TECH

I changed the way I was making notes today. Yesterday I was using the old paper and pen method as I was only getting 90 mins battery life out of my laptop. The days are quite long 9-7 and with some food in the evenings was struggling to write things up so have started to use my mobile to take notes. Thankfully it does have a qwerty keyboard all be it a little small but the battery easiiy lasted all day on it.

Exchange 2007 failover and high availability

Current model is based on hardware failure available but no data failover.

Single Server Mode is called local continuous replication – this is based on log shipping.
It’s designed to copy data locally to a second disk.
Failover is a manual process.
There is some resource hit on the server due to the extra disk activity.
Allows for larger databases due to faster restore

Clusters failover node.
Only the mailbox server role only role cluster aware all other roles have to go on other servers. Up to 8 nodes always with at least 1 passive node is supported in a cluster but if you want data replication only two nodes are currently supported.

Exchange 2007 supports up to 50 databases

Cluster Continuous Replication
2 nodes only
No shared storage
Local copies of both the data and the logs
Local Quorum on both servers.
File share created on another server that is as quorum witness. Recommend it goes on hub server. This is used in case the two nodes cant comminute with each other.
No resource hit like on local replication

There is a blogcast demo of this here http://msexchangeteam.com/archive/2006/08/09/428642.aspx

An initial copy of database is taken
As changes to logs are made they are copied to the other node and then replayed
Only replicated closed logfiles not bits of logfiles are replicated.
New service to handle this - microsoft exchange replication service

Each storage group has single set of logs
Each log set has prefix E00 E01 etc
E00 is active and then once used up renamed to next inc number and new E00.log created

Dirty failover - if active node fails then open log isn't avail on passive node as they only get copied once closed. Passive node carries on with some data loss. Starts logs from the number that the original passive node had. There are settings to control this but it’s a toss up between getting the server up and running quickly or recovering all data. There are other ways to get back some of that data without waiting for the original server to be fixed.

If active is repaired it comes up as passive. If logs are still there then there will be a mismatch due to uncopied logs.

Repairs to database can also cause a mismatch to database such as a defrag. This is called divergence. Replication service can detect this.

Log file sizes are now reduced to 1MB from 5MB to reduce information that could be lost.

Vss backups can now be done from passive node. This gives you snapshots in time of the database. Backup notifies the active node to tell it that a backup in process. Once backup complete passive note tells active node and changes made to the active node.

There is a transport dumpster on the hub server. This is only enabled when using CCR. If the recipient of a mail is on a CCR server then its put in a CCR dumpster on the hub. Recommend that this is set to 3 x max message size. The dumpster is used to replay back messages that may have been lost in a failover situation due to the lost log files.

Circular logging. Not supported

---------------------------------------------------------------------------------------------------

Softgrid application delivery

There are 3 components to Softgrid
Sequencer
Server
Client

Sequencer - standalone machine. Recoding process that watches changes made when you install the app. App can also be configured to set the default settings in the application such as file paths etc. Have to also launch the app so it knows which files are necessary to run the app and what are the additional files
Creates 4 files. The files go onto a delivery server that can service 000,s of users per dual processor system

Every time an application is run its logged so you can track who is running what.

Softgrid is integrated into sms so people with sms can still leverage the benefits that sms gives with asset management and tracking. Sms can meter softgrid as well so it can watch both local and softgrid apps in one place

Softgrid is 4 years old so have a number of large existing customers all ready

Softgrid very useful for citrix environment as you don't need to build silos due to application conflict. Bad behaved apps may still need to be silo’d but conflict silos aren't needed

Desktop optimization pack - available in Jan. Includes
Softgrid
Diagnostics
Asset inventory
Advanced group policy

Works out roughly 10$ per user per year. Hell of a lot cheaper than when I looked at this a year ago.

Softgrid for terminal servers is different product

A good reason to install office locally is that COM is only possible between a virtual app and the desktop not between two virtual apps as most apps need to talk to office and office shares between apps install local is a good idea