Explore Microsoft Windows Server 2016 Software Defined Data Center

Explore Microsoft Windows Server 2016 Software Defined Data Center


Good morning, how’s it going?>>Good.>>Are you guys enjoying the show so
far?>>Yes.>>Yep, good sessions,
good speakers, cool. Who in here was in my
session yesterday? A number of you, okay. So I don’t have to repeat the joke
about my name from yesterday.>>No.
>>Right? That’s okay. Let’s skip that. [LAUGH] Well, thanks again guys for
being one of my sessions. For those of you that were
not in my session yesterday, I’m sorry for you. It was a good session, right? Just kidding. Everything is being recorded. All the content is
available online as well. You can see the recorded
from midnight, you can see the recorded
from Tech Summit. It’s up to you. So don’t panic that you
are missing good content. You can review everything
online as well. I am Vinicius Apolinario. I made a joke about my name
yesterday, I won’t repeat. I usually do, but I won’t. I work on the Windows Server team. Actually in the marketing
organization for the Windows Server stuff that we do. So I’m a technical person
inside of the marketing group, so we can have
a technical conversation. Actually, that’s the idea
of this session, to talk about the features inside
of Windows Server 2016 for software-defined data centers, so
think about compute, storage and networking, and
what’s new in Windows Server 2016. Those are the things that we
are going to cover in this session. I prefer that when we have
this kind of sessions that you guys take advantage
of being here actually and so you guys are able to ask questions. So if you have any questions while
I’m presenting although I can barely see you guys, just raise
your hand ask the question. We don’t have any mics in the room
so I will repeat the question so everyone can hear the question. Although this is kind of a small
room, I’ll repeat the question and we can have more of a conversation
than actually just me presenting and you guys just hearing. I think that’s more beneficial for
everybody, so let’s get started. Like I said,
this is the probably the most data center focused session that
you’ll find in the Tech Summit. The reason why is because we are
really focusing on this session on Windows Server 2016 and what we
have in Windows Server 2016 for compute, storage, and networking. So what’s new in
Windows Server 2016? So because I want to have this
conversation with you guys, let me start by asking
some questions. How many of you guys are already
using Windows Server 2016? Okay, a good number. How many of you guys
are using 2012 or 2012 R2? That’s good. 2008 or 2008 R2, yeah. 2003? 2000?>>[LAUGH]
>>NT.>>[LAUGH]
>>No one using NT anymore? That is a great story. I come from Brazil and I was in the
Brazil subsidiary of Microsoft and we got someone from engineering team
visiting one of the largest bank in Brazil and then the guy said, look,
I hear everything you are saying. It’s all great. But you know what? We are still using Windows NT. At that point the engineering guy
that was talking to the customer said, if I was your customer and I knew about that,
I would end my account right away. And at that moment,
okay let’s switch gears, let’s talk about something else. We know that you guys have to
support old applications or legacy applications. So there’s a big investment in
Microsoft to make sure that and this is what we are going
to talk about and that’s the reason why
I’m talking about this. We are making a big investments on
things that we understand you guys are going to take advantage of,
right? In the new products. At the same time we understand
that you guys have to maintain what you already have running. So what are the things
that we are looking for when we think about software
define data center? Where we are in
the evolution in terms of what’s coming and
also what we are bringing from the legacy that we have
inside of your companies, as well. Because of that, what are the
customers that you’ve got facing in terms of upgrading, in terms
of performance, in terms of cost, flexibility, reliability, and so on? And what are the features inside of
the product that we will actually help you guys take the next step,
right? So this is the main
focus of this session. So like I said,
let’s start looking back, right. So if we look back through
the investments that we did and what was the focus of the product
at that specific point in time, when we launched Window Server 2008,
R2, along with the System Center components to
manage the features in the platform. So think about Windows Server
as the platform and System Center as
the management base, to manage the platform that we
release with Window Server. So that was the moment in time
where we actually introduced virtualization. Let me rephrase that, Server X86 virtualization
to the platform, right? Because if we think about
virtualization you probably know that virtualization is
a very old technology. It’s been out there for decades, but we are talking about hypervisor
server X86 virtualization. So basically, we introduced
Windows Server 2008, and then we launched an update
that introduced Hyper-V 2008. I don’t know how many of
you remember that, but Hyper-V was an update for
Windows 2008, right? So at that moment, what we wanted was to actually take
advantage of virtualization and having virtual machines being able
to better use the hardware and start taking some advantages
of virtualization like it’s easier to back up,
it’s easier to deploy and so on. In the Window Server 2012 we
introduced some new features that started to change
the virtualization market, right? So we all know that for example
VMotion is a feature from VM where they launched this feature, we came
up with Live Migration that is the equivalent to VMotion in the VM
space, in the Microsoft space. But in 2012 the Window Server
2012 we started to do some things that VM was actually starting
to catch up to Microsoft. So think about channel off and
a live migration okay. So VMotion, VMware, but for
VMotion and live migration to work you need a charge storage and
then you move the virtual machine between the hosts that
you are trying to move, right? Shared nothing live migrations
removes the need for shared storage. And why did we launch
something like that? Because we heard from customers. Customers said, look, live migration
or VMotion are great, but we need to move virtual machine regardless
if they have a shared storage. I have a stand alone host,
I have another stand alone host, and I want to be able to move
the virtual machine between the two hosts regardless of
the shared storage, right? If we think about this kind of
thing, Hyper-V replica there are a number of other things that
we launched with the product that VMware came up with equivalent
in their side as well. So what I’m trying to say here is
that for a long period of time we’ve been in this battle
of catching up on features. Right, but if we think about
the features we launched to 2012, and 2012 R2 as well, we were actually looking at
the pure infrastructure, right? So what can we do to
the infrastructure to support the applications that are running on
top of this infrastructure inside of virtual machines, right? Now, things started to
change at some point. And that’s why in Windows 2012 R2,
our design point was actually Azure because then when you move to
a design point like Azure, you change the perspective and
you have to change the way you do the infrastructure to support
the virtual machines or the applications. Let me give you an example. If my application can see
default in the infrastructure, why do I need high availability? Think about it, my application can understand if the infrastructure
has a problem. So my application by itself can
remediate The application or the infrastructure
problem via software. Why do I need a cluster? One of the technology that is
coming out that is similar to that, you ever heard about Micro Services,
right? Containers, right? Those technologies, they are moving to a level where
the infrastructure’s actually no longer that important from
the application standpoint, right? Now, that doesn’t mean that
the infrastructure is not important to support the application. It’s just that it changes
the agility that is needed from the infrastructure for
the application to run. And that’s the reason why software
defined data center is so important. Software defined data center,
if you think about it, is infrastructure catching up to
what development is doing, right? In order to deploy an application,
I have to spin up storage, compute, and networking. So let’s take one of those,
networking. How long does it take to reconfigure
VLANs in your environment? Is it something easy? Is it something fun? Right, I understand the pain. The developers don’t. They don’t care, right? The only thing that they know
is that when I need just pin up an application, it takes about two
weeks to deploy the application. What if we could deploy this
network configuration via software. And that would take minutes. In some cases hours, but
it will take minutes to deploy this network configuration
for my application to work. That’s the difference from
software defined data center to the old way of doing infrastructure,
right? So when you think about
2012 R2 in Hyper-V, a lot of people will still think
you know what I’m sticking to VMware because we’ve
been using VMware. Or I don’t really see the phase
that you guys are doing so one of the things that we decided
to do is there is a slide that basically list the features that we
launched between 2012 and 2012 R2. And I don’t expect you to read and
understand and actually see what each
of the features are. I’m just showing you that we’ve been
doing a lot just for Hyper-V, right? We’re not talking completely about
storage or networking, just for Hyper-V. The things that we are doing to
support your applications, and make sure that you have the best
in class hypervisor available to support your virtual machines,
right? Now everything that I’m talking
about is this notion that we need to rethink our data centers, right? And the main reason why we need to
change this is because if we think about the pressure that exists
between the application, and the application owners, and the development people,
and the infrastructure. Let’s think about what
is the requirements for it from the application perspective. They need agility,
they need to deliver faster cadences of the updates
of the application, and so on. And from the IT perspective, by the way I’m assuming that most of
you guys are from the IT side right? Do we have any developers here? No, okay, so can I say bad
things about developers? No one cares? Okay. [LAUGH] So if you think about what
are your requirements? You have an SLA, you need to make
sure that applications are up and running, you need to make sure
that everything’s secure. You need to make sure the
performance that the application has is the performance that was agreed
with your customers, right? So those are things
that you care about. If we compare the things
that application owners and developers want and infrastructure,
that’s the main reason why IT is seeing inside of the company
as the blocker, right. Because every time someone
tries to create something new, we are the first ones to say,
okay, that’s not gonna work. I have an SLA to maintain. I can’t do that. But so SDDC,
Software Defined Data Center, is the enabler for
the IT and infrastructure to change the way we do
IT in the data center. We move from the traditional way to
a cloud model because cloud model is what enables IT to deliver
what application owners and developers are looking for, right? So, all this introduction
is basically to say look, when we launch Windows Server 2016
along with System Center 2016, those were the things
that we had in mind. So if you think about in Windows
Server 2016, what are the three things that we tried to achieve
by launching Windows Server 2016? First, we wanted Windows Server
2016 to be the most secure OS possible for the application and
the hypervisor, right? So the ones of you that were
in my session yesterday, you know what I’m talking about. Then we want to make sure that
Windows Server 2016 along with Systems Center is
the best-in-class platform and management for the data center. Of course,
we do a lot of things in Azure and OMS, the Operations
Management Suite. But for
the on-prem side of the house, we have Windows Server 2016 as the
best in class software defined data center solution today, right? And then the next step, which is
the next session that I have after lunch today,
is the application platform, right? Application platform
being containers, Nano Server how Window Server can
really support the new types of applications that are coming out. Software defined data center is
really in the middle, right? Because software defined data center
provides the three components computes, storage, and networking that touches
security of your data center. And also how you’re going to
move your applications to the next level right? So this is Windows Server 2016. So, Again let’s talk about compute,
storage, and networking. So let’s start with
software-defined compute. And when I say software-defined,
one of the things you have to remember is everything
that I’m going to talk about now, instead of talking about
this transition and this kind of communication between
infrastructure and application. I’m going to talk about the features
that we have in the platform. One of the things that we have to
remember while I’m talking about the features itself is everything
that we are putting in Windows Server 2016 is actually
one of the components of the software-defined data center
story that we have for you to support the new applications that
you have inside of your environment. Now, let’s start taking
a look at what were the drivers to do compute
in Windows Server 2016. So these are things that we heard
from customers in terms of compute. Performance is very important. I need to be able to deliver better
performance for my applications, and that means that software
needs to talk to hardware in order to deliver that
performance that I’m talking about. The other thing is reliability. So, I can’t have any
kind of down time because of infrastructure
transition, errors, or failures or because I’m just changing the
configuration of the infrastructure. Changing the configuration of
the infrastructure shouldn’t cause downtime to the application, right? And the last one is flexibility. So, I need to be able to first make sure that the applications
that I have are running. And also I need to be able to choose what is the platform that I
wanna run my application. So one of the things that we did as
well was to make sure that no matter the application framework, no matter
the operating system that you are running on, we wanna make sure
it runs great on hyperview, right? So, these are the features
in Window Sever 2016 for compute to address the areas
that we are talking about. So from performance perspective and we are going to enter in more
details on each of these. We have RDMA,
high performance live migration. So we are still investing
in live migration. We understand that we have a lot of
legacy application that still runs on virtual machine. And if you have any kind of
planned maintenance, you need to move the virtual machine around
and you need to do it faster, right? Virtual machine Kiwi,
VM load balancing, start ordering of virtual machine
in SMB, investments in SMB as well. All of these are related
to performance. So we wanna make sure that, in terms of performance,
as well as scalability, right. We are the best in class. Now, one of the things
that you won’t see, Is a comparison like is
Hyper-V faster than VMware? Right. You won’t say something like that. Why? Because VMware does not
allow us to do that. One of the things that
is the VMware product their terms and agreements. They don’t allow any third party,
including Microsoft, of course, to present any kind of
research or performance comparison. So one of the things that I would
actually provide to you guy is test Hyper-V if you are not
using Hyper-V today. How many of you
are using VMware today? The vast majority of course. How many of you are using Hyper-V? How many of you are using
both Hyper-V and VMware? Okay, for the ones of you
that are not using Hyper-V, I really encourage you
to really test Hyper-V. So first of all Hyper-V is free. You have the Hyper-V server
option that you can download, or if you have any Windows Server
license you can also use Hyper-V. Because we can’t prove
that we are faster, that we have a better performance,
test for yourself. Install ESXI host in your hardware. Test the performance. Install Hyper-V test the performance
and take your conclusions. Reliability and some of the things
here are specifically about 2016, only coming from 2012 R2 like for
example cluster OS rolling upgrade. Great new feature but
it’s for 2012 R2. We are going to talk
in more details. Shooted virtual machines, I won’t
talk about this one today because we talked about this one yesterday. I encourage you guys to
check what this feature is. It’s way ahead on whatever other
hyper advisor you may think about, protection against
the rogue administrator, or compromise account from
your fabric administrators. Flexibility we’re going to talk
about some of the things that are also for storage and compute, but makes
sense from the compute perspective. So, let’s take a look at
some of the features. Before I do that,
one of the things that, during the technical previews
of Window Server 2016, a lot of customers asks us
is not Microsoft going to extend the numbers in
terms of scalability? And the question that I usually
come back to customers and ask is are you really using more than 4 TB
of memory on the physical server? Are you using more than
320 logical processors? Are your virtual machines using
more than 1 TB of memory and more than 64 logical processors
per virtual machine? And some customers sometimes say,
yes, I have one workload that
needs more than that. Does anyone in this room
needs more than this, in Windows Server 2012 R2? Yeah, that’s usually the answer. No one is using more than this. But, yeah, you know what? There’s one customer that
came to us, and said, look, I need higher numbers. I need to run more than 1 TB per VM, I need to run more than 64 virtual
processors per virtual machine, and by the way my hardware
is way bigger than that. You guess which customer is that? Azure, Azure is the largest
Hyper-V customer. And believe me when I say this, the Hyper-V that they run
is exactly the same, right? The proof of that is that we
are giving Azure Stack to you guys later in this year. And guess what,
that’s the same hypervisor. Now the layer on top
of the hypervisor is a little bit different,
that’s Azure Stack, right. But the hypervisor
is exactly the same. And they came to us and
said look, this is not enough. I’m running workloads for
machine learning, log analytics,
I need more than that. So, for Windows Server 2016,
we came up with these numbers here. 24 TB for the physical server,
in terms of memory, 512 logical processors, and
like huge virtual machines, 12 TB of memory for
virtual machines and 240 virtual processors
per virtual machine. So that’s just a proof that
if you have any workload that you wanna virtualize in terms
of performance you can get it. Take a look at the Windows Server
blog that we did in conjunction with the SQL team where we
are talking about how SQL Server 2016 is taking advantage of these
numbers in terms of performance. It’s amazing what
those guys are doing. So let’s take a look at
some of the features here. Now RDMA is probably one of
the main things that we did in terms of performance, especially
compared to our competitors, right. We are the only one to support RDMA. And the one benefit that we
can talk about RDMA is really think about live migration for
example. The process of doing a live
migration is basically looking at the memory, storage, and the state
of the virtual machine, and moving that virtual machine from
this host to that host, right? Now in order to do that,
you have CPU cycles, right? So you have a performance impact
every time you run V-motion or live migration, right? When you use RDMA, basically
what you do is your offload all that process of the live
migration process to that NIC, to the network card, right. So in terms of impact to
the performance of the CPU, you see very little impact for that
performance because you do that. One of the things we’re going
to talk about is the storage, hyper-converged storage. When you think about
hyper-converged, you absolutely need RDMA
because when you start to scale you need to make sure
that the processing needed you have your shared storage
between the multiple nodes. You need RDMA in order to
off load the process from the processor to the network card,
right? That takes us to high
performance live migration. We have some videos
available up on YouTube and Channel 9 showing performance for
live migration. I mean it’s crazy what we can
do with RDMA for live migration. VMQ is something that we
launched for Windows Server 2012. In Windows Server 2012 we introduced
a VMQ, that basically, when you have multiple virtual machines running in
a single host, you need to make sure that they have the correct
access to the network card. What we did for Windows Server 2016,
is going even further when we created a multi-queue that basically
combines VMQ with var SS and we have also VRSS to
the virtual machine as well. So in terms of performance there’s
a lot inside of Windows Server 2016. Another thing that we realized is
look, we are in terms of density getting to a point where hosts
are running more and more virtual machines, and we are running out
of capacity inside of those hosts. Now, that’s not true for all the hosts that you had
inside of your environment. So, one of the things
that we did was first, VM Load Balancing and
the Start Ordering. So they are kind of different but
they complement each other. So the first one is Start Ordering,
and Start Ordering really is if
I have workloads that depend from each other, like for
example, I have a website, but for that website to work the SQL
database needs to be on. For the SQL database to be on, I need my Active Directory as well,
right? So I need to start my virtual
machines in this exact order in order to have my
application up and running. So in Windows Server 2016,
we introduced at this through the platform not just to
the management layer anymore, right. I need to remember to get
on the Skype every time I’m going to present. And VM load balancing does
exactly what the name says. So, if I have multiple virtual
machines running in a single host, and I have another host that still
have some capacity available. Now Hyper-V can see that and move that live migrate
the virtual machine between those hosts to move the processing
from one host to the other. So I can have equivalent
performance in both of the hosts. SMB is basically our protocol to For storage, right? So if you think about scale out file
server, basically what we do is we bring the communication should by
using the SMB protocol, right? So, one of the things that SMB adds, is the capability of
using multichannel. So, think about servers
with more than one NIC, instead of using
just one single NIC, you can have multipath for
SMB in Windows Server 2016 as well. In terms of reliability,
like I mentioned before, one of the things that we are trying
to do for Hyper-V is every time you have to change the configuration
off a Virtual Machine, or you have any kind of
transient failure. We want to make sure that
the workload is just still up and running, right? So, for example, let’s say I have
a problem with the cluster service inside of one of
the nodes in my cluster. Usually what that will cause is the
virtual machine to fail over from host A to host B, right? So, that means the virtual
machine is shut down, and started in the next host, right? If I have any failure in the host
now, in Windows Server 2016, the first thing that
the cluster will try to do is keep
the virtual machine on. And see if the error
resolves itself, or if there is transient error
like a storage access, or network access, or
something like that. So, we wont just turn off
the virtual machine, we either safe the state, or we just keep
the virtual machine running, right? So, for example, in the case of
the cluster service goes out, the virtual machine
continues to run, right? Although the virtual machine is
running on top of a cluster, if the cluster service
is not working, the virtual machine
continues to work. Another thing is,
we enable now a hot add and remove of memory and network. Of course this is not something new
for you guys that are using VMware, but for Hyper-V, this is something
that we add for in the Server 2016. So, we understand, we are again
catching up on features. So, if you have
a Windows Server Hyper-V host and you have a virtual machine up and
running, and you wanna change the memory
of the virtual machine. Usually what you will have to do
is turn off the virtual machine, change the memory, and then bring
the virtual machine back on. Now, we can change the memory
of the virtual machine. There is one point here, which is the guest OS needs to
support that as well, right? So, the guest OS needs to be Windows
Server 2016 or Windows 10, right? Network does not require that
the operating system to be 2016 or Windows 10. It could be 2012 if I’m correct, but
you can add additional networks, additional NICs to
the virtual machine. This is a great one, especially for customers that are using
guest clustering. And guest clustering is one of the
things that I think that we are way ahead of our competitors. Guest cluster being I
have virtual machines and instead of doing the cluster
in the infrastructure layer. I do this, or the hypervisor layer, I do this cluster in the virtual
machine layer, right? So, I have physical host, I can
have cluster in the physical host. I have virtual machines, I have cluster for
those virtual machines as well. Now, in order to do that, the virtual machine will also
need shared storage, right? How do I enable guest
clustering in that case? Well, in Hyper-V you
have four option. You can use virtual fiber channel
for the virtual machine, so they can talk to your regular fiber
channel storage device, and so on. Another option is you can
have shared VHDX, right? The VHDX being the equivalent for
the VMDK for those of you who are running Yammer. So, I have a virtual disk file that is shared between
multiple virtual machines. And all of the virtual machines can
access the same file at the same time, right? One of the things that we added for
Windows Server 2016 based on the response from customers that
are using this technology is, I wanna be able to resize this
file without stopping the virtual machines, or the workload that those
virtual machines are running, right? In production checkpoints,
this is a minor one, but it’s very interesting. How many of you guys
really uses snapshots in production environments? Okay, this is just double checking. You guys know all the effects
of checkpoints, right? Snapshots, right? If you don’t know what
I’m talking about, please stop using checkpoints,
right away. First of all, checkpoints or
snapshots are not backup solutions. You all should know
that at this point. But you can cause problems to
the application when you restore. By the way, creating a snapshot, there’s absolutely no problem
in creating a snapshot. The problem is restoring
the snapshot, right? When you apply a snapshot
to a virtual machine, that virtual machine
goes back in time, right? So, you can cause problems to
the application when you do that. Especially applications that are
multi-node like Active Directory, Exchange and so on. Databases and so on. Now, one of the things we did for Windows Server 2016 is
production checkpoints. Now, what production
checkpoints are, are the regular snapshots that
you guys are familiar with. But when you do that, we start the V
assess inside of the guest OS. So, that’s snapshot is basically
application aware, right? So, the application whether
you restore the checkpoint or the snapshot, the application
is aware of that, right? So, just wanted to make sure
that you guys are not creating problems for your environment. For reliability, there are a number
of things that we can talk about. First one, and this is the one
feature that I really like, not for now, I think, this feature,
in the future, will represent a lot. And what cluster OS rolling upgrades
is is basically if you think about every time you have to upgrade
to a new version of Windows Server, there’s always this process
of upgrading your servers. But especially for a cluster, it’s usually a very painful process,
right? And the reason is you have this
cluster of a number of hosts, and if you want to move to a new
version of Windows Server, you have to spin up a new cluster. Now, you have a number
of options to do that. You can buy new hardware and bring
the new hardware to the new cluster, and then start moving the workloads
from this cluster to that cluster. Or you can evict one of the nodes
in your cluster if you have enough capacity to do that, and move
the work loads to the other nodes, then take this hardware and
bring to the new cluster, and then you can do the same thing. The real problem of this process is you have to spin up a new cluster,
right? So, Cluster OS Rolling Upgrade is
nothing more than allowing you to bring a new Windows 2016
to your existing cluster. So, in that case,
you have a mixed-OS mode cluster. What that means is, you have nodes
running Windows Server 2012 R2 and that’s the requirement. And you’ll have also Window Server
2016 running on the same cluster. So now you can live migrate or fail over workloads between
versions of the operating system. You can live migrate a virtual
machine from 2012 R2 to 2016 in the same cluster, right? And the approach we took to do that
is almost the same as we have for Active Directory. So, Active Directory you
have the domain, and forest functional level, right? For cluster, you have the cluster
functional level now. So, what happens is if you have
Windows Server 2012 R2 running, you have the 2012 R2
cluster functional level, you can bring 2016 to
that functional level. Once you move the work
loads around so you can update the host,
then you can update as well the cluster functional level
to Windows Server 2016. So, all the other features
that we are talking about for cluster are also available for your cluster without stopping
the virtual machines, or just with a small failover for
your other workloads, right? VM resiliency is basically
what I was talking about, in terms of making sure that if
I have a transient failure or if I have a minor failure in the
cluster, the virtual machine won’t be failed over or
won’t be stopped because of that. The other thing that we have is
the fault domain-aware cluster. Historically if we think
about cluster, the cluster in the Microsoft story is I have nodes,
and if this node fails, I have other nodes in the same site
to failover that workload, right? With Window Server 2016, what we are
doing is we are bringing the notion of failover to modern nodes. You can establish racks,
you can establish sites, you can go further from just nodes,
node failover, right? So one example is I
have site Chicago and site Seattle for
the same cluster, right? And if I have a failure
in this Chicago site, I can failover the workload
to Seattle, right? Of course,
this is a extreme example. I’m just trying to convey that,
you can have multiple sites for the same cluster, right? In terms of flexibility, QoS,
or Storage Quality of Service, this is something that
we are bringing to 2016, as an evolution from 2012 R2. So 2012 R2 introduced
costs per disk. Now you can apply policies
based on workload, you can use System Center
to even further have more granularity in terms of how
you apply the policy for cost. For Linux support, we are extending
the number of communities and companies that we are working with
in order to make sure that those workloads are supported in Hyper-V. And this is a very interesting one. So if you run a Hyper-V environment
or a VMWare environment or I would say whatever
hypervisor it is. You know that sometimes you
need to update the integration tools or the VMWare tools
that you have inside of the virtual machine, right. So one of the things that
we are doing for 2016 is we are making the integration
services part of Windows update. So whenever there’s an update
available for integration services, it will automatically update itself. Right, of course, it goes through
all the process of the business or configuration manager that you have,
but this allows you to have the integration services in the
guest OS, updated through the host. Compare it to the host, right? So with of all that, one of
the demos that I chose to show you today is the VM load balancing. I’m not sure if there’s audio
in this one, so let’s see. There isn’t.
So, let me just explain what we have
here before we get started. What you see here if
you’re not familiar with Microsoft, here you have
a cluster called EldenC. This is Elden Christensen’s
environment. You have the rows,
virtual machine, and the nodes supporting those roles. And in this case, I have two nodes. One of the nodes is up,
running four virtual machines. And one of the nodes is down. So all of the processing is
running in a single node. What he is going to do is he’s
going to spin up the second node, you go ahead and
right-click Start Cluster Service. And what happens here is
if we go to the Roles, you’ll see after one minute because
that’s the time that it takes from the moment that we check
the capacity of the nodes. You see that the first
virtual machine is being live migrated to
the second node, right? Now after 30 minutes,
what will happen is the Cluster Service will check
the capacity of the host again. And if necessary, it’s going to
migrate another virtual machine. What we did was we
forced the process, so we don’t have to wait 30 minutes. And you can see that the next
virtual machine is moved to the second node. So the workload is balanced
between the two nodes that I have running now, right? In here we forced the 30 minutes,
but you actually don’t
have to do anything. It will automatically live migrate
virtual machines around to make sure that the workload is balanced
between the multiple nodes that you have in your environment, right? It’s one of those things that
I’d say is pretty cool, but you don’t actually
have to do anything. All right.>>[INAUDIBLE]>>Yes.>>So the question is, I don’t wanna
have a specific virtual machine or a specific set of virtual
machines in this specific nodes. Yes, you can do that. You can create preferred hosts, you can set a virtual machine or
a set of virtual machines, to evict those nodes,
you can balance that as well. And you can turn off the load
balancing here, right, if you want. Okay, so with that let’s No,
I wanna move. I wanna, yeah. Let’s switch gears to
software-defined networking. And this is the most
interesting one, I believe, in terms of trying to explain the
value of software-defined network, and so let’s see if we can do that. When we talk about networking,
how many of you guys have ever configured a VLam in a real
enterprise grid environment. I do. Is it a fun process? Yeah. How long does it take to,
well actually let me rephrase that. Another interesting thing. Imagine your company today and imagine that you have to
reconfigure all the VLANs. [LAUGH] That’s not a fun process,
right? That will take time. That will, a lot of things
should take in consideration. And the reason is, everything
is tied to the hardware, right. All the configuration
that you have to do, you first, some things that most
of our customers are saying to us, we have multiple vendors and it’s
hard to stick to a single vendor. We do when we can, but it’s hard to. Even the same vendor, they have
different versions of software. So, sometimes
the management solution for switch A doesn’t talk to switch B,
right. Reconfiguring all the switches,
routers and so on is painful. And every time that someone’s
deployed a new application, you have to reconfigure something in
the switches, or the routers, right? So what happens is that, you have multiple layers in your
network that you need to configure every time you’re deploying
an application, right? What we are doing with networking
is the same that we did for hardware, right? If you think about it, every time
you had to deploy an application in the physical server days, you have
just spin up a new server, right? What we are doing is,
we are taking all of the hardware topology that you have for
networking, and this is some of the challenges
that I talked about. So what we’re doing
is we are combining all of the capabilities from your
network devices into a single layer that abstracts the physical network. Should the logical network or
the virtualized network. If you think about it, it’s exactly
the same on what we did for servers. A virtual machine sees
the hardware as if that hardware belongs to that virtual machine,
right. A virtual network will do exactly
the same, but instead of a server hardware, it’s a network hardware or
network device, right? So on top of that, what we do is we are bringing the Azure inspired
Software Defined Network solution. So this is basically, we’re gonna talk about
name of the features here. So we are bringing the same
capabilities that we have in Azure to run in our host, in our nodes
inside of the Azure Data Center, and we are bringing
to Windows Server. So the main component to enable
all of this to work is this guy, the network controller. So the network controller is
a feature inside of Windows Server 2016 that will oversee the multiple
Hyper-V hosts that you have, and you can also integrate with
your physical network, right. And because it’s integrated with
your physical network and also we owned a virtual switch running on
Hyper-V, now we can run the virtual network on top of the physical
network that you have, right? One of the things that is
interesting about this, so aside from the network controller,
what we are doing here, this is not new, we already had. Virtual networks in
Windows Server 2012, and 2012 R2, and system center. The only difference is, well first, system center is not
a requirement anymore. You can bring system center and that
will help you manage the virtual networks that you have, but you can
do everything through PowerShell, and that’s the demo I’m
going to show in a minute. But the other thing is in 2012 and
2012 R2, we were using the NVGRE protocol. That is not a Microsoft protocol,
but that is the protocol
that we adopted. In 2016, one of the things that we’re doing
is we are supporting VXLAN as well. So if you are using VMware today
it’s a easier transition to the Hyper V and
the Window Server networking, virtual networking if you
wanna do that migration. So how does that work? So when you can abstract the network,
the physical layer from the network, to the software, what it can do is
there are no limits now anymore in terms of blocking
application to be deployed. Via software I can do
everything that I’m going to talk about in those slides. So and
then I’m going to show in a demo. So here I have kind of a three-tier
application, so in this case we have Active Directory in the back
end instead of a database. But you have the web servers, the
virtual machines for the web tier, you have file servers here that
communicated with the web server and the file server needs to talk
to active directory in order to run properly. The first thing I want to do is
I wanna create the subnets for those tiers that I have. And I have a PowerShell
script to do that. On top of that we wanna make sure that those subnets are part
of the virtual network. So if you ever deployed a virtual
network in Azure that’s exactly what to do in Azure. You create a virtual network, you create the subnets and
you’re good to go, right? But what’s the point
on creating this if you can’t talk to
the outside world. So one of the things that we have in
Window Server 2016 is a gateway that it can connect the virtual networks
to existing networks in an Internet. We also added a load balancer so
it’s not the regular load balancer that you guys are familiar with from
older versions of Window Server. Is a brand new load balancer
the one we’re using in Azure. You can create ACLs in order to say
what is the traffic that is allowed between the tiers and which tiers
can communicate with each other. And you put that as a virtual
network on top of your existing network. So here are some of
the things that are new. So the network controllers. The other things that you see
are things that were there in Windows Server 2012. And with system center virtual
machine manager you are able to manage. So we are decoupling the virtual
network from systems center. You can use Windows
server only now and we are also adding some supports for
example, VXLAN and so on. So let me show you a demo, so you can better understand
what I’m trying to say here.>>Let’s find out what’s new, Windows Server 2016 with
software defined networking. Before we get started, let’s first
understand the problems software defined networking for
SDN is solving. Applications are innovating
at a rapid pace. And no key infrastructure, with
its complex set of connections and devices, can keep up to meet
the needs of those applications all while ensuring constant uptime. Server virtualization
solves this problem for servers by placing
the application into a virtual machine that is separate from
the hardware on which it’s running. With SDN, we create a network
virtualization layer on top of the physical network so that
the requirements of each of your applications are met
consistently and independent of the configuration
of the physical devices. This makes your applications
run more reliably and securely while reducing the churn
in your physical network config. Even if you grow your physical
network infrastructure, you don’t have to worry about
breaking the apps that run on top. And your apps can immediately
take advantage of new resources. Let’s go hands on to
see it in action. We’re gonna start by creating
a virtual network for two-tier web application. We give this application it’s own
network, keep it isolated from any other applications on
the same infrastructure. In this network we’ll
create two subnets. One for
the front end web servers, and one for the back end file servers. We start by creating set of objects
that represent virtual network. And within these, we add a virtual
subnet for each of our app tiers, we then apply these settings
to create the network. Once the network’s created,
we can attach each of the VMs to it. In this case we have a helper
function to do this. We call it once for
each of the VMs we wanna connect. Now let’s see the results of
one of our web server VMs. IP config shows the IP address
that we’ve been assigned. We can do a ping to see basic
connectivity to another web server. We can open up a browser and connect
to the default IIS web page on the server to show that yeah,
it’s working. In this demo we created two subnets. And since these two subnets are part
of the same virtual network, Virtual Router got
added automatically so VMs can communicate across subnet
boundaries with no extra effort. This is great for one VM. What happens if the load
on your application grows? Load balancing becomes critical
in order to scale out. And so Windows Server 2016
Datacenter includes an all software Load Balancer for use with SDN. You can use the software Load
Balancer by creating virtual IPs or VIPs on top of your
application tiers. These VIPs are the single entry
point into the multiple VM instances of an application that
make up the tier. As the needs of
the application grow, the number of VM instances
can grow as well. The load balancer will
continue to make sure that the load is distributed evenly. To achieve this in the next
[INAUDIBLE] after selecting a front-end virtual IP for
my services. Creating a back-end address pool to
contain the VMs that will respond to the request to the VIP. And finally I am connecting the two
with some load balancing rules. In this case I’m saying,
TCP on port 80 of the VIP will get distributed to
the backend pool members. Apply the load balancer
configuration. The next step is to add the pool
members to serve the request. I do this once for
each of my web server VMs. Applying it only
takes a few seconds. Now let’s see the result. I placed a uniquely colored page
on each of the Web Server VMs. So depending on which VM serves
the page, I get a different color. Of course, if I go to the main page,
the app that I’ve actually deployed, in this case a blog,
you can see it’s also working. The software load balancer also has
the ability to insert internal FIPs within the virtual network for
load balancing between the tiers. While it’s not required for the file
servers in this example, it can be used for other apps that require
databases or application gateways. So you know your services don’t live
in a vacuum, they need to connect to Active Directory,
databases and other servicers. For that STN provides
hyperconnectivity to the networks containing other existing services
in the same remote data center. The application network finds their
own gateway configuration that once again uses shared infrastructure
to connect to the outside world, with minimal involvement
from the network admin. Types of gateway connections
are as flexible as the needs of the application, and
include Multitenant, VLAN Routing,
Site-to-Site Tunneling, GRE Tunneling or
any combination of the three. To show you this in action, for this application we’re gonna connect
our virtual network to VLAN L-340. Here we’re starting one
of the web server VMs. This time, we’re trying to
go out from the network. I start to ping to
show it does not yet have any outbound connectivity,
so let’s run the script. Okay, the setting’s been applied. While we’re waiting for
the configuration to go into effect, let’s take a closer
look at the script. The script first retrieves
the objects that I’ll be connecting. In this case, my virtual network
is being connected to a VLAN in the physical network. This VLAN Id, 1002, has been
preconfigured in my physical switch by my network administrator. It then creates a virtual gateway
service on the virtual network and adds a connection referencing
the VLAN’s subnet from the physical network. It gives the virtual gateway
an IP address on the VLAN subnet. Switch back to the VM to watch as
the connectivity gets established. Wait for the ping,
there you go, connected. And we can do a trace route to
see each hop that it takes on the way out. There is the distributor router. In the virtual gateway, the physical
router, and finally the destination. And now I have the full
ability to send traffic in and out of the virtual
network via that VLAN, with this ability to create and
connect to your network. The ability to protect
it is critical. Cyber crime costs organizations
millions of dollars each year. SDN can help here as well by using
the included distributing firewall, you can create micro segmented
boundaries around individual tiers for your application. Permitting traffic to only those
flows that are required and denying any flow that is not
needed for the app to operate. When these rules are applied to the
subnets, each VM automatically gets the required rules that
are unique to its tier, even if that VM gets
added at a later time. Let’s add some protection
to our web app now. We’ll start by defining
the rules for our Web tier, and then apply those rules to the Web
tier subnet That’s all it takes. Let’s go to the VM. We’ll start by flushing DNS, to show that DNS lookups
are still permitted. Now, we do the ping, we’ll get
the IP address back for the name. But the ping itself is blocked, because we did not allow that in
the rules We’ll open up the explorer to check to see the file server
access is still permitted, cuz that’s important for
the application to run. We’ll use the UNC path and
connect to it by name. And there we can see we can browse
the file share without any issue. We can also open up Internet
Explorer to verify that the app itself still runs,
by connecting into port 80. So you can see to the webpage
it’s still getting served. You aren’t limited to
Microsoft only technology. If you’ve already tooled up by
investing in a third party with a virtual appliance that runs on
HyperV, you can add an instance to that virtual appliance, to the
virtual network and send traffic to it by taking control of
the virtual networks router table. This is called user-defined routing. Here we are adding Linux
appliance to our virtual network. You’ll see the Linux
appliance to the right, and you can see it has some
background traffic to start. Cuz we haven’t set up the rules yet,
so let’s run the script you can see it’s applying this routing
table to the virtual network. And within the second you can see
all of the traffic destined for the dot 2 subnet, is now being sent
to the Linux virtual appliance for it to inspect and
forward as it sees fit. Everything you’ve seen has
been done in real time, done by the application owner
without having to take time away from multiple administrators. And it was done without losing any
of the performance you expect from your high performance hardware. When you’re deploying your next app
in your existing infrastructure, think about how much easier it could
be with the agility of SDN and go try it with
Windows Server Datacenter. It’s all included in the box.>>Okay, two things that I
know you guys are thinking. First, I’ll ask some
questions in a minute. First, how many of you guys
really love PowerShell? Yeah, I should have asked, how
many of you guys hate PowerShell? Not hate,
you don’t like PowerShell, okay. So the question is I will never
write that PowerShell script.>>[LAUGH]
>>Right? We know that. It’s all available in GitHub, right? You can download all the PowerShell,
there’s all the documentation and the support for you to create
your own PowerShell script for your own environment, right? Second one is not a question,
but, I don’t know if you noticed, Windows Server 2016 Datacenter,
that’s important. All the SDN, the networking,
and storage capabilities, I’m going to talk about,
are only available on Datacenter. Now if you have
a virtual environment, you probably already have
the data center licensing because you are running
multiple virtual machines. For those of you that
are running standard, it’s important to know that those
things that we are talking about as well as storage is not
available for standard, right? Question?>>Yeah,
you basically said [INAUDIBLE]. Is that correct?>>Well, if you-
>>In the traditional sense of the word [INAUDIBLE].>>No,
the difference is that network is a virtual network running on
top of any physical network. So if you think about DMZ
in terms of isolation, that’s kind of similar. But their difference is on top
of that same network I can run multiple virtual networks that
are isolated from each other, and then through the gateway you
can connect all of them, right? But the main difference is you can
run multiple networks on top of the same network. And there’s another difference,
virtual networks because they are completely isolated, you can run
the same IP address in this virtual network as well as in the other
virtual network and the other virtual network and they can
communicate with the external world. You can’t do that in
a physical network.>>So would that be considered
[INAUDIBLE] standards? As far as security and [INAUDIBLE].>>Correct, yes. So the question is, will that
meet the DMZ requirements or standards, yes.>>[INAUDIBLE]
>>Yes, yes it will. Let’s just move on to the storage. I still have ten minutes and I still
have some things to show you guys. So let me go ahead and
start storage. I will stay here. I have another session. I will be here the entire day to
answer questions in case you have more questions. Challenges that customers face, of
course you all want to move faster. So the same way we have for
networking you wanna be able to deploy storage for
the application regardless on if I have to add more, regardless
of my physical infrastructure for storage, and cost. Cost is probably the main reason
why people are looking for software defended storage. Because, let’s face it,
storage is very expensive, right? Every time I have to deploy
a fiber channel device, or every time I have to add more ports
in my fiber channel environment, one single port is extremely
expensive, so cost is the main one. And of course, flexibility for
my application as well. So if we think about what we’ve
been doing with Window Server. What we did was, we added to
Windows Server 2012 and 2012 R2, the capability for Window Server
2016 to be the storage, sorry, the 2012 and 2012 are to be
this shared storage device for your Hyper V or virtual machines or
applications Right. The only thing that if you think
about Windows Server 2012 and 2012 R2 was missing, is the ability to
use its own discs from the server. So, in 2012 and 2012 R2,
you can have multiple servers as the back end storage in terms
of connectivity to your servers. But then,
you need a shared storage for those servers to access the disks,
right? So if you think about a very
expensive storage device that you have in your company, you have this
black box that is managed by your storage vendor and there you have
disks and then you have a software that does the duplication,
tiering, volumes and so on. Zoning for
your fiber channel environment or as the connectivity for
the servers and so on. Now, basically what we did was
we opened the black box and we looked at, you know what? We can take the software
out of here and bring it to Window Server,
that’s what we did. But we still needed the box for
the disks, so a is an excellent solution for storage for Windows
Server 2012 and 2012 virtual. In 2016, we thought, you know what? Let’s remove completely the need for a shared storage from
a storage vendor. You can use it if you want but
now with Windows Server 2016, you can use the disks
inside of the server to be represented as the volume for
your application. So what happens here is that
when you create a volume in Storage Spaces Direct,
which is the name of the feature. You use the disks of the servers and you create a volume on top
of all of those disks. So we enable
a Scale-Out File Server. We have a policy for plus or
IOP per virtual machine per service per tenant and you can manage all
of that through VNM or PowerShell. So when we were thinking on how we’re going to
approach this we saw two solutions. And the first one was the converge
solution which is close to what we have today, right? The converge solution is you
have you application servers, which in that case
is a Hyper-V host. So my Hyper-V host is running
my virtual machines and it connect to my storage device. The only difference is that in that
case the storage device are multiple servers with internal disks that
are represented should the Hyper-V host as a single volume, right? As a shared storage. Now, there’s another option for
customers that are adopting new storage devices today
called Hyper Converge. Did you guys ever heard
about Hyper Converged? Yep, it’s one of the trends
in the IT today. Hyper converged is nothing
more than bringing together application and storage, right? What I have here is I have,
in that case if you look closely you have eight nodes,
two clusters. The top cluster is my
Hyper-V cluster and the bottom cluster is
the storage cluster, right? Here I have four nodes. And I have one cluster with
the virtual machines and the storage in the same node. Right? Now what are the pros and
cons of each of the scenarios here? The pros of hyper-converged
is it’s cheaper, it’s easier, but
doesn’t scale very well. Because the performance of the nodes
is impacted by both the performance of the applications, in this case the virtual machines,
as well as the storage, right? So up to 12, even in some cases 16
nodes, you can use hyper-converged if you’re running large or
high performance applications. So you probably need to separate
the virtual machines or the applications from
the storage layer. So then you want to go
to the converged or disaggregated solution, right? The cool thing is Windows Server
2016 data center provides both options out of the box, right? If you compare to vSEND, vSEND
compares to this side of the house. vSEND only delivers
hyper-converged scenarios. And by the way,
please test the performance. We use RDMA and VMware doesn’t,
so I can’t say it, but I would like to, so
please test the performance of Storage Spaces Direct
in Windows Server 2016. So this should be the last slide,
but it’s not. So we support any option that
you may want to use for storage. For instance,
I don’t like Storage Spaces Direct. I don’t like
the hyper-converged idea. I wanna use my regular
fibre channel device with my regular vendor, fine, use it. Well, we support that. I wanna use a NAS,
because my applications doesn’t require performance,
everything’s running well. I wanna use NAS. Go for it. No, I wanna use
the Microsoft options. You can use, we have a third
party storage device or you can use with internal disks. It’s up to you. In terms of partners that
are supporting us in this journey, we do have some partners. I recommend you guys take a look at Windows Server Software-Defined
Program, that you can buy hardware and software together,
along with services. But better than that,
let’s see it in action, right. So this is a demo for
Storage Spaces Direct. What we have here is we have
three servers with each of the servers have four spinning
disks and two, what is it? NVMe, NVMe disks or flash devices. So no shared storage, no fancy
cable, just ethernet, right? What I’ll do is I’ll
open PowerShell. I can do this through
the graphical interface. And if you see the disks that
I have in each of the servers, so as I mentioned, six disks. What I can do here now is I have to form a cluster to be able to,
all the servers work together. So I’ll create a cluster. And that’s a simple PowerShell line. New-Cluster with the name of
the nodes that I wanna create the cluster. There you go. Of course, there’s a warning saying
that you need to see the report. And if I look at the nodes in
this cluster, I have three nodes. Now what I’ll see is I’ll take
the disks from those nodes and I’ll create a pool to be able to
make those disks work together, and create a volume on top of that. So first thing is I’ll enable
ClusterStorageSpacesDirect. And this comment will basically do
all the heavy lifting of configuring the disks and checking the configuration from
the multiple nodes and so on. There we go. So oops,
I think I missed part of the, Yeah, so now if I look at the pool, I should be able to see all the
disks from all the servers, right? So now I have the six disks
from each of the servers. So now that I have a pool,
I can create a volume to support the applications that are going
to run on top of this. To create a volume,
there is another command. I will create five volumes, if I’m
correct here, of 50 gigabytes. There you go. So that’s it, we’re done. We used three servers with six
disks, each of the servers. We created a pool from those
disks inside the cluster, and on top of that pool I can create
as many volumes as I need for my applications, right? Now one of the things
that happens is, what if I bring a new
server to this environment? What do I have to do? Nothing. You bring the server, and Storage
Spaces Direct will detect that the server is part of the cluster. That cluster has, so
let me add the cluster here. I’m sorry,
let me add the node to the cluster. Storage Spaces Direct
will detect that this is a Spaces Direct cluster, and
it will bring the disk of that server to, To the pool. There you go. And the disk is automatically
added to that pool. So now it’s more space and
performance available for my applications to use it. I really need to close Skype. Sorry, guys. Okay, pretty cool, right?>>Do all the disks have
to be identical or similar?>>That’s a good question. Does the disk need to be identical,
or? The answer is no, but we highly recommend that
you do add identical disks. And the reason is you lose
performance if you don’t.>>Does it know how to
classify what’s fast and what’s not fast or you-
>>No, it classifies by itself, so it checks if this is SSD or
NVMe disk, or a spinning disk. And then you have the tiers. You have the performance tier and
the capacity tier, is automatically.>>Just the two? Just two tiers?>>Two tiers, yeah.>>Is there a service that checks
the health of the [INAUDIBLE]?>>Yes, that’s the cluster service. Yeah, I know there
are more questions. I’ll just move along here. I still have one more topic
to cover before I finish, and that’s Storage Replica. So one of the things that we
added to Windows Server 2016 that a lot of customers
asked as well is, look, every time I buy a storage
device from my storage vendor, I also have to buy an additional
device for my other site. And I also have to buy the service
to replicate from site A to site B. Now we just showed you that first
with Storage Spaces Direct, you don’t need fancy
storage any more. And in terms of performance,
it’s great. I really recommend you to take
a look at the Windows Server blog where we published some of the
information in terms of performance. The next step is, if the cluster
can be stretched out to multiple sites and
I don’t need a fancy storage device, can I replicate between
the multiple sites? Yes, you can with Storage Replica. It’s part of Windows Server
2016 Data Center. It’s built into the product. And it works synchronously or
asynchronously, depending on the latency that
you have between the sites. Right, so
one of the things that they were. Are you guys familiar with
the Seattle region or the region where
Microsoft is located? So we have Redmond where
the Microsoft campus is. And we have Bellevue, which is
another city close to there. I don’t know what would
be the equivalent here. But anyway, they are close cities. And one of the things that we
did was we created a site of the cluster in Redmond, and
another site in Bellevue. And the latency meets
their requirement, so we have synchronous replication
between Storage Spaces Direct deployments inside of the cluster
inside of Windows Server 2016. You know what that means? That means that every single
data that I have in this cluster is replicated automatically
between the two sites. If I have a virtual machine running,
I can failover the, I can even live migrate
the virtual machine. But in other cases if I have a SQL
database inside of the cluster, for example, I could failover
the database without losing any data because the site
A is complete gone, right. So this is amazing. Because we ran out of time, I had
a video that I was going to show for storage replica. This video is available online so I highly recommend that
you take a look at it. But guys, a lot of stuff for
you to go home and look more for it. I highly recommend that you take
a look at the Windows Server blog, where we published how SQL Server
can take advantage of this. And what is the performance for
storage in Windows Server 2016, as well as the networking and
the other stuff. I hope you like it, I’ll be here for
some other questions. Thank you.

Leave a Reply

Leave a Reply

Your email address will not be published. Required fields are marked *

Copyright © 2019 Geted Tabs Online. All rights reserved.