Well, it has been about a month since my last post. Apologies... My time was spent preparing for a trip to "campus" (Redmond) to attend an "Exchange IT Fellowship" which was an in-depth course on how Microsoft does "IT" from an Exchange perspective. We have our own internal IT department that manages our own Exchange deployment. It was a very cool experience. The deployment is like SBS on steroids!! ISA2004 doing web publishing of the Exchange virtual directories in IIS. Forms-based authentication in IIS (pretty OWA logon screen), SSL bridging back to IIS, etc. The coolest thing is that we use our own native anti-spam technologies on our "gateway" servers - no 3rd party products! Recipient filtering, sender filtering, tarpitting, RBL lookups, IMF, and SenderID. About 85% of the inbound mail on any given day is rejected/filtered as spam. I actually saw Exchange System Manager for our corporate servers. I got one of the instructors to click on queues and there were about 1200 in a "retry" state. The vast majorities were "typo's", meaning petergal@Microsft.com (missing the "o" in Microsoft) or email@example.com (should be sbcglobal.net). We run Exchange in a clustered environment and we have specific SLA's (Service Level Agreements) that govern the way we manage/update our servers. When one of the "active" nodes fails over to a "passive" node...we do not automatically "fail it back" to the previous active node whenever it comes back online. The reason why is that it takes the "node" offline for about 4 minutes, which blows the SLA of 99.99% uptime for the month out the window. Another very cool thing was seeing the actual hardware. Bit of an overkill for an SBS deployment!! LUN's with fiber channel adapters. The fiber channel adapters on the LUN enclosures are actually running Windows XP Embedded on the controller itself <very cool>. It is amazing the amount of research that goes into the hardware required for a specific Exchange deployment. How many users? How much data? What is the SLA for downtime? What is acceptable latency? How many users are going to use OWA/PPC/RPC over HTTP? We have several tools to help determine the appropriate hardware for an Exchange deployment including the JetStressUS and LoadSIM. IOP's per second was an oft used term (Input/Output per second). A "day in the life" of an Exchange server consists of 8 hours user access, 8 hours for online maintenance and 8 hours for backup. We actually backup to file on another drive (NTBackup) and then backup that file to tape. This helps reduce the time to backup the databases. We also use MOM2005 to manage the Exchange environment. The datacenter has 5 large overheads that display MOM statistics as well as various other "data".
Also, something about Seattle made me want to drink large amounts of coffee!