Shulou(Shulou.com)06/03 Report--

Extended clustering is the main application scenario of Windows Server 2016 storage replication. Through the combination of storage replication and WSFC, cross-site cluster storage replication is realized, which helps enterprises better achieve cross-site disaster recovery with lower RTO RPO, and ensures that when the site fails over, the transfer will not fail because of storage.

In fact, Microsoft did not first put forward the concept of extended cluster. VMare VSAN,IBM SVC put forward this concept as early as a few years ago, and each manufacturer has its own practical understanding of the concept of extended cluster.

Take the VSAN extension cluster as an example. For VSAN, the extension cluster is an extension of super-converged storage nodes, extending the original rack in the computer room to multi-campus in the same city, or remote cluster architecture. After realizing the VSAN extension cluster, two copies of the virtual machine storage on the VSAN will be stored, and each component can be stored in a primary site, a secondary site, a primary site and a secondary site. "there are two copies of each piece of data, each ensuring that one copy is replicated to another site, while the virtual machine reads storage optimized, extending the cluster architecture, where each virtual machine reads storage from the local site, combines it with DRS, and fails over from DRS to the appropriate site."

VSAN extends the characteristics of the cluster architecture

1. Save storage costs, and extended clusters can be implemented entirely by local VSAN storage

two。 Virtual machines are bound to each site to ensure that virtual machines normally run at the site where they should be run.

3. Combined with witness components to achieve automatic failover, if the virtual machine site is down, it can be restarted at another site.

4. It is realized by the super fusion function itself, without the help of other software.

5. To achieve double active users, not one site master, the other site is completely unavailable, both sites can store virtual machines normally, and the virtual machine is replicated to the other site

6. There will be at most one copy of each component and cannot be copied to multiple sites

7. Will occupy 50% of the total resources, reserved for disaster recovery, this part of computing resources and storage resources need to be reserved, otherwise the virtual machine cannot be fully started in the event of a disaster.

Microsoft's extended cluster is different from the concept put forward by VSAN,IBM SVC. In fact, Microsoft's extended cluster is not the cluster itself, or super-converged software, storage virtualization software, but the combination of the storage replication function on the system and the cluster function to achieve disaster recovery on the basis of high availability, and the combination of the two to achieve business continuity.

As we all know, Microsoft cluster itself supports multi-site deployment. Lao Wang and everyone have specifically mentioned before that Microsoft multi-site cluster deployment needs to consider network, arbitration, storage. In storage, Lao Wang also talked to you about the importance of storage replication. In traditional clustering, two nodes are connected to a shared storage, but in the case of multi-site, you need to implement that both sites have storage. Because if you store in one site, if you find a site-level disaster, even if another site can take over, but because there is no storage, the same cluster cannot operate, so an important aspect of multi-site clusters is to achieve storage replication, which used to be implemented by devices, or third-party software, such as Starwind,SIOS,Symantec VVR.

Microsoft has implemented block-based storage replication in Windows Server 2016, which can be achieved only by adding functions to the operating system.

For Microsoft extension cluster, it combines storage replication with clustering, using asymmetric storage architecture, that is, site 1 connects to site 1 shared storage, site 2 connects to site 2 shared storage, and the storage size on both sides is the same. if you meet the requirements of storage replication, you can extend the cluster.

The configuration of Microsoft extension cluster can be completed in the graphical interface of Cluster Administrator. It will configure the disks that meet the requirements of both sites for storage replication, and support the deployment of multiple replication groups in the same cluster to achieve multi-master and dual activity. when one of the sites fails, the extended cluster will automatically fail over, promoting all the replication group storage at the other site, and then the cluster application will be online at the other site. Because a failover cluster is used, the Microsoft extension cluster has the lowest RTO. After a failure, the cluster will automatically complete the failover without human intervention. If the synchronous replication architecture is used, zero RPO loss will be used, and if the asynchronous replication architecture is used, data loss may occur.

The main difference between Microsoft extended Cluster and Microsoft Hyper-V replication is that

1. Extended cluster is automated failover, Hyper-V replication requires manual

two。 Extended cluster can only recover to the most recent point in time, Hyper-V can recover to multiple optional points in time

Microsoft extends the features of cluster architecture

1. At present, we still need to use asymmetric architecture, that is, both sites are connected to shared storage, and local disks and SDS architecture cannot be used. The future version of maybe will be changed.

two。 Using two sets of asymmetric shared storage, the underlying layer can be SAS JBOD (which can be used with storage space to support SDD HDD hybrid architecture), SAN, Share VHDX or iSCSI, and persistent retention is required

3. Each replication group requires active and destination data disks and log disks

4. Fully implemented in windows server without the need for other software

5. It is the combination of storage replication technology and cluster technology, which can automate failover and storage switching.

6. In an extended cluster architecture, the source data disk must be a CSV or a traditional file server cluster role before it can be replicated

7. Multiple replication groups can be established to achieve multi-master double activation.

8. Storage replication technology will occupy 50% of the total resources of the cluster and be reserved for disaster recovery. This part of computing resources and storage resources need to be reserved, otherwise a disaster cannot be fully started.

9. Mainly used for file server load and virtualization load

10. Support for planned unplanned failover storage

11. It can cooperate with cluster site awareness technology, cluster placement technology, priority local site failover, read optimization, etc.

By comparison, we can see that the two types of extended clusters have their own advantages, but in the final analysis, they are all for the purpose of achieving high availability of cross-site cluster storage, so we can give a preliminary definition of extended clusters for the time being. on the basis of realizing cross-site clustering, the use of device replication technology, or super-fusion technology, or replication technology to achieve high availability of storage to ensure that when the site fails. Disaster recovery is not affected by storage.

Extend several categories of cluster storage processing

1. Device replication: represented by EMC,Netapp and Huawei

two。 Third-party software replication, represented by Symantec,SIOS,Vision,Starwind

3. Hyperconvergence or storage virtualization replication: VSAN,IBM SVC

4. Native replication of server operating system: Microsoft extension cluster

Microsoft extends the configuration requirements of clusters

1. Active Directory domain environment, which provides Kerberos verification of each node in the replication process.

two。 Each Site node connects its own Site storage to ensure that each Site storage is not visible to another Site

3. Each Site replication node requires at least two disks, one data disk and one log disk

4. The format of data disk and log disk must be GPT. MBR format disk is not supported.

5. Both data disks and partitions must be the same size, maximum 10TB

6. Both log disks and partitions must be the same size, at least 8GB

7. The source data disk needs to be configured as a CSV or cluster role

8. Storage replication uses port 445 (SMB-replication transfer protocol), port 5895 (management protocol of WSManHTTP-WMI / CIM / PowerShell), port 5445 (iWARP SMB-only required when using iWARP RDMA networks)

Planning recommendations for Microsoft extension Cluster

1. Considering RTO / RPO and cost, if it is a critical application, you can use the extended cluster synchronous replication architecture, which ensures the lowest RTO and zero data loss RPO, but requires higher bandwidth, and synchronous replication recommends that the latency of the two sites does not exceed 5ms, or the distance does not exceed 30km, so the synchronous replication extension cluster is suitable for networks with different campuses in the same city with high bandwidth and low latency. The availability of the application can be ensured as much as possible. If the cluster application is not critical and can accept data loss for a short period of time, then you can consider the extended cluster architecture of asynchronous replication. the latest windows server 2016 already supports asynchronous replication to extend the cluster. In previous versions, only synchronous replication was supported. The advantage of using asynchronous replication to extend the cluster architecture is that it does not require high bandwidth, acceptable latency, longer distance, cross-region, or multinational. The disadvantage is that if a failure occurs suddenly, the data may not be replicated to the secondary site in time, resulting in data loss, so engineers need to choose the appropriate architecture based on the actual enterprise situation, whether to use synchronous replication to extend the cluster, or asynchronous replication to extend the cluster, or hyper-v replication, ASR, or other products.

two。 It is recommended to use SSD or NVME SSD for the log disk. Storage replication first writes data to the log disk. Good log disk performance can help improve write efficiency.

3. It is recommended that you plan for larger log space, which allows faster recovery from larger outages, but consumes space costs.

4. Synchronous replication extends the cluster to prepare reliable and high-speed network bandwidth. It is recommended that 1Gbps start, preferably 10Gbps, network card supports RDMA better, synchronous replication scenario, if the bandwidth is insufficient, the write request time of the application will be delayed.

5. In the actual scenario, it is recommended that at least four nodes extend the cluster, cooperate with site awareness technology to achieve normal local site transfer, and transfer to the secondary site in the event of a disaster.

Extend other Microsoft technologies that clusters can integrate

Deployment: Nano Server,SCVMM

Management: PS,WMI, Cluster Administrator, Honolulu,SCOM,OMS,Azure Stack,Azure ASR,DPM

Integration: Hyper-V,SOFS,SMB Multichannel,SMB Direct, duplicate data deletion, ReFS,NTFS

Thoughts on the Integration of Microsoft extension Cluster and other functions of WSFC 2016

With the ability to extend clustering, engineers can better think about the design of multi-site clusters

For example, with site awareness, site awareness features are stored, so that failover within the same site is always a priority.

Adjust cross-site failover detection parameters with site heartbeat detection function

Combined with VM elasticity technology, storage elasticity technology realizes instant break processing.

Cooperate with cloud arbitration technology to realize extended cluster witness

Microsoft extended Cluster implementation

Environment introduction

This experiment simulates the architecture of two sites, Beijing site and Tianjin site, each node has a server and an ISCSI, each node connects to its own site storage to achieve an extended CSV-based cluster, the cluster hosts Hyper-V highly available virtual machine roles, normal storage and virtual machines operate in the primary site, and the primary site is transferred to the secondary site in the event of a disaster.

AD& Beijing ISCSI

Lan:10.0.0.2 255.0.0.0

ISCSI:30.0.0.2 255.0.0.0

16Server1

MGMT: 10.0.0.3 255.0.0.0 DNS 10.0.0.2

ISCSI:30.0.0.3 255.0.0.0

Heart:18.0.0.3 255.0.0.0

Tianjin AD&ISCSI

Lan:10.0.0.100

ISCSI.30.0.0.100

16Server2

MGMT: 10.0.0.4 255.0.0.0 DNS 10.0.0.100

ISCSI:30.0.0.4 255.0.0.0

Heart:18.0.0.4 255.0.0.0

At present, each node has been connected to the ISCSI storage of each site and formatted as GPT,NTFS disk, 10GB data disk and 8GB log disk respectively.

16server1

16server2

Install failover clustering, storage replication, and file server role functions for each node

Similarly, before implementing the extended cluster, it is recommended to test the environment first. The test process uses the Test-SRTopology command to complete the test, which can be used after completing the storage copy function. The testing process will evaluate whether the existing environment meets the storage replication requirements, check whether the disk size, partition size, bandwidth meets the requirements, log size, replication IOPS, initial replication performance, etc. Finally, according to the results of the evaluation, the html report will be produced.

To execute the Test-SRTopology command, you need to generate IO for the disk to have an effect. Here Lao Wang uses the Diskspd command to generate an IO test.

Diskspd download address: https://gallery.technet.microsoft.com/DiskSpd-a-robust-storage-6cd2f223

Diskspd.exe-c1m-d300-W5-C5-b8k-T2-O2-r-w25-h s:\ test.dat

Generate test report

Test-SRTopology

-SourceComputerName 16server1 # Source computer

-SourceVolumeName S: # Source data disk

-SourceLogVolumeName R: # Source log disk

-DestinationComputerName 16server2 # Target computer

-DestinationVolumeName S: # destination data disk

-DestinationLogVolumeName R: # destination log disk

-DurationInMinutes 1 # specifies the test time, 10-30 minutes is recommended for production environment

-ResultPath C:\ SRTest # report generation path

Wait for the test to complete, and open the report path to see the storage replication test report in html format. The report will show whether the current environment meets the basic requirements of storage replication, whether the performance meets the expectations, and if not, how to make adjustments. It should be noted that this test must be meaningful only when IO is generated on the data disk, otherwise the test data will not be obtained.

After the test is complete, we can implement the extended cluster.

The idea of implementation is as follows

Create a cluster

Add Cluster disk

Add source data disk as CSV or cluster role disk

Execute the Cluster disk replication Wizard (extend Cluster Wizard)

Select destination data disk, log disk

Select the source log disk

Select synchronization mode

Select synchronization initialization step

Create cluster SRcluster, configure cluster quorum for file share quorum, or cloud quorum, or quorum disk other than stand-alone replication

Just created the cluster, open the disk, you will find that there is not a single disk, because we neither open S2D nor use a shared disk, so it is empty by default

If we need to configure the extended cluster, we need to enter an additional command so that the cluster can read all asymmetric shared disks

Get-ClusterAvailableDisk-All | Add-ClusterDisk

After the input is complete, all disks can be seen in the cluster, and because we are the architecture of asymmetric disks, two disks should always be unconnected, because not all disks are visible to all nodes

Add the source data disk to CSV, or assign the traditional high availability file server role to the source data disk

At the cluster shared volume that has been added, right-click replication-enable

Start the extended Cluster configuration Wizard and select the target data disk

Select the source log disk

Select the destination log disk

Select the initial synchronization operation and specify whether to merge or overwrite the destination side by the source side

Configure replication mode, synchronous replication or asynchronous replication. See Lao Wang's first Storage replication blog about the difference between synchronous replication and asynchronous replication.

Configure consistency groups, choose to optimize sorting performance, or enable write order. If you plan to deploy SQL FCI On CSV by StorageReplica or other cluster applications that require write order, you must choose to enable write order.

OK,We Done it! After running the wizard, we can see the changes in storage in the cluster.

Previously unavailable disks become SR groups, replication roles are displayed, and source site log disks are automatically promoted to CSV

You can see the extra storage column below the disk information, where you can see the replication status of the current storage replication.

After initializing replication, the normal state of replication should always be continuous replication

After the failover, storage replication and cluster convergence in the test plan can be said to be very intelligent and convenient, for example, if we implement storage replication through two nodes and run CSV to provide services, if we know that we need to do maintenance, we can directly move the source data disk and log disk to the destination disk, and then set the node to maintenance mode, and then we can carry out maintenance operations for the source site.

Click the source data disk log disk and select move to 16server2

After moving, you can see that the current storage replication has completed the planned maintenance inversion, and the 16server2 becomes the source and the 16server1 becomes the target. If other roles are hosted on the 16server1, you can remove it for maintenance.

Although we can also see the storage content in the CSV on 16server1, please note that 16server1 looks at CSV and sees the content provided by 16server2 through CSV redirection coordination. Because we have moved the storage replication to 16server2, the 16server1 source master node will not be able to access the storage. At this time, if the application is still running on 16server1, it will operate in the way of CSV redirection, and the efficiency will be very low. Therefore, if you perform the reverse operation of storage replication, it is recommended to remove the role on 16server1 as soon as possible, and then return to the online role after maintenance.

After we get the CSV, we can run the cluster load on it. It is recommended to use Hyper-V,SQL 2014 or later, or directly use the traditional high-availability file server. Here, the official website does not say that it supports SOFS, but only that it supports the traditional high-availability file server. Lao Wang guesses that it may be due to the switching of storage replication that SOFS cannot complete transparent failover, so it is not fully supported yet. Maybe will change in the future.

To sum up, Microsoft extension Cluster is nothing more than two architectures.

Super convergence, storage replication node itself and then run Hyper-V or SQL to achieve high computing availability and storage disaster recovery

Converged, storage replication node itself provides file server UNC path for front-end use

In this example, we try to install a virtual machine in the cluster and run on the data disk CSV. Remember that in a single replication group, only source-side data disks can be used, and other disks cannot be used.

Let's first simulate a storage failure. The current data disk CSV runs on 16server1 and the virtual machine runs here. We simulate a storage disaster and disable ISCSI directly on the ISCSI server connected to 16server1.

As you can see, the cluster can sense that the storage replication master node cannot connect to the storage offline, and automatically switch the storage to the 16server2 master node immediately, always ensuring that there is storage on one side that can be read and written.

For virtual machines, due to the 2016 VM storage bomb, the loss of storage for the virtual machine will not cause the virtual machine to crash, but will freeze the virtual machine IO, set it to a paused state, and re-release IO if the storage is restored within a certain period of time.

If you turn off the VM storage bomb and try again, the virtual machine detects that the storage is missing, and because the CSV volume is used, the virtual machine will continue to run on the 16server1, but will use CSV redirection to access the 16server2 storage because 16server1 has lost the connection to the storage.

Through this experiment, we can understand storage replication technology, VM storage resilience technology, CSV technology and virtualization technology together.

The extended cluster can sense storage failure and fail over. When one of the Site nodes loses contact with the storage, it will automatically switch the primary site storage to the secondary site for reading and writing.

2016 when VM is enabled by default, the original intention is to ensure that when the storage is cut off instantly, the business will not be affected, the IO will be frozen and the recovery will be released immediately.

If there is no instant disconnection from your VM to storage, then you can turn it off to VM. When VM detects that the local storage is missing, CSV will play a role in redirecting IO to other nodes with storage access qualifications, but note that the performance of the virtual machine will be significantly degraded at this time. It is best to move the virtual machine to the site where the current storage group is alive.

VM storage bomb is mainly used to deal with the problem of instant disconnection, but if it is not restored for a long time, it will also extend the downtime. Therefore, it is recommended that if there is no instant disconnection scenario, turn off VM storage bomb and allow the virtual machine to run in CSV redirection, or move to the main site of the storage group after transfer.

Next, we simulate the disaster of the whole site, and the computing and storage resources of the main site are not used. Stop the ISCSI server and shut down the primary node.

As you can see, the first storage is automatically transferred to 16server2 to provide read and write

Virtual machines are also automatically converted to 16server2 to provide services.

This is to extend the charm of the cluster and achieve double disaster recovery of computing and storage resources, which can allow storage and computer errors without affecting the business. When a disaster occurs at the site level, the primary storage of the storage replication above will be automatically transferred to the secondary node to provide services, and the hosted SQL, virtual machines and file server resources will then fail over and go online.

When the primary site is restored, the storage replication reversal will not be performed automatically, and the primary node of the replication group will still be responsible for the previous secondary node. If you want to reply, you can manually move the CSV volume on the interface.

After the primary site is restored, the storage group is still at 16server2 as the primary site

Select to manually move cluster shared volumes and reverse replication back to 16server1

At this time, the virtual machine will not automatically move back to the main site, but will continue to run in 16server2 in the way of CSV redirection. It needs to be moved back to 16server1 manually. If it cooperates with site awareness and storage site awareness, CSV can sense that the site is back and move back to its own site. After 1 minute, the virtual machine feels that it and CSV are no longer the same site, and follow CSV will automatically move back to the site. Achieve virtual machine resources and site binding, always run on the site that should be run, and always avoid the problem of CSV cross-site redirection.

Three points to pay attention to

By default, the site failure virtual machine will not fail over immediately, because 2016 of the VM bomb × × ×, it thinks that a short instant break does not need a failover, so it will not fail over for a period of time. This function is turned on by default. If you find that the virtual machine does not find the transfer, but out of a state that is not monitored, you can manually remove it directly, or turn off the VM bomb × ×, about VM play × × × introduction. Please refer to Lao Wang's article https://blog.51cto.com/wzde2012/1963604

For site failures, virtual machine resources are usually rebooted at another site, unless it is too late for normal shutdown, can be released from storage, or real-time migration, otherwise, if it is a direct power outage, it will only be rebooted at another site.

"extended cluster non-transparent failover, with a delay of 10-30 seconds when failing over at the site level, depending on the quality of the network, because storage needs to be transferred before roles are transferred."

New features of WSFC2016 need to be taken into account when implementing extended clusters to determine whether the transfer results are in line with expectations.

Through the above two experiments, we can see that the extended cluster can handle three levels of disaster.

1. Can sense storage failure: choose to face VM storage elasticity or CSV redirection. If the virtual machine resource is running and the connection to the storage is suddenly lost, it will enter the frozen state by default in 2016. Freeze all the IO of the virtual machine, wait for the storage to recover, and then release the IO. This design is designed to avoid the storage flash problem, which is not suitable if your environment does not have a storage flash break, because during the freeze period. All IO cannot be done. On the contrary, if VM storage elasticity is turned off for the virtual machine, the virtual machine will directly enter the CSV redirection state. Although IO needs to be forwarded in east-west direction, IO can still be performed although it is slow, and specific choices need to be made according to the actual scenario. Only Hyper-V resources face this problem of VM storage resilience and CSV redirection, but not for SQL and file server loads, which directly fail over or redirect.

two。 Can sense node failure: if a single node is down, it will automatically transfer the primary storage copy, role or virtual machine carried by that node.

3. Can sense site failure: if the entire site is down, the primary storage copy, role or virtual machine hosted by the site will be automatically transferred.

Optimization suggestion

Considering network factors, refer to the content about multi-site cluster network mentioned in Lao Wang's disaster recovery blog.

Combined with WSFC 2016 site awareness, storage site awareness, preferred site

As recommended by Microsoft, the best practice is to deploy an extended cluster of at least four nodes, two nodes at the local site, and two nodes at the remote or city site.

# configure site failure domain awareness to achieve failover within priority sites

New-ClusterFaultDomain-Name Beijing-Type Site-Description "Primary"-Location "Beijing Datacenter" # create a Beijing site failure domain

New-ClusterFaultDomain-Name Tianjing-Type Site-Description "Secondary"-Location "Tianjing Datacenter" # create Tianjin site fault domain

Set-ClusterFaultDomain-Name 16server1-Parent Beijing # add a Beijing node to enter the site failure domain

Set-ClusterFaultDomain-Name 16server2-Parent Beijing

Set-ClusterFaultDomain-Name 16server3-Parent Tianjing # add Tianjin node to the site fault domain

Set-ClusterFaultDomain-Name 16server4-Parent Tianjing

# configure CSV follow Site and apply Follow CSV

Get-ClusterSharedVolume | Get-ClusterGroup # get the CSV group name

(Get-ClusterGroup-name CSVClusterGroupName) .PreferredSite = "Beijing" # configure Beijing site CSV follow Beijing site

(Get-ClusterGroup-name CSVClusterGroupName) .PreferredSite = "Tianjing" # configure Tianjin site CSV follow to add site

After this optimization, we will get this effect.

The failure domain is shared storage on this site: storage replication is automatically transferred to other sites, and if CSV is transferred, the virtual machine will follow CSV in the past to avoid facing CSV redirection and VM storage resilience.

The failure domain is a single host node at this site: virtual machine or cluster roles are automatically transferred to other hosts at the same site

The failure domain is the shared storage of this site and all nodes: storage replication is automatically transferred to other sites, and resources are automatically started at other sites with storage.

Storage replication supports the creation of multiple replication groups in a single cluster. It should be noted that one replication group has at least four disks, and two replication groups need to prepare eight disks.

By deploying two replication groups, we can achieve the double activation of multiple replication groups. For example, Beijing is the master of the first replication group, Tianjin is the master of the second replication group, and Beijing is the backup of the second replication group.

In this way, the cluster computing resources can be better utilized, and half of the storage resources are consumed.

If you deploy a replication group with multi-master and double activation, it is recommended to use site awareness and storage site awareness to transfer resources to the local site first, follow CSV, and avoid CSV redirection.

A typical scene

1. Realize the double activation of multiple replication groups of multiple SQL instances, and use multiple replication groups on a set of WSFC clusters to ensure the double activation of multiple SQL instances.

two。 Super converged architecture, the node acts as both hyper-v node and storage replication node, and can handle disk level, node level, site failure

Extend cluster troubleshooting:

Store replication event logs: application and service logs-Windows-StorageReplica-Admin

Storage replication performance counter pointer

Cluster Administrator Lo

Cluster event Manager Log

ClusterLog

Dumpfile

Through the above introduction, I believe you have seen the extension of the cluster function, which is a combination of Microsoft WSFC and storage replication functions, which can be perfectly integrated in disaster recovery time, automatically complete storage replication switching and cluster role switching, and be able to deal with disk failures, node failures, and site failures.

Hope that storage replication can be optimized in the future.

1. Support for local disk, SDS architecture

two。 Transparent failover can be achieved

3. Optimize disk locking problem

4. It can be integrated with the WSFC2016 VM load function. If the VM load can sense the site, it can allow the application to load balance within the site and follow the site sensing and storage site sensing rules. At present, once the cluster uses storage replication, it is easy to use the VM load function, because the VM load balancing function currently cannot sense the site, so it is possible to migrate the virtual machine to other sites. CSV will not follow the migration. Therefore, it will cause CSV cross-site redirection. If VM balancer can sense the site, then extend the cluster, and load balancing can be performed within each site, automatically controlling the load balancing of each node.

5. Can support one-to-many storage replication, and cluster-to-stand-alone extended replication

6. Can be integrated with more Microsoft applications

In Microsoft's entire enterprise application ecosystem, apart from storage replication, there are many other replication products. What are the differences and matching points between storage and comparison?

The difference between Hyper-V replication and storage replication

Hyper-V is also supported in Standard Edition, while Storage replication only supports data Center Edition

Hyper-V replication uses port 80 or 443, storage replication uses SMB 445

Hyper-V can support the selection of certificate verification or non-certificate during replication

Hyper-V supports multiple recovery points and can choose to recover after a disaster

Hyper-V replication can be all disks of a virtual machine. Storage replication does not support replication of system disks.

Hyper-V replication is designed for virtual machines to better handle application consistency issues

Hyper-V replication requires manual failover unplanned, and the storage replication extension cluster can achieve automatic failover.

To sum up: hyper-v replication and storage replication have similarities in many points. They are both storage-independent and disaster recovery functions. The difference is that storage replication is more focused on ensuring the high availability of the underlying storage, while hyper-v replication can better understand the VSS application of the virtual machine above. Hyper-v replication now has environmental assessment tools, extended replication, ASR, replication progress view. Relatively speaking, at the disaster recovery level, it seems to be more comprehensive than storage replication. The biggest difference between storage replication and hyper-v is that it can achieve automated failover natively, while hyper-v replication needs scripts or ASR to achieve automated failover. Using hyper-v replication can get cheap disaster recovery, but there will be delays in RTO and RPO during native disaster recovery. Using storage replication to extend the cluster can achieve the lowest RTO and zero RPO loss at the cost of high bandwidth and low latency networks.

Storage replication has more application scenarios than hyper-v replication. Storage replication can be used as long as there is OS. It can be used on Guest Cluster, any cloud platform, any virtualization platform.

Exchange DAG temporarily does not support the underlying storage replication architecture.

The difference and matching point between SQL Always on replication and storage replication

AlwaysOn replication is not just at the block level, it knows more about SQL

Replica can be read-only, but storage replication is not supported for the time being

Support for eight asynchronous copies or two synchronous copies

Backup of target copies is supported, and storage replication only supports backup of source copies

SQL AG requires SQL Enterprise Edition authorization. If there is no authorization, there is no way to implement SQL AG. At this time, you can cooperate with storage replication to achieve storage replication protection of SQL instances.

The difference between DFS FRS and storage replication

DFS replication is at the file directory level and storage replication is at the partition level

DFS only supports replication of files that are closed. There is no such restriction on storage replication.

DFS and AD site integration uses site topology, storage replication is not integrated with AD site

DFS is distributed and can be read by each node. Storage replication backup site cannot read it temporarily.

DFS can provide a unified external name, name access and replication functions are separated, storage replication does not provide a unified external name.

DFS is mainly used to copy closed files, information worker files, storage replication is mainly used in hyper-v, file server, SQL, private cloud scenarios

Storage replication technology itself is only a disaster recovery technology, which helps us to achieve disaster recovery without the help of hardware devices. With cluster technology, we can extend the cluster and help us ensure the integrity of site disaster recovery. However, storage replication technology is not a backup technology. You still need to backup the source data disk to prevent data from being deleted by mistake. It should be noted that storage replication only supports backup of the party that can be read and written on the source side. If you need to back up from the standby node, you need to perform reverse replication first.

The above is the content of the extended cluster in this article. I hope it can bring harvest to the interested friends!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.