Backup and recovery solutions based on the MDS 9000 Series

Purpose

This White Paper aims to discuss backup, recovery architecture and solutions, and introduce the features of the Cisco MDS 9000 series multi-layer guide and network switches.

Introduction

Global Enterprises with key business data on servers require continuous application availability. For a large amount of data generated by applications such as supply chain management SCM, Enterprise Resource Planning ERP, and customer relationship management CRM, enterprises must protect their security at all costs. The data must be stored on tape at least periodically to prevent data loss. However, as the amount of data increases, not only requires a larger storage capacity, but also requires faster and faster servers and more backup time. Generally, if data backup takes several hours to complete, data recovery takes the same time. For many enterprises, such a long recovery time is unacceptable, because it will lead to loss of revenue due to long service downtime. Therefore, in many cases, tape backup is only the minimum requirement for disaster recovery DR planning.

To ensure 99.999% availability of these enterprise applications, the high availability of each link must be considered in the storage design. All enterprises must develop a disaster recovery plan that resolves the problem of prolonged downtime and allows seamless recovery to standby sites in the event of a major fault. In the DR program, in addition to tape backup, enterprises often use replication technology to remotely replicate the entire data center. Therefore, in addition to restoring data from tape, the current recovery plan may also include the use of remote backup data centers for recovery. There are many factors that may cause faults, which are difficult to predict in advance, including:

• Device fault
  
• Application fault
  
• Human error
  
• Natural and non-natural disasters

Every enterprise must prepare for disaster recovery at any time and determine which key data must be continuously accessed. To determine the most important position, function, or application of an enterprise, business impact and risk analysis must be performed. As a ing image of the primary data center, the remote data center can provide comprehensive access in the event of a disaster. Apart from tape backup, many DR solutions include copying data images in real time. In addition, the replication technology provides many other functions to meet various application requirements. Although the replication technology can help enterprises quickly recover from catastrophic faults, errors may also occur during the replication process. Therefore, enterprises also need to use tape backup as a way to archive valid data. This article will mainly discuss the tape backup technology, architecture and various methods in the entire DR program.

Tape backup

In today's enterprise environments, most application servers are directly connected to dedicated tape drives through parallel SCSI connections. As the number of tape devices to be managed increases proportionally to the number of application servers, the deployment and maintenance of dedicated resources are expensive. However, direct-to-tape drives ensure performance because only one server uses the drive. In cost consideration, the Direct Connect tape drive is gradually transferred to the network backup mode, that is, the tape drive is mounted on the LAN and shared by multiple servers. Shows typical LAN backup methods. In this mode, both data and backup traffic are transmitted over the same LAN. Network Backup can improve tape utilization and manageability, but it brings about the following problems.

Figure 1

First, a large amount of backup data increases LAN traffic, which may reduce application performance. To reduce interference with data traffic, backup tasks are generally performed during off-duty hours. However, as the data volume increases, the backup time is getting longer and longer, and will gradually extend to work hours. At present, many enterprises require 24x7 operations, leaving backup time shorter and shorter. Second, sharing a LAN between backup traffic and application traffic may affect or even disrupt backup operations. Third, it is expensive to share a LAN between backup and data applications, because firmware upgrades or instability in one application environment will lead to downtime in another environment. In order to solve these problems, the Administrator proposed application and backup separation. As shown in the following figure, in the new implementation scheme, the customer will port to the non-LAN architecture and separate the backup traffic from the application. Currently, many customers have begun to deploy dedicated backup storage networks.

The implementation of the current backup process must be completed manually, which is not only laborious, but also prone to manual errors. Generally, backup errors are not easy to detect, which may lead to data loss and further failure. In addition, tape not only requires manual insertion, rewind, and removal during operation, but also needs manual transportation to the storage location after recording. Because of the high cost of manual management, robots can be used to improve tape management. Centralized backup can organize tapes together in pools, that is, multiple data backups can be performed on the same disk. After a tape is fully record, the backup will be performed using the blank band in the pool, improving manageability. Reducing the number of tape drives to be managed helps reduce costs.

The increase in backup will lead to an increase in application server operations, because data must be retrieved from the disk and then written to the tape. The application server is usually busy processing a large amount of data that is sensitive to latency and performance. It needs to use an additional CPU cycle to process data movement and planning. Therefore, implementing backup is too heavy for the application server itself. The purpose of the serverless solution is to eliminate the performance impact of backup on the application server itself. This architecture transfers the role of data mobile and connected man-in-the-middle to a dedicated backup server. A powerful backup solution must solve the following problems:

1. Optimized the use of backup resources, including tapes, drives, and operating hours;
   
2. Reduces the impact on application traffic;
   
3. Isolate the backup domain to reduce the CPU burden on the application server.

The following describes the most common backup solution implementation methods in detail.

Non-LAN backup

With the use of non-LAN backup, each application server can directly move data to tape devices through the storage network, instead of using a dedicated backup server. Using a well-known shared storage method, each application server can act as a media server and directly transfer backup data to tape. Each server then requests a tape drive and occupies it during Backup. In addition, the application server can use tape drives in tape library units rather than shared tape drives. Data Flow in the storage network reduces LAN traffic. In addition, tapes can be shared by multiple applications and shared to tape libraries and drives. At the same time, LAN is still used to transmit metadata, that is, to track the Data Location change table between the backup server and the client, while the actual backup data is transmitted through the storage network. Shows a typical non-LAN implementation solution.

Non-LAN backup separates data domains from backup domains, but does not reduce the CPU burden on the application server, because the application server still needs to obtain backup data from the disk. The serverless backup shown below can solve this problem.

Figure 2

Serverless backup

Serverless backup means that when the backup data is transferred from the disk to the tape, the data does not go through the application server itself, so data backup will hardly affect the applications on the server. Backup data can be moved from disk to tape without traversing the server. This process reduces the CPU burden on the application server because the application server is not involved in the backup process. The mechanism for direct data movement between disks and tapes is the SCSI extended replication command. Components that execute the extended copy SCSI command can be configured in the SAN switching domain Fabric) or server software. Current serverless Backup implementation methods, such as Legato Celestra Power and Veritas NetBackup, manage the entire process through the backup server. Shows the typical serverless backup process. Currently, most customers want to deploy independent backup SAN for domain separation. Therefore, these customers may not deploy this solution because the disk subsystem and tape drive of this method must coexist on the same SAN.

Figure 3

Backup industry and Product Overview

Many software vendors have developed centralized backup software, such as Veritas Netbackup, Veritas BackupExec, Legato Networker, and Computer Associates BrightStor ARCserve 2000 Advanced Edition. These products adopt a central backup method, that is, using a dedicated backup server to perform backup and directly connect to the archiving device. In addition, the Central backup method may contain several other software and hardware modules. When using the backup solution, you must understand the various components of the backup solution, as well as the role of the components and the impact on performance.

First, the central or primary server controls the entire backup environment, including indexes, backup plans, customer group definitions, and hardware configurations. In addition, the master server is responsible for recording backup issues and reporting them to the system administrator. This function is sometimes called a connection intermediary.

The second type of server in the central backup system is the media server or storage node. Media Servers are related to the type of backup storage media. Generally, they are tape devices connected through fiber channels or parallel SCSI connections. The Media Server is responsible for transmitting data to tape. The server accepts commands from the master server, including backup files and required media. In addition, Media Servers are also called data mobile devices.

The third and most important thing is to back up the client, which actually refers to the server being backed up. The client software is installed on each system on which the backup service is to be executed. Client software is usually installed on the master server and Media Server for self-Backup.

Main advantages of the SAN backup method

As mentioned above, to solve the performance bottleneck of shared LAN, enterprises have begun to deploy dedicated fiber channel storage networks. Other advantages of the storage network are as follows:

High Data availability

To improve availability and scalability, the storage network infrastructure connects to the storage subsystem using multiple paths, such as disks and tapes. Customers can implement improved disaster recovery solutions, especially in an open system environment. In previous solutions, disaster recovery for remote sites was generally achieved through communication between the LAN server and the server. In network mode, data can also be mapped between two storage subsystems to release expensive server and LAN resources. After transferring backup traffic to the storage network, you can limit the impact scope of intra-domain faults, that is, to prevent LAN data traffic faults from affecting backup traffic or backup traffic faults from affecting data traffic.

Low TCO

Storage integration allows multiple servers to share storage devices, reducing the number of tape libraries required by the data center. In addition, integration can simplify the redistribution of idle capacity among all servers to improve utilization and efficiency. In addition, after implementing the Enterprise Backup/recovery solution, the customer can also reduce the management and maintenance costs of tape drives connected to the server. If you manually manage each component, the cost is not only high, but also error-prone. If you use a robot to store remote tapes, you no longer need to manually carry the tapes, reducing the possibility of manual errors. In addition, it reduces the possible damage, loss or delay during the handling process, thus improving the reliability. A storage network can create a pool of backup resources, which are shared by multiple servers without the impact of a faulty backup device. In addition, with the scalability, availability, performance, and manageability of shared backup resources, the total cost of ownership (TCO) can be further reduced ).

Flexible backup methods

At present, many enterprises have extended their work hours to support global customers who require them to operate around the clock. There are two backup methods: Hot Backup and cold backup. Cold backup refers to the process in which application data remains unchanged during the backup process, while hot backup refers to the process in which application data is continuously updated during the backup process.

Hot Backup technologies such as writing, copying, and splitting image snapshots use images of raw data generated at a certain time point to implement online backup without affecting applications. Both the write replication and the image separation methods copy data blocks to idle storage to generate scheduled replication supported by most databases. Writing, copying, and splitting images not only manage the ing process of physical data blocks, but also manage the associations with file systems or databases. Storage subsystem vendors that support these methods include EMC Timefinder, HDS ShadowImage, and FlashCopy in the IBM enterprise storage server ESS) or modular storage server MSS using business continuous volume BCV.

The goal of Backup implementation is to shorten the backup time, reduce interference with user traffic, especially peer-to-peer backup, and improve performance. Hot Backup uses remote or local ing to ensure continuous application availability. The following describes the remote ing technology in detail.

Remote Data ing/Data Replication

Remote Data ing or remote replication is the most common method for rapid application and data recovery. This method creates a ING volume in the disk subsystem to provide protection. The volume used by the application is called the master volume, and the ING volume is called the backup volume. Two main remote replication technologies are implemented:

-Host-based remote replication software
-Storage controller-based remote replication of hardware and firmware

Currently, the most common Remote replication facilities available on the market include EMC's Symmetrix RemoteData FacilitySRDF, IBM eXtended Remote Copy XRC), IBM's Peer-to-Peer Remote Copy PPRC), HDS TrueCopy, Compaq Data Replication Manager DRM) and Veritas Volume Replicator VVR ).

Both ibm xrc and Veritas VVR solutions are host-based, software-assisted ing facilities, while PPRC, SRDF, TrueCopy, and DRM are hardware-controller-based remote replication methods. It should be noted that VeritasVVR) provides remote data ing through an IP network instead of fiber channel or ESCON enterprise serial connection.

Cisco disaster recovery products and solutions

As enterprises pay more and more attention to data availability, more and more resources are used to ensure continuous operation. As mentioned above, the purpose of a private network is to ensure the performance and security of backup applications. The smart storage network increases the level of backup and recovery to a new level. In addition, the remote data replication solution can significantly improve availability and meet the new requirements of enterprises through expansion. With the advanced technology provided by Cisco Systems, enterprises can establish end-to-end backup and recovery solutions and disaster recovery solutions in a safer, cost-effective, and scalable manner.

Figure 4CiscoMDS 9000 series multi-layer directors and network switches, from left to right: MDS 9506 guide, MDS 9513 guide, MDs9509 guide, and MDS 9216 network switch.

The Cisco MDS 9000 series multi-layer directors and network switches provide a storage network environment for enterprises and service providers, providing higher port density, switching bandwidth, performance, multi-protocol and reliability. In addition, the series of MDS 9000 products are also oriented to the heterogeneous storage lan that contains multi-vendor products in the entire storage environment. In these environments, the Cisco MDS 9000 Series Products can serve as a central system for providing interconnected and advanced services.

The MDS 9000 series includes the MDS 9500 series multi-tier guide and the MDS 9216 multi-tier network switch. The MDS 9000 series are modular systems that provide high port density and performance for data center applications. For remote data centers used for backup and disaster recovery, you can use an MDS 9500 multi-layer network switch that is small but has the same functions and services as the MDS 9216 series guide.

In addition to the functions that should be provided by switches, Cisco MDS 9000 Series Modular directors and network switches provide many other features and services, including virtual SAN, advanced ISL link aggregation, LUN partitioning, call home, high availability, and seamless firmware upgrade. One of the problems that plague most customers is the lack of manageability and tools that reasonably support storage networks. Cisco MDS 9000 Series Products include powerful embedded exchange domain Fabric) manager applications for configuration, monitoring, and troubleshooting of storage networks.

Virtual SANVSAN)

In many current environments, the backup solution is designed to establish an Independent Parallel Network for backup traffic. From the perspective of technology and operation, although this independent network method is costly, it can provide flexible, secure, and highly available backup solutions. The isolation between the tape storage network and the disk storage network can avoid the application-level impact caused by faults such as power reset of the tape library unit TLU. In addition, the independent network can also prevent the application domain device reset from affecting the backup. In fact, this method has become a practical method to isolate tape devices from the disk subsystem. Through this isolation, Fabric of the switching domain) the impact of configuration changes will be limited to local, so it will not affect the continuous deployment of the isolated SAN.

Although these independent storage networks help to ensure performance and reduce network-level interference, their implementation is expensive. They not only need to be configured with switches, but also increase management complexity. In addition, this method also wastes ports and increases the customer's financial burden. To solve these problems, Cisco uses virtual SANVSAN to completely change the storage network deployment mode. VSAN can establish a virtual independent network on the same redundant physical infrastructure, thus ensuring the security and isolation of different SAN domains. The Cisco MDS 9000 series multi-layer directors and network switches are able to create 4092 isolated VSAN topologies on one physical infrastructure. This method is a bit like a VLAN in Ethernet. VSAN is a revolution in SAN deployment. It uses practical Ethernet technologies and ease of configuration to provide features such as traffic isolation and Security in the SAN environment. The main features of VSAN include:

• VSAN uses the explicit mark on the fiber channel frames to isolate each other to ensure that no frame is missed from one VSAN to another;
  
• The domain ID and FC_ID can be reused in different VSAN;
  
• Each VSAN has its own set of exchange domain Fabric services, including partition servers and domain name servers;
  
• In each VSAN, you can create a partition set as in the normal exchange domain Fabric;
  
• Partitions cannot span multiple VSAN instances.

VSAN adopts mature Ethernet technologies to meet the performance and isolation requirements of Fiber-channel storage networks and provides the best solution. The role of partitioning is to protect edge access and restrict access and traffic flow between devices within the Fabric domain. In addition, VSAN can isolate all services provided to a certain switching domain Fabric, and "encapsulate" them in a VSAN to establish a logically independent switching domain Fabric ).

Figure 5

The storage network domain created with VSAN not only extends the existing security mechanism, but also integrates a vswitch into multiple virtual environments. Different VSAN instances can be completely isolated to ensure that device faults in a VSAN or Fabric in the switching domain are not unstable. These functions can be implemented without installing multiple physically isolated network switches. In addition, VSAN can improve port utilization, reduce the cost per port, and thus reduce the TCO of the storage network. This implementation is very effective because multiple user groups can be organized together to provide services by one vswitch, thus reducing the number of required devices. The Cisco MDS 9000 Series not only provides advanced edge security and services, but also reduces capital and operating expenses.

This architecture not only reduces the total number of SAN or exchange domain Fabric deployed in the data center, but also records the status of backup/recovery and remote data ing domains and application SAN separately. Considering the customer's data center and storage integration model, VSAN can be used to help enterprises implement low TCO solutions.

Multi-Protocol Support

To provide integrated multi-protocol functions, the MDS 9000 series multi-layer directors and network switches integrate FCIP and iSCSI into the same Switching Module. Multi-Protocol solutions provide IP connection methods for disaster recovery and other applications. After integrating multi-protocol services and high-speed and high-capacity fiber channel into one platform, the customer will be able to meet enterprise-level and department-level application requirements independently using the same platform. The integrated FCIP function provides access to remote data centers through TCP/IP to facilitate smooth transfer in the event of a disaster. Although the solution should be deployed depends on the application requirements, FCIP is applicable to most remote replication ing applications. FCIP is suitable for Establishing wide-area connections between two data centers through WAN links at various speeds.

The integrated iSCSI function is a low-cost host connection established using the existing IP/Ethernet infrastructure. ISCSI technology can economically and effectively expand the storage network to low-end servers, thus improving the usage of expensive storage and tape subsystems. This integration is the industry's first integrated fiber channel and IP Storage Solution built on a highly available browser platform. Low-end servers for department or workgroup applications can be easily integrated into the same storage network exchange domain Fabric) to simplify access to company resources. Currently, workgroup applications are no longer connected to fixed storage, nor will they be suspended due to capacity upgrades. In addition, the performance is not reduced, and the cost is much lower. In addition, iSCSI also provides centralized, shared tape, LAN-And serverless backup and recovery solutions for department or workgroup applications.

Obviously, Cisco provides an unparalleled range of outstanding solutions in the industry as it provides a set of smart storage network features and focuses on the availability requirements of data centers. To help enterprises solve business interruptions caused by unplanned downtime and planned upgrades, Cisco products provide many enhanced management features. With FCIP and iSCSI technologies, enterprises can not only reduce infrastructure costs, but also provide integrated management of the two technologies.

Figure 6

In addition, Cisco also provides synchronous ing applications for disaster recovery through products such as the ONS 15540 DWDM switch and the ONS 15454 SONET switch platform. DWDM can provide reliable Metro connections between two data centers, while SONET can provide high TDM bandwidth over long distances. Both technologies provide excellent Transmission Performance for data remotely replicated through fiber channel or FCIP. For details about Cisco's DWDM solution, visit the Cisco Web site: http://www.cisco.com/warp/public/44/jump/optical_platforms.shtml.

Summary

The Cisco MDS 9000 multi-tier guide and network switch series provide the best solution for disaster recovery, backup, and recovery applications. Using various new technologies, the Cisco MDS 9000 Series can meet various requirements of high, medium, and low-end enterprises. The Cisco 9000 Series provides end-to-end solutions for customers who deploy multiple applications in heterogeneous transmission environments. With the MDS 9000 Series, customers can seamlessly deploy new technologies and coexist with old environments. The Cisco 9000 series focuses on performance, port density, investment protection and management, and supports large-scale deployment of highly effective SAN solutions.