Time for my all-time favorite topic,
hyperconvergence! Let me explain why I think hyperconvergence is
THE solution for tomorrow's private clouds. Last time I ended my video talking about
private clouds and now I want to concentrate on one specific architecture
or technology how you can create your own efficient modern private cloud. By the way, if you haven't seen
the previous video of mine I will leave a link in the description field
and I highly suggest that you take a look at that video before
and then come back to this video. Okay, so...let me just once
again take a step back and draw the traditional basic data center how we
have been architecting and constructing data centers already for four decades. And as we remember from the previous video
we have three tiers, if you will. The first one would be networking. Then we have servers. And a third one would be storage. We have to have these three tiers in any kind of
IT environment that we are dealing with. So...we have storage devices here. A couple of them of course for redundancy. We need to have a couple of SAN switches to
connect our storage devices to servers. We probably have some hypervisor
software there to cluster the servers. And on top of everything we have
networking devices to connect our servers to outside world. This is something that we call A 3-tier siloed architecture. 3-tier obviously because we have three tiers:
networking, servers and storage. and then "siloed" because all of these tiers
are quite independent of each other. When you want to operate something,
when you want to manage something, it's always done at that tier level. If you want to do it in storage,
you only upgrade and manage storage. If you want to do something in networking
you do something in networking. And the downside of it is
that if you are doing an upgrade in storage environment although it's
managed and maintained independent of every other tier you still have to
communicate with the server guys for them to know that there's a new storage
device and all the zoning stuff and all those things are are in order. So, that makes it a bit complicated
to manage. And to make it a bit more complicated let's
add a couple of things into the picture. You should never forget the importance
of a backup device. So there's always some kind of backup device
connected to the storage area network. Also if and when we are talking about the
real-world implementations no real enterprise is ever trusting
only one data center. Last time we were talking about redundancy
on a component/hardware level But we also want to be sure that if something
happens to this one data center our data is not lost and our
business continues normally at least as close to normal as possible. So, that's why we want to have
more than one data centers. Let's add our "Site A" here.
So, this would be our primary data center. This is our "Site A". So, that would be our secondary data
center. Just as a backup in case something
happens in primary site. Now, we have to do some kind of
replication between the sites here. Some kind of bi-directional replication in order for us to be sure that all the
same data is available in the Site A as in Site B. This can be quite tremendous task and there are various different
technologies how to achieve this. We can use for example something we call ...WAN optimization which is basically just compressing
the data coming from site a going to the site B. Looking at this picture
already now it's...complicated. We have lots and lots of different
components in Site A, Site B... replication happening here and there,
siloed infrastructure and everything like that. But the complexity doesn't stop there. It's not complex only from infrastructure
and hardware point of view. We have also many different kinds of user
interfaces that we have to use to manage
the whole environment. Hypervisor environment is probably
managed with one type of interface. We have another user interface
for managing servers. Probably we have a few of them there. Storage area network is managed
with different user interface. We have a fifth user interface to
manage storage devices. Backups are managed with
different user interface most probably. And then we still need one user interface
to manage the whole WAN optimisation. So, what I want to do now is
to get rid of as many hardware components from this
picture as possible without sacrificing any of the benefits
that I have in this environment. The second thing I want to do is to
get rid of most of these user interfaces. So what we can do is we collapse
these two tiers completely. So we get rid of independent individual servers, we don't have to worry about storage devices,
SAN switches and backup devices. Instead, we have just one x86
server. "x86" meaning just an industry standard server
running Xeon's and so on and so on. what we have there inside is
a powerful CPU (processor) that's running all our virtual machines,
all the workloads there. Then what we have is
lots of storage capacity. It has lots of disks installed in it. Of course in enterprise environments we are not
trusting any individual component so, we have to have two of these devices here. Then what we do is we connect them
together -- and that's it. Here we have exactly the two tiers:
server and storage tiers inside these server appliances,
as we call them. They are just hidden inside the appliances
and they are not separate devices. So, this would be our "Site A". We need to do still the same for "Site B"
here, we still have the backup site. And the best hyperconvergence solutions
out there nowadays are directly communicating with sites without the need of any special WAN optimization or anything. They use sophisticated file systems to make sure that all the data
is compressed and optimized so that they can be sent over vast distances. This was our Site A and this is now our site B. So, now we have already vastly
simplified the left-hand side picture infrastructure to the right-hand side picture. The second goal I had was to make the
management as easy as possible. All of these boxes, these special server
appliances, are belonging in the same
"federation" as we call it. And one federation is managed with
one single user interface. So, all we need is one user interface
to manage everything. THIS is the beauty of hyperconvergence. We are getting rid of the
infrastructure complexity and we are introducing new simple way
of constructing the data center managed with one single user interface. Now, for obvious reasons...hyperconvergence
is the fastest growing individual technology within the data center industry at the moment. Everybody wants to be in the industry everybody wants to be inventing this and that to the hyperconvergence idea. Because people love simplicity,
people love efficiency and that's what hyper-converged
infrastructure brings to the table. I would like to give you a couple of tips how you can choose the best hyperconvergence
solutions for your needs when you are going through the
options and selection and actually there are only two The first one I want to emphasize is: "As hyper-converged as possible" Now, what do I mean by that? So, if you want to call yourself
a hyperconverged solution the minimum, the bare minimum requirement
is to collapse these two tiers. You have to get rid of separate server tier
and separate storage tier. That's the bare minimum. However, the best hyperconverged
solutions don't stop there but they go the extra mile or two to converge as much infrastructure as possible into
these tiny little appliances So, what we can do is to converge
obviously server and storage but also for example backup
and WAN optimization. So, keep in mind that there are different kinds
of hyperconverged solutions out there Some do the bare minimum, some of them go the extra mile, and some of them go two extra miles. Most typically hyperconverged solutions
can consolidate the infrastructure with a ratio of, let's say, 4:1 or 5:1. So let's say you had five racks of
traditional infrastructure before and when you move to hyperconvergence
you manage only with one rack. That's pretty cool. But some of the best solutions can reach as high as 10:1 hardware
consolidation ratios. The other aspect to this is the data reduction. The most sophisticated hyperconverged
solutions out there they can reach up to 50:1 data reduction. What this means is that if you
have 50 terabytes of data only 1 terabyte is actually written on the disks. So, that's number one. Number two... Make sure that your solution is
a true appliance. What I mean by this - true appliance... is that if we take a look what this hyperconvergence node, this appliance, actually is... So, HCI = x86 hardware + HCI software. HyperConverged Infrastructure software. So we have a common off-the-shelf
industry standard server with special software. Now, once again there are many many different
approaches to hyperconvergence on the market at the moment. There are companies who are only concentrating
on this part of the equation So, they develop their software then they install it on somebody
else's hardware. Quoting one of the true living legends
of the IT world, Alan Kay, who is by the way the grandfather of modern
user interfaces as we know them. He famously said that: "People who are really serious about software,
should make their own hardware" And...there's a lot of wisdom in this. I'm not saying that you cannot get
anything good out of just creating your own software and
using somebody else's hardware. Of course you can. There are plenty of examples on
the market of brilliant solutions that are using software from one company and
then hardware from another company. However, those companies who
are making their own software running on their own hardware are very highly successful because
they can optimize everything. They can optimize their software to run
on their hardware and they can optimize their hardware to run their software. From end customer point of view it's day and night difference in having to have
support from one company what comes to software and
support from another company if there's a hardware failure. We want our hyperconvergence solution
to be a true appliance. We want software and hardware to work in perfect harmony, in sync with each other. And when there's a problem
we call one number. All right so that was my quick intro
to hyperconvergence. If you liked this, hit like. If you want
more of these videos, hit subsribe. Until the next one!
