High availability of New Exchange 2013 features (top)

Microsoft in Exchange high-availability scenarios with the introduction of Exchange 2013, the DAG scenario is becoming more and more mature. Exchange's highly available scenarios vary greatly in the exchange2003 to Exchange 20,103 editions, in the Exchange 2003 era, when Microsoft introduced shared storage, combined with the OS failover cluster, launched a host-level high availability, in which era, High availability is a very expensive solution, the reason is very simple storage is not cheap; in the era of Exchange 2007, Microsoft introduced the LCR, CCR, SCR three cluster scenarios, really widely used in the enterprise or the latter two scenarios, namely CCR and SCR, where CCR does not require shared storage, use is relatively extensive; but in the era of Exchange 2010, Microsoft did not retain and inherit the previous version of the high-availability scenario, but rather the CCR and SCR cluster solution to consolidate, and introduced a new high-availability scenario Dag,dag High-availability scenario one to continue to the exchange 2013, it is believed that the next Exchange version will continue to use this highly available scenario.

Recalling the rapid growth of exchange in the Enterprise mail market over the past few years, which occupies more than 70% of the Enterprise mail system, Exchange's high availability has contributed to its role as the preferred messaging system for large enterprises. There are so many highly available programs in Microsoft that really use very mature and extensive to be considered as shared storage failover clusters (SCR is also part of this scenario), but in these scenarios, there have been problems in the actual application of the enterprise, but also reflected in some aspects of their shortcomings, So Microsoft needs to be constantly revised in the next generation of new versions. Below, let's analyze the shortcomings of each generation (the pros are clear, no need to describe them too much):

One, shared storage-based failover clustering:

This cluster solution is not only applied to Microsoft's Exchange 2000/2003 business system, but there are other Microsoft applications (such as SQL Server) and third-party Windows-based business systems, the advantages of which are high availability, there are many shortcomings, It is host-level high availability, it requires shared storage, and even it does not fully utilize hardware resources to scale horizontally. What is high availability based on the host level. To give a very simple example, the Enterprise has 2 Exchange 2003 configured high availability, when the 1 active node host crashes, the network is interrupted, the application system will automatically activate the passive node, become the active node, all the databases go to this node to run, but there is a problem, If a database in the active node has a problem, or a database to perform maintenance, the business system must be interrupted, because its copy is a single share, and then the passive node as long as no failover, basically idle state, there is a waste of hardware resources, moreover, when the active node hardware resources are insufficient, The passive node is also unable to distribute the task; the last point is that there is a single point of failure in this high-availability scenario, and it is certain that storage at the hardware level provides a dual redundancy mechanism, but this does not mean that a single point of failure can be eliminated, such as storing over-insured replacement storage, storage hardware configuration modifications, or upgrades, Some actions may affect exchange business systems, and most crucially, this high availability does not allow disaster tolerance across sites.

Second, CCR-based failover clustering:

Exchange 2007 provides a CCR cluster for high availability, no shared storage for CCR clusters, and multiple replicas, but CCR clusters still have flaws, and CCR clusters, while having a lot of replicas, remain host-level failover clusters. When only one database copy of the active node is corrupted, it must be that all databases are moved from the active node to the passive node, and the passive node is activated, but all passive node databases are activated, and second, the CCR cluster is still unable to achieve cross-site disaster recovery and the use of passive node hardware resources is wasted. These shortcomings are publicly acknowledged by Microsoft, but in fact the CCR cluster compared to the coexistence of storage failover cluster, there is a technical immature flaw, in the past we helped fix the exchange error, when we found that some enterprises use CCR cluster, when the active node database fails , the passive node cannot be activated, and of course the cause of this problem may be a variety of reasons, and there are no more descriptions here.

Third, Dag-based failover clustering:

At the time of Exchange 2010, Microsoft introduced a new cluster approach, a database-level failover cluster dag. Although Exchange 2013 and Exchange 2010 are both database-level failover clusters, there are significant differences in many areas, first of all, we look at the changes and characteristics of the DAG in Exchange 2010. First, Exchange 2010 introduced a DAG (data availability Group) method for high availability, and dags compared to CCR, with 22 major improvements. 1. It is a database-level cluster, not a host-level cluster, it can distribute a large number of databases on different Exchange servers, each server above the database has both active and passive copies, the full use of hardware utilization, and when a database corruption, You only need to activate the passive copy of this database instead of transferring the entire host; 2. The CCR deployment of Exchange 2010 does not require additional failover clustering, although the bottom-up of failover clustering is significant, in the past, because of the high availability of deploying Exchange, it was first installed to configure failover clustering before deploying Exchange. Once exchange has been deployed, it cannot be added to a clustered environment, and only redeployment can be put into a clustered environment, and in Exchange 2010, a highly available deployment of the DAG is very simple, and any existing exchange 2010 servers can quickly become DAG member servers and vice versa. At the same time, it is also because of this feature, it can achieve cross-site disaster tolerance.

Of course, the Exchange 2013 DAG is still imperfect, and Microsoft has revised and improved the DAG again at the time of the launch, and what are the drawbacks of the exchange the DAG, what has changed in the Exchange And take a look at the next blog post on the high availability of Exchange 2013 features (below).