From the three centers to the dual live data center

From the three centers to the dual live data center

three centres of the two places

There are several forms of realization in the three centers of the two places, which is a typical case.

650) this.width=650; "title=" 3dc.png "style=" Float:none "alt=" wkiom1cnmltthgtlaadtvnzdz8s470.png "src=" http:/ S2.51cto.com/wyfs02/m01/7f/a5/wkiom1cnmltthgtlaadtvnzdz8s470.png "/>

In this case, under normal circumstances, the business runs on top of the equipment in the main room. Primary storage and secondary storage have a one-way synchronization relationship, that is, all the data changes of the primary storage will be replicated ① to the secondary storage in real-time , thus ensuring that two storage data is exactly the same. At the same time, in order to prevent the occurrence of extreme disasters, the data change of primary storage will be synchronized to the storage device of remote disaster room by asynchronous replication ②.

When the main center is interrupted for various reasons, the service can be switched to the secondary room by manual command or automatic software switching.

If the extreme situation occurs, the auxiliary engine room can not run the business, then the remote disaster room has a data storage, you can use it to restore business.

Note: It's just a way of achieving the three centers of the two places, and there are several other ways, such as: ⑴ The remote Disaster Recovery Center also configures servers to allow disaster recovery centers to operate when disasters occur; ⑵ 3 different types of storage topologies. But the basic principle is not very different, here do not repeat.

① synchronous replication ensures that data is fully consistent, but the bandwidth and latency requirements are high and expensive to use for proximity.

② asynchronous replication does not guarantee that the data is completely consistent, there is data loss, but the data transmission bandwidth and latency requirements are low, suitable for remote.  

Dual Live Data center

The advantages of the three centers are to prevent all kinds of damage to the disk array data (excluding software or manual misoperation) risk, the disadvantage is the high cost, and low efficiency of equipment use, especially the auxiliary room equipment can not be used in the normal operation of the business, waste is very large.

As a result, storage device manufacturers have developed dual-live data center technology to improve this shortcoming.

below we take HP XP7 disk Arrays and ORACLE RAC as an example, demonstrate this technical solution.

650) this.width=650; "Title=" dual live. png "alt=" Wkiol1cnmbzcsn47aafmlwcwx2g291.png "src=" http://s1.51cto.com/wyfs02/M00/ 7f/a3/wkiol1cnmbzcsn47aafmlwcwx2g291.png "/>

The core of this program is : Two storage mates, virtual out of a disk array (similar to the host cluster Software floating package technology), the host to the virtual disk array issued IO request, primary storage and secondary storage to work together to complete the host-to-virtual disk array IO request. Primary and secondary storage data bidirectional synchronization, through the internal mechanism to ensure data consistency.

The advantage of this scheme is that the host of the two engine rooms only see a virtual disk array, the internal synchronization mechanism of the two stores is completely transparent to the host, and the host application configuration is simple.

Because the primary and secondary storage has a certain physical distance, if the data synchronization link failure, there will be a "brain crack" situation, this time, the quorum disk is in effect.

The arbitration disk is a third disk device independent of the primary and secondary storage (it is not recommended to use the storage of the disaster room) by FC link is connected to the primary and secondary storage. When the data link of the primary and secondary storage is abnormal, the primary and secondary storage will decide which storage will continue to provide the service through the quorum disk, the storage of the service amount is locked, wait until the data link is restored, and the two array data is restored to normal after the synchronization is completed.

So, what happens if the arbitration disk fails? Very simple, two stores are locked and no service is provided. After all, the integrity of the data is the most important.

what does the quorum server in the diagram do? As the name implies, it is a server or virtual machine, which runs a dedicated program that provides IP Protocol-based arbitration services for primary secondary storage.

for HP XP7 or HDS G The quorum server is not required for series arrays. However, some device vendors do not use disk quorum for various reasons, but are arbiter servers, such as EMC Vplex or NetApp .

In addition, some vendors ' scenarios do not use virtual storage, exposing two physical storage to the host, and then synchronizing the data of the two arrays on the array by other means. I have some doubts about this approach, I hope I can get expert guidance later

Also, the market is not only based on storage implementation of the double live, but also based on the host software implementation of the dual-work, if done well, can meet the needs. But there is one thing that needs special attention: the "brain crack" situation, I think: no third-party arbitration equipment of the dual-live program is not strong enough to cope with the complex situation in the real environment.

on the "brain fissure", some people mentioned the "Lock competition" problem: when the main auxiliary computer room link is interrupted, the storage has the arbitration mechanism, Oracle RAC also have their own arbitration mechanism if it appears RAC lock mechanism to determine the host room equipment to continue to provide services, storage but to determine the secondary room storage to continue to provide services, will lead to "double live" into "double death."

such a situation is indeed a problem, but if you study it carefully RAC locking mechanism, we can avoid this situation by proper configuration, and we recommend the following:

RAC arbitration mechanism uses disks, we simply configure the quorum disk on a virtual disk array to avoid "lock-up contention". Because the RAC arbitration disk on the virtual array, the primary or secondary storage of any one is locked, it corresponds to the host computer room will not be able to access the virtual array lock disk, it is not natural to get the quorum disk recognition, continue to run.

This article is from the "splinters nonsense" blog, make sure to keep this source http://leesbing.blog.51cto.com/1344594/1769519

From the three centers to the dual live data center