Data Center Disaster Recovery Backup
Data center disaster recovery backup refers to the establishment of an independent disaster recovery data center in addition to ensure that when the primary data center stops working due to a sudden disaster, disaster recovery data centers quickly take over all or part of the business that used to run in the primary data center, in order to reduce or avoid losses caused by disaster events. As an important supporting industry in the country, the financial industry has the strongest demand for data disaster recovery, and its investment in this field is also ahead of other industries. For example, a large state-owned bank, one of the top five banks in China, has the largest customer base in China and has invested heavily in the construction of disaster recovery systems in recent years. The OTN equipment of the optical path technology has successfully built a disaster recovery transmission network to realize the “triple” protection of the boards, subracks and networks of the service. By renting the carrier line and self-built the network mode, the TCO is effectively reduced. And fully ensure that the network is “managed, controllable, and operational”. The highly reliable mass bandwidth transmission network lays a solid foundation for the rapid development of ICBC business.
The construction of disaster recovery network
In most cases, the disaster recovery system needs to be interconnected by means of the WAN. Due to the particularity and importance of the disaster recovery, the requirements for the WAN are relatively high:
1) Large capacity:
Under the trend of information explosion, the bandwidth of single storage service has reached 1G/2G or even 4G/10G. There are dozens of such storage services in a SAN storage network, and the total bandwidth is tens of G to hundreds of G, even to the Tbit level, you need enough network bandwidth to transmit;
2) High scalability:
With the rapid increase of data traffic of financial enterprises, storage capacity must be expanded at high speed every year;
3) High scalability:
Real-time performance is a basic requirement of the data storage business. For example, when we go to the bank in our daily life, we hope that this can be received in real time;
4) High reliability:
For financial enterprises, the loss of key business data is intolerable, and the loss of data will cause huge losses;
5) Interface diversity:
5)Although the current mainstream interface type is FC (Fiber Channel), there are still many protocols coexisting in the data storage network, resulting in multiple types of interfaces, including ESCON, FICON, IP, etc. There are also ATM and SDH interface types. In addition, the interfaces provided by mainstream storage vendors such as IBM, EMC, Hitachi, and HP are not fully compatible. These conditions lead to a situation that the types of data storage service interfaces are not uniform and multiple interfaces coexist.
By analyzing the characteristics and needs of data center interconnection, the “Three Centers In Two Places” solution for data center disaster recovery has been lunched.
The “Three Centers In Two Places” solution includes two solutions for disaster recovery in the same city and disaster recovery in different places. The combination of the two schemes constitutes a complete disaster recovery network, achieving data “0 loss” and service “0 interruption”.
[The same city disaster recovery]
The same city disaster recovery mainly uses WDM/OTN technology to realize real-time data backup and real-time service switching to ensure business continuity. Real-time backup in the same city ensures that the data in the primary data center and the disaster recovery center are completely consistent, achieving the highest level of data protection, especially for core and key services, such as applications requiring RPO/RTO close to zero.
First, the WDM/OTN has large bandwidth and strong expansion capability, is particularly suitable for application scenario in which the data storage service has large bandwidth requirement and the subsequent bandwidth expansion is frequent. WDM/OTN is multi-wavelength transmission system with large transmission bandwidth and flexible expansion. Currently, the industry’s mainstream WDM/OTN system is 40/80 waves. If each wavelength transmits 10G of bandwidth, then 80 wavelengths can transmit up to 800G; if each wavelength transmits 40G of bandwidth, then 80 wavelengths can transmit up to 3.2T; if each wavelength transmits 100G Bandwidth, then a single fiber will achieve 8T massive transmission bandwidth.
Secondly, WDM/OTN, as an optical transmission technology, has a very low transmission delay, which is very suitable for the real-time requirements of the disaster recovery system. The transmission delay of WDM/OTN is mainly composed of optical fiber, dispersion compensation fiber and WDM/OTN equipment delay. The transmission delay of the fiber is fixed. It is only possible to reduce the delay from the dispersion compensation and WDM/OTN equipment.
Third, WDM/OTN, as a mature technology form, is very complete in terms of network security and reliability.
Finally, after years of development, WDM/OTN has been able to support all current mainstream storage service interfaces. As long as the WDM/OTN equipment vendors and the mainstream storage equipment vendors have tested the docking certification, the WDM/OTN system will meet the diversified requirements of the data disaster recovery system.
[off-site disaster recovery]
The off-site backup solution backs up the data to relatively distant city, across the geological disaster radius, and can eliminate the damage of business data caused by earthquakes and other geological disasters. Adopt MSTP, WDM and other technologies to build off-site disaster recovery system.
Layer 2 interconnection technology between data centers
In the traditional data center server area network design, the scope of the Layer 2 network is usually limited to the network aggregation layer. By configuring the VLAN across the aggregation switch, the Layer 2 network can be extended to multiple access switches. In recent years, server high availability cluster technology and virtual server dynamic migration technology (such as VMware’s VMotion) have been widely used in data center disaster recovery and computing resource allocation. These two technologies not only require large Layer 2 networks access in the data center, and requires a wide range of Layer 2 network extensions between data centers.
Three interconnection links are usually deployed between data centers, each of which carries different data and functions differently, and the three links are logically isolated from each other.
Server high availability cluster
Cluster refers to the use of Cluster software to associate multiple servers in the network together, provide consistent services and act as a logical server externally. Most vendors (HP, IBM, Microsoft, Veritas, etc.) cluster software requires two-tier network interconnection between servers. Deploying servers of clusters in different data centers can achieve disaster recovery of application systems across data centers.
Design Essentials of Spanning Tree Protocol
When STP domain crosses the access layer of multiple data centers through wide-area link, the failure of one center will cause network oscillation of multiple centers. Therefore, when considering the STP deployment of data center Layer 2 interconnection, it is recommended to limit the STP domain to each data center.
Switch virtualization technology
For the data center server access layer, two access switches are used to access the servers of the same type of service system to meet the uplink requirements of the server’s dual NICs. This topology is usually configured as a VLAN spanning aggregation and use STP to avoid the Layer 2 loop. Virtualization integration of the network aggregation layer is necessary when network simplification is done using switch virtualization, as this is the key network layer that eliminates STP.
The server’s Layer 2 access solution is usually adopted in the data center to achieve flexible expansion capabilities. With the increasing requirement of business continuity and flexible scheduling of computing resources, enterprises will inevitably face the problem of spreading two-layer network among multi-data centers.