Latest posts by Paul Schnackenburg (see all)
- Project Honolulu - A new way to manage Windows Server - Wed, Nov 22 2017
- Use Azure Managed Service Identity (MSI) to store passwords in your code securely - Thu, Nov 9 2017
- Azure Data Lake overview - Fri, Sep 22 2017
The lucky attendees at his sessions were treated to great presentations and were also the first people in the world to hear about the new stuff that’s coming in 2013.
Scott Schnoll - Paul Schnackenbug
SS Ha! You are talking about the internet database availability group (iDAG)?
PS Yes, I am.
SSThat went over pretty well, actually, yes. There were some folks who found it very convincing and actually thought it was a thing we were really releasing.
PS Yes, well, as you might remember, I teach part time at an after high school but before university college, and I did go back to class the next week and tell them about this cool new feature. So I really fell for it!
SS Well, then I would have to retract that earlier statement and say it was largely successful. I fooled everybody. Awesome! Mission accomplished. Yes, we had, you may have noticed several interesting stories coming out from Exchange on that day, so it's something we try to do every year, just have a little fun.
PS It was definitely an insider's joke.
SS I intentionally wrote it to be as convincing as possible. But there were a few hints, you probably noticed there were a couple of giveaways...
PS Yes, there were.
SS The Will I Am reference and things like that.
PS Once I read it the second time I realised I'd been had, but it was a bit embarrassing really, but there you go. So, Exchange 2013 to me feels almost hot on the heels of Exchange 2010; Exchange 2010 is a successful product, I think, for Microsoft. What's 2013 bringing to the table?
SS So many different changes. We do have some of the same terminology that you're used to, but even things like the Client Access Server Role that we had in 2007 and in 2010, but its function is very different from what we had before. We're actually down to just two Server Roles now, the Client Access Server Role and the Mailbox Server Role. The functionality that you used to have in previous CAS and in the Hub Transport and the UM Role, that's now been rolled into the Mailbox role, so now you have a much easier deployment.
We've also done a lot of work around the areas of the store and storage, so for instance the old monolithic store in Exchange that we used to have, that actually we've had since the first version of Exchange, that's gone and replaced with what we call the Managed Store, which is a new sort of a service worker process all written in managed code that completely replaces the old C++ store that we had before.
We've done a lot of work to reduce our IOPS even further; you know from our talk last year that with 2010 we reduced our IOPS 70% over 2007. We've been able to reduce IOPS again in 2013 anywhere from 48% to 76% depending on the client that you're using, whether it's cached or online mode. And we've also introduced new features in the storage area to take advantage of hardware that's not yet available, so the larger, for instance, eight terabyte hardrives. You may know that we have our maximum recommended database size is around two terabytes, so as disk drives get larger and larger, you've got a choice of either wasting a whole bunch of space, if you implement larger drives, or growing the database beyond our recommended size limit.
So, to eliminate those choices and give you a clear path, one of the features we have is called Multiple Databases per volume, so you take an eight terabyte drive, for example, and you carve that up, you make it into a single partition, and you carve up a directory structure that allows you to store multiple databases on that same disk, so four 2 TB e data bases, if you will, instead of one, 2 TB database. So now that means not only, because of the reduction in IOPS and the ability to leverage this new hardware, not only are you able to leverage the new hardware, but you can actually add to the capacity of this system as well, you can have more users supported on the same server in this paradigm.
We also have a companion feature to that called the Auto Reseed and we've done a lot of work in the storage area specifically to target JBOD environments. All the storage features work in all storage environments, but it’s in a JBOD environment in particular, if a disk fails, well, that's a database failure and that will trigger a database failover and if you’re in a paradigm where you're hosting multiple database copies from the same drive, well, now it could be two, three or four databases that have failed and need to be reseeded. So, the automatic reseed feature is one where, when the system recognises that a drive has failed, and failed from an Exchange perspective means that it goes away from the operating system's perspective, so the OS can't see it, or ESE, the Exchange Database Engine can't write to the drive.
If we see those conditions occur, we can take a spare drive that the administrator's already allocated in the box, initialise it, map it, create the necessary directory structure and then automatically reseed all the failed databases. So instead of the operator normally having to do all that work, replace the failed drive, put the new one in there, format everything, do the manual reseed operation, now all of that can be completely automatic. And that's one of many, many examples where we're taking what is normally a lengthy operator driven activity and giving the system the ability to do it automatically, so that now the admin or the operator focuses just on fixing what the core problem is versus restoring data or service redundancy.
PS And the impetus behind those sorts of changes really come from you guys running Exchange at a huge scale in your own datacentre.
SS Absolutely. And what I find interesting is, there's a tendency for some folks to say, as they see us focus on the cloud so much, that the on-premisess product isn't getting any love because of that and in reality, the opposite is actually true, because all the learnings, all the teeth cutting, all the breaking of new barriers, all that stuff is done in the cloud, in our data centre, by us, by the Exchange team. And so the customers don't have to go through those headaches and heartaches in order to learn these lessons. They learn from our lessons and they get all that benefit raining down upon them in the on-premisess product.
So that's one of the examples, you know, when a disk drive fails in the service which is so large and has so many disks, it happens every day, it's an expensive proposition to have somebody there who has to manually deal with this. The larger the service gets, the more you have to reduce costs and being able to have the system self-heal in that way is a perfect example of how we can do that. It's much less expensive to have the system fix itself versus having a human come in and having to do that. And that's just one of many, many places where we can do that now.
An even bigger change is, as you may have heard at the end of the Site Resilience session yesterday, was all the changes we made around the architecture itself, decoupling, CAS and Mailbox from both a namespace and recovery standpoint, and then the worldwide single namespace capabilities that we give you, that gives you much bigger changes. It's great to have the storage changes, but thinking about the bigger scale type of disaster where you lose an entire site or data centre, something like that could take an on-premises organisation in Exchange 2010 hours or days to restore service, and in part because it's a manual process in 2010; we require humans to make the decision because of all the connected moving parts.
In 2013 we've separated that all out, everything has got failure isolation, everything has its own island, the namespace is decoupled, CAS and Mailbox don't require each other in a site anymore. We’re also much better on the network, we moved the session affinity into the product and away from the load balancer; we've done away with extra namespaces we don't need any more by changing all of the Outlook clients to use Outlook Anywhere, so that’s Outlook HTTPS over TCP, and as a result of all these changes, you can now have independent data centre failover of the namespace or of the DAG or of both.
So now, if, say for example, in my primary data centre, I lose my load balancer, so in the 2010 paradigm, that's an unfortunate scenario, unless you happen to have a second one right there that can take over, which a lot of customers don't. Now you have to do a manual data centre switchover and that’s a whole lot of stuff that has to get changed, including DNS, which means data’s replication, data’s latency, which in some parts of the world, dealing with that is not very good when you consider the multitude of clients. So with all the work that we've done to separate this, now I could, in my primary data centre, lose half of my DAG and the other half can go and failover automatically and everything can continue working seamlessly.
Or if I lose the load balancer and I have another one, we could automatically failover the namespace and be serving clients from a different data centre and have those requests simply proxied over HTTP back to the Mailbox server which is unaffected. So now, instead of restoring service, now you're just replacing the failed piece of equipment and any admin will tell you, who's been in this situation, that the stress that the administrator's under when they have to restore service is so much higher than the level of stress they'd be under just replacing a piece of failed hardware. It's like night and day. And then when you consider the failover aspect is now largely happening at the client through DNS Round Robin and redundancy of the namespace, now the recovery time objective just went for every organisation from, even if they could get it down to 45 minutes in 2010, now it's gone down to about 20 seconds in 2013. So it's a huge, huge win in 2013; stuff people asked us for in 2010 that simply wasn't possible, that by re-architecting everything we made possible in 2013.
PS Which again, sounds very much like a feature that comes from your Cloud and Office 365 experience.
PS You've gone to Outlook Anywhere for all clients; which used to be an optional protocol, now it's the way everybody connects; what's the thinking behind that?
SS The thinking behind that has to do with some of the architectural changes that I just talked about. If you have a database availability group (DAG) and you extend that to two data centres, just to sustain that environment you need to manage eight different namespaces, two of which are for the Outlook RPC endpoint for the internal domain joint clients who would use MAPI over TCP, they need that RPC client access array object that we introduced in Exchange 2010.
We moved away from that in part to do away with those namespaces because they have to go on the certificates which mean money for the customer, but also in part to standardise by using HTTP as the access protocol for almost all clients. Yes, we still have POP3 and IMAP 4 and yes there's still obviously SMTP and UM clients, but all the other access methods, Outlook Web App, Outlook Excel, Exchange ActiveSync, Exchange Web Services, and even Remote PowerShell, all of those are HTTP and that gives us a very reliable and stable protocol to use. It also eliminates the problem of DNS latency, because now we don't have to change DNS anytime we want to move service around. It also eliminates the RPC client access namespaces and it unifies everything into a protocol that was designed to go over WAN versus RPC which we use over WAN today but it wasn't really designed to go over WAN because it really prefers a high bandwidth / low latency requirement.
HTTP on the other hand is just the opposite, it was designed for low bandwidth and high latency and so that gives us the ability to have, you know, for instance, a customer or a client out in APAC accessing the client access server that might also be in APAC but then that CAS server is proxying the HTTP request over the internet to a Mailbox server in North America and that's doable now, where it wasn't in 2010. In 2010 CAS talked to RPC in the Mailbox server so they had to be really close. We could have CAS and the client be far away, because that was HTTP, but we were still talking RPC there. Now that we're talking HTTP from the client to the CAS and proxying that from the CAS to the Mailbox, now we're much better as a network player.
PS So all of this really is server stuff which will make Exchange administrators happy. What does the end user see? What's the change for the people who are using Exchange 2013 as their email platform?
SS Sure. So, of course, it's going to depend on the client that we're using, so Outlook users, you're going to see a big change with the new Outlook client, obviously it's designed for the up and coming workforce and the next generation of devices; it's also designed with the social integration in mind as well. This means information workers have a big win, because they have tools that are ready to go with the latest operating system, and with the latest hardware, but even though it's brand new, it's still going to be natural and intuitive to them because they're going to recognise all the attributes of Outlook that they enjoyed.
It might have a visually stunning new UI that, you know, they'll find gorgeous, but when it comes down to it, people don't just stare and love the UI, they want to get work done. And so they'll be able to do that kind of stuff as well, they'll be able to understand what all the Outlook functions are, whether they're on a Tablet using the touch mode or whether they're using a kiosk PC or their desktop or their phone, for that matter, and then they're also going to benefit from a lot of the integration that we have across Office that we didn't have before.
The integration that we have with SharePoint, for example, with this new feature called Site Mailboxes. That is what it sounds like, it's an Exchange Mailbox paired up with a SharePoint site, and you get the benefit of all the collaboration between the two without having the duplicate content stored multiple places, without losing version control, because the way we’ve implemented is we’ve split out the functionality, so the Mail items we store in Exchange. The documents and things that people email around or post to SharePoint, they get stored in SharePoint and the experience is great because I could send out an attachment to folks on a Site Mailbox and the attachment will actually get removed from the message and put into SharePoint automatically and then in its place will be a link to where it lives on SharePoint.
So you're not constantly sending out emails with documents, here's this version and you have to keep track of it now if you send it out to other people, they have to keep track of it; instead we all get to work from the one same source but we don't have to change the way we work in order to do that. I can still send you an attachment in email and have it just do that automatically behind the scenes.
And then of course, the stuff that we've done with Lync, where we already did have some integration with Lync in the previous versions, we're taking that farther by introducing not just the presence of information around there but also having things like a full fidelity Office experience inside things like Lync meetings for example. I don't know if you got a chance to see some of Andrew Ehrensing sessions on Lync? The ones that will stand out from the information worker's standpoint would be the meeting improvements, in 2010, for example, we always show you the active speaker, whereas in 2013 we can show you all of the speakers in sort of a nice panel view, so you can see them all, not just whoever happens to be talking.
We can do things like give you HD photos that can be exposed through Exchange, through Outlook, through SharePoint, through Lync, and we can also do other things like give you full fidelity in multimedia experiences in Lync as well. I could be in a Lync conference call with a bunch of people and I can literally pull in a multimedia video and play that and they'll see it stream just as if they were watching it on a native local copy as well. So there's these huge improvements, but of course, to do a lot of that work, which is, you know, dealing with larger data sets, heavier communication protocols, as you start adding streaming video to these conference calls, a lot of changes have to made on the backend, not just to facilitate that functionality but to make sure that it can scale the way it needs to, as well.
PS I've been an Exchange administrator for quite some years, because I've worked with small businesses, that is a scenario that is changing, since Small Business server, in a year and a bit is no longer going to be available. Most small businesses, if they continue to use Exchange, they’ll been using hosted Exchange, either Office 365 or by somebody else going forward, that's probably the scenario we're seeing. So is this a case of Exchange having always been an Enterprise product and then being shoehorned into the small business space for a while and now we're just seeing that's what always was the case, or did you change Exchange again to become even more Enterprise because you can leave the small business space behind?
SS That's an interesting question, I never really thought of it like that.
PS It's because you've never worked with small businesses.
SS Well, actually, that's not true, I work with them all the time and not all of them are SBS customers, many of them just bought the regular retail product and for that, you know, we have the scaling division that scales to five databases per box server and there's a lot of customers who are happy with that and we support all the features in Exchange that we do in both the Standard and Enterprise so they can benefit by, you know, high availability and site resilience, they were just limited in the number of databases they could have.
So I think, Exchange, our target has largely always been anybody who needs email. So that's going to be customers of all sizes; it's going to be small customers, it's going to be big customers; we use terms like small business and Enterprise, the reality is, they’re all companies, they're all organisations, just the fact that one has more seats than the other doesn't really change the functionality of the application, it doesn't change how they would use it, it doesn't change how we might market it or develop it. And in fact, I even talked about this in the Architecture Session that I delivered, the whole point of the major architecture changes that we've made was to move to this building block, or this brick sort of model so that we could scale from the very small shops all the way up to the bigger shops.
So we recognise that there are two ends to the spectrum and I think largely Exchange does a pretty good job on both ends. Now, SBS, of course, they were taking it one step further, on the premise that many of these shops simply don't have an IT person on staff or even the people on staff have no knowledge about how to run server based products and that is an excellent niche to go after, those people needed that kind of help and it was great that we were able to provide it to them.
But Exchange 2013 is actually going mean a lot easier management and as an administrator you probably recognise that things that you used to do that would take you maybe five or six steps in 2003, for example, you're doing in one step in 2010 and in one step in 2013; so we've always been able to target the lower end; you don't have to be a super guru, for example, in order to employ a DAG. We’ve DAGs so easy to do that you literally just follow the instructions, you use the GUI and go through the wizard and next thing you know, you've deployed it, just like, you know, the high end enterprises, just like the biggest companies and very similar to how we would deploy in our own service as well, you know, it's within the reach of all customers now.
I think that barrier largely disappeared around the 2007 / 2010 days actually, I think we became more in reach of the smaller shops than ever before. And to continue with that, I think a lot of that is also about options, too, and so now with the option of small businesses not needing to do the on-premisess product if they don't want and being able to go to a hosted offering of some sort, I think that meets the same needs that were targeting with the SBS product line.
PS What competition do you see for Exchange? Like in a small business space, because that's where I work, on small business servers, when the death of SBS was announced there was a lot of people who got very upset about it and they said: oh, well, there's Kerio Mail Server, for instance, which has made a bit of a name for itself in the small business space and there always used to be, you know, Domino or Lotus Notes or whatever. Do you guys see these as competitors today
SS So from a competitive standpoint, as with any product, your competitors are basically, you know, any of the places where they're not running on the latest version of your software, so that could be, you know, any non-Exchange environment out there and it could be earlier versions of Exchange as well. There are customers who still run Exchange 2003 who haven't, for their business reasons, found the reasons to move them forward. So there's always going to be that sort of competition; you want to bring those people forward, you're glad that you were really successful with them, but you also realise that you're going to benefit from getting them off a ten-plus year old messaging system and onto something more modern, and it's not just about the changes in the software code, it's also about the changes in the paradigm itself.
If you're on 2003 you're running small mailboxes which aren't part of today’s paradigm; if you're on 2003, you're taking backups which aren't part of today's paradigm; if you're on 2003 you don't have native data redundancy, again, which isn't part of today's paradigm. So, we want those people to come forward just as we would want people who aren't running Exchange yet to come and check out our platform and use it for all the advantages that we've been talking about.
PS A very diplomatic answer. Do you have a percentage of the number of seats of Exchange on-premises versus Office 365?
PS As in, you don't know the answer or you can't say?
SS I haven’t anything to share with you at this time.
PS Politicians. One thing I've seen about the new Exchange administrative console is that it’s web based.
SS Yes, EAC.
PS EAC is the now the only interface to manage Exchange, I mean, apart from PowerShell. One of the things I really like about the Exchange Management Console is that it told you what it was going to do in the end of the PowerShell script; I found that a great learning tool, especially for casual administrators or people who are trying to learn Exchange; that's gone in the EAC.
SS Yes, we have received that feedback, we're aware of that. We are working towards full parity with EMC, with the Management Console that we have. In most places you'll find it, that's, you just happened to pick on one of the very, very few places we haven't quite finished up yet.
PS Okay. Finished up yet, as in probably going to be in the RTM version or coming later?
SS At this point, it'll probably be a post-RTM deliverable.
PS Okay. The Edge server Role is gone?
SS That's also a post-RTM deliverable as well. So for the time being, if a customer has Edge, either 2007 or 2010, keep it in place, don't even need to touch it, you can move everything else to 2013 and everything will just continue to work. If you don't have Edge yet, if you're using something else, again, the same thing, just keep that in place in you want and move the internal site over to 2013 and everything will continue. But we do expect to have an Edge 2013 Server post-RTM.
PS Okay. So, what's the one killer feature in Exchange 2013 that you think will drive people to upgrade?
SS So, I think, from my perspective, and maybe it's because it's my favourite area, it's got to be the data centre failover capabilities. I think the multiple databases for volume is awesome and Auto Reseed is awesome, but I don't think we're going to see those come to fruition until those larger drives are available, because right now if a customer's got a two terabyte drive, three terabyte drive, that's perfect for, you know, one database per volume, but as they move forward into 2013, then they're going to be buying perhaps more hardware, and at those times when they start buying these larger drives and they start being able to take advantage of these features, that's going to be killer for them.
But in the meantime, I think it’s all the work that we've done in the architecture and around the namespaces and the ability to have a single global Namespace, the ability to have a data centre failover happen, all of that just, it changes everything. Like I said, the recovery time goes down to 20 seconds for a data centre level event. You can't beat that. And literally, I don't know how we're ever going to beat that. 20 seconds, I mean, the data centre's gone, 20 seconds later, everything's ready to go over there. You can't beat that.
PS No. I think that's it, really.
SS Thank you.
PS TechEd here in Australia is really the first time you got to talk about Exchange 2013 wasn’t it? The preview wasn’t released at Tech Ed US or Europe?
SS That’s right. And needless to say, we got a lot of questions there and we all knew it was coming, but we couldn't say anything and it was probably more difficult for us than for anybody else. I mean, it seems like 2010 just came out, but the reality was, that was coming up on three years now.
PS It does really feel like it was yesterday, it really does.
SS I know. And for those of us who've been working on 2013 for all this time, like 2010 feels like a distant memory.
PS Thank you very much for your time, Scott.
SS Thank you. It's a pleasure, Paul.