Within a modern data center there are new and interesting challenges that the IT department needs to address to adapt to the new way of thinking regarding building services instead of focusing on an explicit product. The most important fact is that there will never be a product in the market that will cover or solve all the needs of a company or organization. Today's IT department must change the way it looks at solutions; a single product will never be the answer but combining different technologies and building up services will provide a much higher value for the organization or company regarding both time consumption and cost.
The modern data center is all about building services that empower the possibilities the different products bring to the table. By combining your backup software with the monitoring software and using the concepts of automation, you can create something that will develop with time. There is no whitepaper that you can simply download that will let you in on all the secrets to build a flawless Backup as a Service (BaaS), Restore as a Service (RaaS) or Disaster Recovery as a Service (DRaaS) solution.
System Center Data Protection Manager 2012 R2 (DPM) is the latest release and with it comes a new product group team in India. The team has in the few past months provided high scale development and improvements to the product and they will strive even harder to optimize DPM and also bring new features and supported scenarios to the table.
In this chapter, we will cover the pre-installation tasks for DPM 2012 R2. We also will provide you with the information you need to understand how DPM 2012 R2 works and get started with a well-planned implementation of your backup, restore, and disaster recovery scenarios.
For many years, the focus for protecting a data center has always been backup. There are many vendors and third-party solutions still claiming in the market that they are the best and most suitable backup solutions for your data center. If your purpose in performing the backup operation is being able to perform a restore, then the majority of the third-party software and vendors out in the market will not be able to help you. The most challenging part of providing a restore plan or a restore scenario for a service is that the third-party software and vendors are focusing on providing you just a backup, not the ability to restore. This is where DPM 2012 R2 comes in. It all started with the release of the Windows Server 2003, and with that release came the possibility of providing online snapshots using a new feature called Volume Shadow Copy Services (VSS). The APIs for the solutions were quickly adopted by third-party software and vendors but no one could, in the beginning, get it up and running in a fully optimal way. The idea regarding the VSS architecture was a great one and a big step forward regarding how to manage snapshots of different workloads like file, SQL, or Exchange in order to provide a consistent data snapshot.
Microsoft understood that if this architecture was to blossom in the way that they wanted it to, they needed to create backup and restore software themselves. In 2006, the first version of DPM was released with the focus only on protecting the file workload. A lot of companies tried it but few were adopting the new software since they still hosted and ran their data center with the traditional IT mindset.
During the past years Microsoft did not invest heavily in DPM, but under the new team that is driving the DPM solution from India, there have been a lot of investments, making DPM the primary choice for private cloud protection. Microsoft has invested in both developers and experts, pushing the product forward and making it a natural choice of the modern data center.
You may ask what is the thing that differentiates DPM from other backup software on the market. The answer is quite simple; restore based on supported and optimal scenarios with the power of easy integration building the BaaS, RaaS, and DRaaS for a modern data center. DPM knows the Microsoft stack architecture the best as it was written by Microsoft for the Microsoft technologies present in the data center. Combining DPM with the majority of the System Center family members and Azure technologies, you are able to provide an extremely high-end solution for your data center; it is all about design.
DPM only adapts to the current technology and solutions that Microsoft has developed for their workloads. DPM will never, for example, make any decisions on its own regarding SQL restore. The SQL team has clearly defined how a supported and optimal restore should occur and this is what DPM understands and adapts to; the same goes for all the Microsoft-branded workloads.
DPM 2012 R2 has the capability to back up production data sources to disk, tape, and Azure. For more detailed information regarding the DPM disk pool please read the Planning the DPM disk pool recipe. For more information regarding how DPM uses tapes, tape management, and Azure integration, please read Chapter 11, Azure Integration.
With DPM you are able to back up the most common workloads in a modern data center. The workloads that DPM will natively back up using the underlying VSS are:
System States and Active Directory
The following figure shows the DPM back up process:
For detailed information regarding the different versions DPM 2012 R2 can protect, please read the support matrix http://technet.microsoft.com/en-us/library/jj860400.aspx.
The primary DPM server is the first line of protection that you deploy for your production workloads. For systems that use transactional log systems (Exchange and SQL), you are able to replicate the production data from the protected data source to the primary DPM server every 15 minutes. This is also applicable to the file workloads.
Regarding the DPM disk pool, DPM has only one demand: the disk that should be used for the DPM disk pool must be presented in the operating system that has the DPM software installed as a local disk. DPM is able to operate with pass-through disks, Direct Attached Storage (DAS), Network Attached Storage (NAS), or Storage Area Network (SAN) storage. It is also possible to communicate directly with a storage using the iSCSI initiator within the operating system of the DPM server. Please note that this is not a good, optimal, or in any way decent approach to a DPM disk pool; never use iSCSI.
When DPM 2012 R2 starts protecting a workload, it creates an initial copy of the production data source that it has to protect. This is stored in the DPM disk pool in a dedicated volume that uses the same GUID ID as the protected data source. This volume is called the replica volume. The replica volume will hold the only full back up that the DPM ever will do of the protected data source. All changes in data, also known as data churn, that occur after the replica has been created, will be synchronized to DPM disk pool. When creating a recovery point, the data is stored in the recovery point volume that also uses the GUID ID of the protected data source for identification.
The most important component of the DPM server architecture is the DPM database also known as the DPMDB. The DPMDB holds all the configurations made on the DPM server, protection groups, agents, backup schedules, and so on. It is of great importance that the DPMDB is protected via a secondary DPM server when you build the DPM-DPM-DR scenario, copied to file, or archived to tape.
DPM depends on its DPM agents that will perform a VSS request for the protected data sources and have that data replicated from the production environment to the DPM server.
For further information regarding DPM, there are numerous places you can visit online. There are both Microsoft websites and MVP blogs that provide you with good content and decent information:
The DPM TechNet forum http://social.technet.microsoft.com/Forums/en-us/home?forum=dataprotectionmanager.
DPM TechNet web pages http://technet.microsoft.com/en-us/library/hh758173.aspx.
As the market develops, Microsoft understands the new challenges that companies and organizations face and therefore constantly develops its product stack to be able to meet the new requirements.
Microsoft has made some improvements and added some more architectural functions to DPM 2012 R2 release. The following list presents and explains the enhancements made to the product:
DPMDB cluster support: The database that stores all the configuration for DPM (DPMDB) can now be placed in a SQL cluster environment. This removes the standalone challenges that existed in the previous versions. With this enhancement comes reliability, consistency and most important, scalability.
Backing up of virtual Linux servers: DPM 2012 R2 can now protect virtual Linux servers running on Hyper-V using the technique "Backup using child partition" or "online snapshot" in Hyper-V. Note that only file-level protection is supported,not application-based protection.
Virtual deployment: It is now possible to deploy and manage DPM via System Center Virtual Machine Manager (SCVMM). You can install DPM on a virtual machine, and configure storage using
.vhdxstorage pool disks that are shared through the VMM library.
Optimization for online protection using Azure: Microsoft has optimized the express-full technology used when synchronizing the protected data from the DPM server on-premises to Azure via the Windows Azure Backup Agent.
SQL server media: For any new DPM 2012 R2 installation, the SQL server now needs to be pre-installed locally or remotely.
The new releases in the R2 version of the DPM software will make it possible for companies to meet some of the new challenges they are facing.
There are some parts of DPM that are very important to have knowledge about. They are:
VSS interaction with mini drivers and the DPM filter
Local security groups
The following are some catalogs under
%systemdrive%\Microsoft System Center 2012 R2\DPM\DPM that are important to the DPM server functionality:
End User Recovery
The End-User Recovery catalog contains the MS file for the active directory schema extension and configures the DPM server to enable the feature End-User Recovery.
Protection Agents contains two catalogs, which are AC and RA. The AC catalog is used by the DPM agent coordinator when pushing new DPM agents to the server hosting the data sources that it needs protection for. The RA catalog holds the binaries for the DPM agent and can also be used as a remote repository when manually installing the DPM agent.
DPM provides a lot of logs that you can read to gain an understanding of what has happened and why. All logs that the DPM server software produces are stored in the Temp catalog.
For DPM to be able to continually provide data source protection, even if protected virtual machines are making an outer-cluster migration that also includes the migration of storage, DPM needs to integrate with SCVMM. The catalog VMMHelperService contains the binaries and DLL files for creating the integration between the DPM server and VMM server.
The catalog Volumes actually contains four catalogs, and three of them are important to know about. The DiffArea catalog contains shortcuts to the recovery point volumes in the DPM disk pool. The Replica catalog contains shortcuts to the replica volumes in the DPM disk pool. The ShadowCopy catalog contains the catalog Database Backups, which will hold the backed up DPMDB when you trigger a DB backup using the DPMBackup executable.
Last but not least is the XDS catalog, which contains all the XML schema files for the DPM software.
The DPM 2012 R2 software depends on both local security groups that are referred to when the
DCOM object that controls the DPMRA service is used to initiate a VSS Request for the protected data source. If there is a problem with rights in the
DCOM object, the DPMRA will not be able to provide a snapshot since the chain of configuration is broken.
Since the local security groups are critical, it is important that their purpose is explained. The local security groups are:
Distributed COM Users
The members of the Distributed COM Users group are the computer accounts that the current DPM server has access to. You will also find specific service accounts here regarding your SQL server hosting the DPMDB as well as user accounts.
DPMDRTrusedMachines members are the other DPM servers that provide a DPM-DPM-DR configuration and via the membership of this group, have access to listing the protected data sources on the primary DPM server and take usage of the DPM writer.
The group DPMRADCOMTrustedMachines contains the primary and the secondary DPM server computer accounts.
DPMRADmTrustedMachines contains the computer accounts that have an associated DPM agent with the DPM server.
The group DPMSCOM contains the computer account for the SCOM management server that is used for monitoring and management of the DPM server via SCOM and the Central Console feature.
MSDPMTrustedMachines contains the computer accounts that have an associated DPM agent for the DPM server.
The last group is MSDPMTrusedUsers and this group is used by the centralized management features.
With the introduction of VSS, the purpose of making a full copy of the production data is obsolete. Using the VSS as a starting point, DPM provides an extremely fast and reliable change identification process.
DPM 2012 R2 relies on the VSS architecture and provides advanced change tracking functionality based on the mini-filter driver stack along with a DPM file filter.
The Copy-on-write technique is the one that provides the possibility of making online snapshots during production hours. This technique identifies the block-level changes and "freezes" the blocks making them ready to be copied. If there are other blocks that were changed during that process, the VSS identifies those. Once the initial identification has finished, the VSS starts managing the changed blocks.
Describing the whole picture, there is more to it than just using the standard VSS architecture. Not only is Microsoft DPM dependent on the VSS functionality but it also uses parts from the mini-filter driver stack and a dedicated file filter named DPM file filter to make the change tracking process of block-level changes fully supported, as well as optimal and fast for a DPM protection scenario.
For further and deeper information regarding how Volume Shadow Copy Service works, refer the MSDN website http://msdn.microsoft.com/en-us/library/windows/desktop/bb968832(v=vs.85).aspx.
When DPM creates a recovery point, there are a number of underlying technologies that make it possible. As mentioned earlier in this chapter, the DPM server can take advantage of a system that uses transactional logs and adapt to the change journal. Creating a recovery point in DPM is a part of what Microsoft calls Express-Full. The Express-Full operation is the function that is triggered when the backup job is initiated from the DPM sever.
When DPM backs up server workloads, the DPM server triggers the backup job. When the DPM server backs up Windows clients, the client triggers the backup job.
Block-level change tracking process
Synchronization of changed data
Creation of a recovery point
This recipe will enable you to plan your DPM server deployment and most importantly, make it possible for you to provide a starting point for your BaaS, RaaS, and DRaaS. The recipe will also cover network consumption and archiving possibilities.
There are a lot of questions that must be sorted out before you start planning your DPM server deployment. First is the classification process of the data sources that you would like to protect. A very common strategy that many companies that are still providing traditional IT to the company or organization are using is to back up everything they can backup once a day. However, this is not a good approach. There are, of course, several servers within your data center that need a high frequency of backups but not all the servers are equally important. It is essential that you adapt your business continuity plan before you have any implementation done. Start by identifying your services and then break down the services into components to clearly see how or why they are of importance to your business.
Use the different classes and sort the components that build up the service into these three categories:
As you may realize, the Gold class is more important than the Silver and the Bronze ones. Use these classes to differentiate your data and services so you get a clearer view.
From a more technical perspective, there are some considerations that need to be addressed:
The total amount of data that should be protected
Internet access to remote sites and cloud services
SQL Server installation
Virtual or physical deployment of the DPM server
To be able to understand the number of DPM servers that you need to deploy, you must first know:
How much data to protect
Network limitations between different remote sites
The need for building up disaster recovery scenarios
Every DPM server can manage 120 TB of storage for its disk pool. For example, 40 TB of those could be addressed to the file workload or 25 TB could be addressed to the SharePoint workload. In the scenario where you are managing multiple domains that you need to provide protection for, you can create a two-way transitive trust between the domains. In that case, the DPM server will operate in both domains without any limitations. This, however, comes with a security risk and if you have not created a two-way transitive trust between the untrusted domains you should think twice before you go for it. If you are able to set up a two-way transitive trust without causing any security risk, follow this blogpost to learn how to do it: http://robertanddpm.blogspot.se/2011/09/backing-up-cross-domains.html.
For the other scenario, to provide restore capabilities for untrusted domains, refer to Chapter 10, Workgroup Protection and CBA.
DPM can be deployed and provide a global restore capability strategy for your company or organization. If your bandwidth is greater than 512 kb then DPM will be able to provide its services. However, you should always include the restore scenarios when you are building up your backup strategies that reflect your business continuity plan. In many cases, you should deploy a DPM server on-site or build up your BaaS service using Windows Server Backup feature and have that integrated with Azure for archiving purposes, if you don't have the funding. Nevertheless, you are able to both throttle the DPM agents that are installed in the remote office protecting data sources and enable the on-the-wire compression for optimizing network consumption.
The SQL Server is no longer a part of the installation media for DPM 2012 R2, which is a good thing. Now you need to think before you act. A majority of the consultant companies needs to understand SQL more and also realize that if you have a poorly set up SQL, you will have a bad experience with the product hosting its database on that SQL server. Remember to set up your SQL server using service accounts, dedicate a RADI 1+0 for the DPM and monitor the SQL performance using System Center Operations Manager with a proactive monitoring approach.
With the release of the UR3, DPM 2012 R2 can be completely virtualized in all thinkable scenarios. You have the ability to use Synthetic Fiber Channel for accessing and using a physical tape library and also the ability to provide deduplication of the DPM disk pool using the SOFS architecture. When building a BaaS, RaaS, or DRaaS for a modern data center or a data center that still adopts the traditional IT approach, it is crucial that the DPM server is running on Hyper-V and is of instances that build up the service. Since a service in a modern data center is highly resilient and highly automated, you will benefit in many ways by building your BaaS, RaaS, and DRaaS using virtual DPM servers.
Physical tape library
Virtual Tape Library (VTL)
You should not just consider one of these three as an option. Focus instead on the requirements of the BaaS, RaaS, and DRaaS. Some protected data sources should be provided as archiving possibilities but maybe only once per month, go with Azure. Where protected data sources need an archiving solution but should also be able to restore quickly, go with VTL. In case the data sources need an off-site secure solution, go with physical tapes.
Having all the information presented in an organized manner, you can now start designing the structure of the BaaS service. A piece of advice here—take small steps towards your goal and never rush an implementation. You will probably bump into a challenge or two, so it is of key importance to work using a well-defined structure.
For more information regarding the setting up of SQL as a part of the prerequisites, please read Chapter 2, Installation and Upgrade, which also provides you a walk-through guide of the steps for DPM 2012 R2 installation.
The best way to start the planning phase for the DPM disk pool is to have the initial amount of data that you would like to protect and also a theoretical data churn. With these two components, you can easily determine the exact storage needed for your DPM disk pool.
System Center Data Protection Manger is able to use any type of disk that is presented as local attached storage in Disk Management.
From the UR3 release, you can now build a total virtual solution for the DPM server technology using Hyper-V. This also comes with the great possibility of building a virtual disk pool with VHDX files that could be stored on a Scale Out File Server (SOSF) and enabled for deduplication. With the approach of making a virtual solution you can save and optimize disk consumption very easily.
DPM 2012 R2 is able to use:
Direct Attached Storage
Fiber Channel Storage Area Network (FC SAN)
A very important fact is that iSCSI should not be considered your primary choice due to some challenges that often occur when leveraging that technology. The most common challenge is that the initiation of the iSCSI target fails and therefore the entire DPM disk pool fails. iSCSI will work in smaller deployments with DPM but if your ambition is to provide a more stable solution, you should consider SOFS. If your company does not provide an SOFS, you should use a DAS and provision VHDX files to the virtual DPM servers. As you can clearly see, the recommendation is to build as much virtually as possible.
Keeping track of the storage performance is also of key importance. Providing a decent proactive monitoring and performance approach is something that you will benefit from in the long run. Build up a proactive monitoring concept using System Center Operations Manager and see how the impact of the performance will change according to the number of backup jobs, workloads and synchronization frequency of the DPM servers.
Microsoft recommendations regarding how to determine the actual size of the DPM disk pool are from a minimal perspective. Multiply the total amount of protected data size by 1.5. For instance, if you want to protect 10 TB of data you need 15 TB of storage from a minimal perspective. The recommended approach is to multiply the total amount of the protected data size by three. So if you want to protect 10 TB of data, you need 30 TB of storage. These calculations will, of course, generate a lot of overhead in the DPM disk pool since there are several factors that are not presented in the equation.
Exchange data = Data source size x (1 + log change) / (alert threshold - .05)
SQL Server data = Data source size x (1 + log change) / (alert threshold - .05)
Windows SharePoint Services data = Total size of all databases/ (alert threshold - .05)
Virtual Server data = Data source size x 1.5
For Windows Server system state = (Data source size x 3) / 2
For file data = (Data source size x 3) / 2
Recovery Point volumes:
Exchange data = 4.0 x retention range in days x log change x data source size + 1600 MB
SQL Server data = (2.5 x retention range in days x log change x data source size) + 1600 MB
Windows SharePoint Services data = (1.5 x retention range in days x log change x total size of all databases) + 1600 MB
Virtual Server data = (Data source size x retention range in days x 0.02) + 1600 MB
For system state = (Data source size x retention range in days x 2) / 100 + 1600 MB
For files = (Data source size x retention range in days x 2) / 100 + 1600 MB
The log change is:
6% for Exchange
6% for SQL
10% for SharePoint
As you can understand, you will get a nearly exact estimation but the downside is that it will take a long time to calculate.
If you are protecting just Exchange, Hyper-V, and SharePoint you should use the storage calculators that are available for download at this site: http://www.microsoft.com/en-us/download/details.aspx?id=24375.
Providing decent hardware that hosts the DPM disk pool is very important. You don't need premium disk solutions for the DPM disk pool but can use decent hardware that can easily scale out.
Adding the DPM disk pool could be done via the DPM console or via PowerShell. For more information regarding the subject, please read the Adding disks to the DPM disk pool recipe in Chapter 3, Post-installation Tasks.
It's important to know the limitations of the Windows Server operating system. DPM 2012 R2 has the following limitations that you should keep in mind while designing BaaS, RaaS, and DRaaS:
600 volumes in the DPM disk pool
32 spanned disks in Disk Management
The disk that should be used for the DPM disk pool must be a Dynamic disk in Disk Management
The DPM disk pool can store 9000 snapshots
Opening just the right amount of firewall ports with the right direction of communication will provide you a more high-end security approach. DPM uses Microsoft standard communication ports, but for some features, there are a few other TCP ports that need to be opened.
135 / TCP
DPM specific ports
5718 / TCP
5719 / TCP
53 / UDP
88 / UDP
88 / TCP
389 / UDP
389 / TCP
137 / UDP
138 / UDP
139 / UDP
445 / TCP
6075 / TCP
1433 / TCP
1434 / UDP
80 / TCP
443 / TCP
50000 – 65000 / TCP
4022 / TCP
5723 / TCP
Having the Windows firewall enabled would be considered the most natural thing. However, many companies rely on a physical firewall as their first line of defense meaning that their Windows firewalls are disabled.
An easy approach is to create a Group Policy Object (GPO) that holds the configuration for the Windows firewalls. Use the Advanced mode for firewall configurations so you can easily provide the necessary configurations.
One important thing regarding the direction of communication is to understand who is initiating the communication. When DPM is protecting server workloads, the DPM server will call for the DPM agent to start its VSS request, but when DPM is protecting clients, the DPM server will wait for the DPM agent present on the client to call in.
You could also limit the actual port range for the high-end ports to a specific port range. For instructions on how to do this, you can refer to this article: http://blogs.technet.com/b/dpm/archive/2011/06/28/how-to-limit-dynamic-rpc-ports-used-by-dpm-and-protected-servers.aspx.