I’ve tuned into a webcast aimed at the System Center Influencers and I’m going to try blog from it live. Microsoft’s line is that System Center is the way to manage SharePoint because Microsoft understands the requirements.
SharePoint often started as some ad-hoc solution but grew from there to be mission critical and containing urgent business data. Administration is complex: users, file server admins, web admins, database admins and web developers.
System Center Improves Availability:
- DPM backs it up the way it should be.
- Operations Manager monitors health and performance.
- Virtualisation (VMM managed) can allow for rapid deployment with minimal footprint.
- Configuration automates management
- Service Desk will add more benefits
This is the norm for System Center. Centralised management with delegation is how System Center works. For example, a Sharepoint administrator could deploy a front end server in minutes using the VMM 2008 R2 self service portal. A quota will control sprawl but the network administrators don’t need to be as involved.
OpsMgr Management Pack
- There is a new monitoring architecture. There are physical and logical components where the physical entity rolls up to a logical entity.
- Monitoring is integrated into SharePoint so the SharePoint admins can see the health in SharePoint
- There will be a unified management pack instead of the current 2007 split management packs. The discovery process will identify the roles installed on an agent machine and only utilise the required components.
We’re shown an OpsMgr diagram that shows the architecture of a SharePoint deployment. If you haven’t seen these, they are hierarchical diagrams that give you a visualisation of some system, e.g. HP Blade farm, Hyper-V cluster, SharePoint farm.
The 2010 management pack allows you to monitor a particular web application in SharePoint 2010. The management pack is more aware of what components are deployed where and the interdependencies – sorry I’m not a SharePoint guru so I’m missing some of the terminology here.
Rules administration has been simplified. There is a view in the Monitoring pane to view the health of all rules for the SharePoint 2010 management pack. I like this. I’ve not seen it in any other management pack. The SQL guys should have coffee with the SharePoint folks :)
Three are 300% more discoveries and 1293% more classes and 300% more monitors than in 2007. That is a huge increase in automated knowledge being built into OpsMgr to look after SharePoint 2010. There are 45% fewer rules. This is a good thing because there is duplicated effort being reduced for IIS and SQL management pack to reduce noise. Microsoft assumes you’ll install those other management packs. approximately 150 TechNet articles are linked in the pack to guide you to fixing certain detected issues.
Data Protection Manager 2010
DPM 2010 is due out around April 2010. It important to Hyper-V admins because it adds support for CSV. DPM allows you to backup to disk and then optionally stream to tape. You can also replicate one DPM server to another for
SharePoint 2003 and WSS 2.0 are backed up basically as SQL. You need the native SP tool to complete the backup..
SharePoint 2007 and WSS 3.0 is backed up using a SharePoint VSS writer. Every server (web/content/config/index) gets an agent. DPM reaches out to “the farm” and can back up everything required.
DPM is designed to know what to back up. 3rd party solutions are generic and don’t have that. For example, a new server in the farm will be detected. The DPM administrator needs to authorise this addition.
DPM 2010 does something similar with SharePoint 2010. However, it is completely automated, allowing your delegated VMM administrators or Configuration Manager administrators (SharePoint administrators) to deploy VM’s or physical machines.
One of the cool things about DPM is that it doesn’t have specialised agents. It’s using VSS writers. That means there is 1 agent for all types of protected servers.
We get a demo now and we see the DPM administrator can just select “the farm” and back that up. There’s no selecting of components or roles. The speaker only sets up his destination and retention policies.
DPM 2007 is noisy, e.g. data consistency checks. I’ve seen this when I did some lab work. The job wizard allows you to either to perform a heal/check if a problem is found, on a scheduled basis or not at all. This is a self healing feature.
Recoveries can be done at the farm level, an individual content (SQL) database. SharePoint 2007 can restore a site collection, a site or a document. This requires a recovery farm, i.e. a server, consuming resources and increasing costs. SharePoint 2010 with DPM 2010 does not require a recovery farm. You can directly recover an item into the production farm. Trust me, that’s huge.
The release candidate for DPM 2010 comes out next week.
- Web role, Render Content: Virtualisation ideal
- Query role, Process Search, Queries: Virtualisation Ideal
- Application Role, Excel Forms Services: Virtualisation ideal
- Index role, Crawl Index: Consider virtualisation– small amount of crawling, and drive space used to store the index (VHD = maximum 2TB, although you can go to pass through disks for more).
- Database role: Consider virtualisation – OK for smaller farms.
My advice on top of this: Monitor everything using VMM and Operations Manager. You soon see if something is a candidate for virtualisation or if a VM needs to be migrated to physical.
If you run everything on a Hyper-V 2008 R2 cluster then enable PRO in VMM. Any performance issues will allow an automatic Live Migration (if you allow it) to avoid performance bottlenecks.
If you are going physical for the production environment then consider virtual for the DR site if reduced capacity is OK. For example, your production site is backed up with DPM. You keep a Hyper-V farm in the DR site. Your DPM server replicates to a DR site DPM server. During a DR you can do a restoration. Will it work? Who knows :) It’s something you can test pretty cheaply with Hyper-V Server 2008 R2. Money is tight everywhere and this might be an option.