calculating data；

At least one memory module, for managing the data.

7. all-in-one according to claim 6, it is characterised in that the memory module includes at least two flash cards.

8. all-in-one according to claim 7, it is characterised in that the memory module is divided using Share-Nothing's At least one described memory module is fused into concentrating type logical resource pond by cloth technology；

The database server reads the data using RDMA zero duplication technologies from the memory module.

9. all-in-one according to claim 6, it is characterised in that at least one described database server with it is described at least Pass through IB network services between one memory module.

10. the all-in-one according to any one of claim 6 to 9, it is characterised in that the database server is general X86 servers.