Category Archives: UC&C

One of the things I like most about Microsoft’s approach to software engineering is their focus on making data-driven decisions. For example, if you remember the advent of Windows 2003, you may remember that MS spent a ton of time analyzing crash causes for Windows 2000. They found that a large percentage (I want to say ~ 60%, but I might be making that up) were caused by faulty drivers– so they introduced the concept of signed drivers, and the ability for admins to disallow unsigned drivers.

For a more recent example, take a look at the Office 2007 ribbon. Love it or hate it, the ribbon came about because we told MS that’s what we wanted, through the Customer Experience Improvement Program (CEIP). CEIP allows you to choose to send usage data to Microsoft so they can get anonymized data about how their programs are being used.

The Exchange component teams want to start making more use of CEIP data opportunity to improve customer experience and drive E15 planning by gathering data in several high impact areas. Microsoft’s David Espinoza listed some of the data items CEIP could potentially gather:

• How many users have turned on SMS notification?
• How many NDRs are generated, for what reason?
• How much latency exists between the time a user sends an SMS message and the time the phone picks it up from the Outbox?
• What are the most frequently observed cmdlet errors?
• What Exchange Server roles are the most often installed in virtualized (VMWare, Hyper-V) environments?

As you can see, none of these are particularly sensitive (or even that interesting) for a particular site. However, in the aggregate, data items like this give Microsoft a great deal of insight into what features customers are using, how well they’re working, and where they could be improved.

Microsoft is asking everyone running Exchange 2007 or Exchange 2010 to turn on CEIP, so I thought I’d help spread the word. The actual process is simple; you can enable CEIP on individual servers or for the entire organization with a simple Exchange Management Shell cmdlet. Give it a try!

Darn it, Dr. J beat me to the punch:

Remote wipe in Exchange ActiveSync is only useful when a user loses his or her device, and even then, it is lacking since you cannot reach out to the device and wipe it. Remote wipe in Exchange ActiveSync is utterly useless when people are terminated from their employer.

In the case where you fire an employee and want to remove your organization’s confidential data from it, there’s a big ol’ hole that Jesper describes quite well. There are various mitigations that might seem to apply, but most of them revolve around preventing someone from connecting in the first place, or in blocking their ability to connect after you fire them or whatever. For example, you could use client certificates so that only devices with certificates could connect, but then only as long as the client cert remains valid. That doesn’t solve the wipe problem, though.

A related problem: the current device wipe implementation on most phones resets everything and completely erases the phone back to its factory state. Users lose all their apps, personal data, and so on: not a great experience for them (though one school of thought says that you just fired them, so who cares?)

It seems like it would be reasonable to do two things. First, allow sending the EAS wipe message to a device even if it fails to authenticate. If the device has an existing sync relationship, and it tries (but fails) to authenticate, just send the pending wipe message to it anyway. Second, give admins the choice of whether the wipe message requires a complete wipe or only deleting data that came from the organization originating the wipe message in the first place.

I literally had never heard of this feature until I found this excellent post by Elie Bou Issa. Turns out that the Calendar Repair Assistant, or CRA, is documented, but I hadn’t run across it.

In brief, the CRA’s job is to ensure that calendar data items stay consistent across the organizer and attendees’ mailboxes. This is a hard job given how many different clients may be in use, and how many of them (I’m looking at you, Apple) have problems handling some kinds of Exchange calendar events. Elie’s article explains things quite well, so I have nothing more to add than a delighted fist-pump that this feature even exists.

I’m doing a TechNet webcast on 16 March at 11 am. The topic: Exchange 2010 unified messaging, and what’s new and updated relative to Exchange 2007 UM.
Register here, and I’ll see you there!

Sometimes I get voice mail messages for which Exchange 2010 refuses to provide a voice mail preview. It says “oh, bother; the message was too long.” Wouldn’t it be great to set the threshold for “too long” in Voice Mail Preview?
Too bad; you can’t.
Well, OK. Let me elaborate: you can change the maximum length used when you use a transcription partner. (Wait, what? You didn’t know that you can outsource transcription? Yep. I’ll write more about that later.)
However, there’s no supported way to change the length threshold for messages transcribed on your own system. There’s a good reason for this, though.
All speech recognition systems rely on statistical models to help them “understand” what the speaker’s saying. These models help the speech engine predict what sounds are likely to come next given what’s being said right now, and improved models are a big part of why speaker-independent continuous speech recognition has finally reached the point of actually being useful.
There’s a problem with this approach, though: messages that don’t match the model are really, really hard to recognize. When a speaker says “My phone number is four”, if the next phrase is “syzygy exploding potato” it’s going to be hard for the model to keep up since it’s expecting a few more numbers, not random crapola. The more the message diverges from the model’s expectation, the worse accuracy gets.
This ties into long messages how? Glad you asked. The longer a message is, the more likely it is to be a specialized message– in other words, one that doesn’t match the model well. Consider the difference between a short message (say, “Hi, honey, it’s me. Don’t forget to pick up some milk on the way home.”) and a speaker reading this blog post to your Exchange UM server!
To avoid this situation, Exchange won’t transcribe any message longer than about 75 seconds. That’s not a big deal, given that something like 99% of all voice mail messages are shorter than 75 seconds, but it’s still a bit of a hassle if you need transcription for longer messages. Of course, then you could always use a transcription service, about which more later.

One of the coolest things about Exchange 2010 is that it’s designed from the ground up to deal with multiple Exchange organizations. There are all sorts of features that take this into account, like the ability to designate equivalent dial plans, set up federation trusts, and so on.
One feature you may have missed is that the Exchange Management Console can be used to manage any Exchange organization, not just the one associated with the domain you’re logged into. This comes in especially handy if you’re using Exchange Online, or if you need to manage more than one Exchange organization from the same machine.
Turns out, though, that you can only install the Exchange 2010 management tools if your machine is joined to a domain. It doesn’t matter which domain; any one will do.
This seems odd at first glance. After all, lots of other management applications allow single-seat management across Windows domains. In fact, the Exchange 2010 control panel (ECP) does this. There actually is a good reason for the restriction, though. The Exchange 2010 EMC uses Kerberos to authenticate and encrypt the data channel used for remote PowerShell. This is simpler to deploy and manage than dealing with yet another set of SSL certificates for HTTPS… but the use of Kerb requires that the management workstation be joined to a domain. There you have it.

I’ve been meaning to blog about this for a while, but lately I’ve been doing a lot more work with the Exchange 2010 management shell, and this has really been getting on my nerves.
In Exchange 2007, you could use tab completion to figure out which cmdlets existed for a given task. Need to do something with public folders, but not sure what cmdlets are there? Just type “get-pu” and hit Tab, and voila! you start seeing public folder-related commands.
Exchange 2010 breaks this!
The reason is simple: tab completion doesn’t work in remote PowerShell. Every time you launch EMS, you’re making a remote PS connection, even on the same machine. Therefore: no tab completion for you.
Worse, the get-help cmdlet won’t accept wildcards any longer. The same issue is to blame.
There’s a partial workaround for this issue, which is to use get-command (or just “gcm”) instead of get-help. gcm will give you the list of cmdlets, and then you can use get-help to get help on an individual cmdlet.
I don’t think this is a great situation. In fact, it stinks. I learned PowerShell in large measure by trying to tab-complete various things to figure out what existed, and as I’m now trying to train a few other folks on basic PowerShell, the lack of this feature is a big pain in the butt.
However, complaining to the Exchange team won’t help much because this is a PowerShell issue. So, if you find this bug annoying, you can let the PowerShell team know on their blog (or here; I’ll forward a link to them). Maybe this will be fixed in a future release… hopefully not a far-future one.

Last month I wrote a bit about MailTips and Group Metrics processing that happens on Exchange 2010 mailbox servers. Now for some deeper follow-up.

First, E.J. Dyksen has a good post on general MailTips troubleshooting on the Exchange team blog. Go read it. (Don’t worry; I’ll wait for you to come back.)

Back already? Great. Now let’s get to it.

Group Metrics creation is on by default; it’s controlled by the organizational GroupMetricsGenerationEnabled setting. Every mailbox server that generates an OAB will also generate Group Metric data for the entire organization. Other mailbox servers will not generate any GM data, unless and until you enable them to do so.

Clients ask the CAS for Group Metric data. A given CAS server will assemble a list of Group Metrics servers, then use AD sites and site link definitions to find the "best" server for getting a copy of the Group Metrics data. The list contains servers that generate OABs for that CAS, plus mailbox servers that are explicitly enabled for Group Metrics generation. (Note that this means that CAS servers that don’t host any OABs also won’t host any Group Metrics data.)

When does all this happen? By default, Group Metric generation happens at midnight, plus or minus three hours. You can change that time using the GroupMetricsGenerationTime parameter, in which case the +/- 3-hour offset doesn’t apply. Also, after you change the time, you’ll notice that the next Group Metrics generation follows the old schedule, then the new schedule kicks in.

There’s no way to force regeneration from the UI, although you can stop and restart the Microsoft Exchange Service Host service on any server to force an update. However, updating the Group Metrics data won’t force the CAS servers to pull the new update; for that, you have to restart the Microsoft Exchange File Distribution service.

I am an early adopter (but then if you’ve been reading this blog for more than 10 minutes, you knew that already). Recently, two fairly new pieces of technology have been working together in a, shall we say, unexpected way.

First is Outlook 2010, which I’ve been running for some time. Overall I like it a lot; it performs well, it supports multiple Exchange accounts, and it has lots of grace notes (like the configurable "Quick Steps" feature) that make using it both easier and more pleasant than earlier versions.

Second is the newest release (3.1) of the Facebook client for the iPhone. It has the option to automatically sync Facebook data (including profile links and pictures) for contacts it finds in your iPhone address book.

Put these two together, and what do you get? Pictures of people you don’t really know showing up in Outlook, like this:

Yes, that’s right: any time I exchange mail with someone whose e-mail address is registered with Facebook, I get their picture! In this case, a college student bought a book I had listed on Amazon, and she was writing to ask whether I’d shipped it or not. Imagine my surprise to see a picture of her and her two BFFs (at least that’s who I assume the other two girls are.)

What’s making that happen? Outlook 2010 has a feature called "Suggested Contacts" that automatically adds the e-mail addresses of people you correspond with to a new "Suggested Contacts" folder. This replaces the old .nk2 file that earlier versions used for nickname autocompletion. Unfortunately, Suggested Contacts appears to most applications (well, the ones that aren’t Outlook 2010) as a regular Contacts folder. On the iPhone (and in Mail.app) that means that people you’ve exchanged e-mail with show up in your contact list until you manually purge them.

The Facebook app on the iPhone is trying to be helpful, so it looks for people in your address book—which now includes the contents of Suggested Contacts—and downloads their pictures. Ta da! Instant confusion.

The contact-picture feature is one of my favorite Outlook 2010 enhancements, so I’m not going to turn it off. Likewise, having up-to-date pictures of my actual Facebook friends is a neat feature, so i"m leaving it on as well. For now, that means that I’m stuck occasionally seeing pictures of people I don’t know—part of the price for being an early adopter, I guess.