馃攳
Optimize Operating Costs with Cloud Economics in Azure - YouTube
Channel: Microsoft Mechanics
[0]
(dramatic beats)
[4]
- Welcome to Azure Essentials,
[5]
in the next few minutes I'll
explain the fundamentals
[7]
of Cloud Economics in Azure.
[9]
Now whether you're
running existing workloads
[10]
or architecting new solutions in Azure,
[12]
I'll share some lesser-known
tips to help you navigate
[15]
the economics of the cloud
for your organization
[17]
so you can optimize your
operating costs in Azure,
[20]
based on your specific workloads.
[22]
So let's start with the key question:
[24]
how does cloud pricing work?
[26]
Well, cloud billing is mainly
tied to compute and storage,
[29]
and this includes the underlying
software licensing fees.
[32]
Costs accrue via a
pay-for-what-you-consume model,
[35]
versus the up-front server infrastructure
[37]
and software licensing costs
[39]
you would typically pay
on-premises in your data center.
[41]
Put another way, it's
the difference between
[43]
operational expenses, OpEx,
[45]
and capital expenses, or CapEx.
[47]
Now, there are a few ways
to consume Cloud services,
[49]
but because many of you
[50]
are used to managing
your own infrastructure,
[52]
today I'll focus on infrastructure
as a service, or IaaS,
[55]
and dig into how
[56]
the cloud consumption
model in Azure applies.
[58]
Now the first thing to note is
[60]
the cloud is infinitely flexible.
[62]
It's not one size fits all,
[64]
and as you literally pay
for what you consume,
[66]
for the best pricing
[67]
you need to consider how
you will consume resources
[69]
for your specific workloads.
[71]
Now the overall premise and
major advantage of the cloud
[74]
that you're probably most familiar with,
[76]
is the power it gives you
[77]
to elastically scale compute resources
[79]
in response to different
peaks in your business.
[82]
This is great for unpredictable workloads,
[84]
where you can add and subtract
resources as you need them,
[87]
resulting in variable cost.
[89]
Think of this as paying per
night to stay in a hotel,
[91]
only when you need to.
[93]
Now, if a portion or all
of what you're consuming
[96]
is more consistent,
[97]
for example, a batch
process that runs every day
[99]
using the same resources on a schedule,
[102]
what we call a predictable workload,
[104]
well, here we have options for that too.
[106]
Here, if you make an upfront commitment
[108]
to consume at a certain level
for a longer period of time,
[111]
you can benefit from fixed
costs at reduced pricing.
[114]
Now in this case,
[114]
it's like paying for an
extended stay in a hotel,
[117]
where you are getting
a better nightly rate,
[118]
because the hotel doesn't
have to worry about
[120]
keeping the room occupied.
[121]
Now striking the right
balance across both models
[123]
is the first step towards saving costs.
[126]
In fact, variable and fixed
cost models for consumption
[129]
aren't an either-or decision,
they complement each other.
[132]
So for your steady, predictable workloads,
[134]
you can establish a
fixed-cost, low water mark
[137]
for the level of compute
that will always be used,
[139]
and reserve at that capacity.
[141]
Now there are a few
options to achieve this,
[143]
with Azure Reservations and Dedicated Host
[146]
that we'll explain more in a moment.
[148]
And you can layer on top of that,
[149]
variable resources for your seasonal
[151]
or demand-based activities,
[153]
where elastic computing makes sense,
[155]
and where you can consume on demand
[157]
or automate against specific
capacity thresholds.
[159]
Now, across both of these
consumption approaches,
[161]
one of the most important levers
[163]
to keep your costs in check,
is software licensing,
[165]
which can make up to 30
to 40% of your cloud bill.
[168]
For example, if you're an
existing Microsoft customer,
[170]
the good news is you
can bring your existing
[172]
and even future Windows
Server and SQL Server licenses
[175]
with you into Azure, including
any negotiated rates,
[177]
using the Azure Hybrid Benefit.
[179]
This brings dramatic savings versus
[181]
the default Service Provider
License Agreement or SPLA,
[184]
which is often rolled into
compute costs by default,
[186]
across all cloud vendors
offering Microsoft software.
[189]
Now one tip here, if you
don't already hold licenses,
[192]
you can save costs by acquiring licenses
[194]
with an enterprise agreement,
[196]
or with a software subscription,
[197]
before provisioning resources in Azure.
[199]
Visiting aka.ms/LicensingEssentials
is going to arm you
[203]
with lots more useful information.
[204]
That said, beyond Microsoft software,
[207]
with Azure Hybrid Benefit,
[208]
you can even bring your
Red Hat Enterprise Linux
[210]
and SUSE licenses with you.
[212]
Now we've covered the three
main areas to focus on
[214]
to manage costs,
[215]
let's drill more deeply
into best practices
[217]
for paying for what you consume.
[219]
Of course, this starts with understanding
[221]
what you really need to run
your workloads in Azure.
[223]
If you're lifting and shifting
an existing infrastructure,
[226]
a key tip here is to not just bring things
[228]
like-for-like into the cloud.
[230]
This often leads to over provisioning.
[232]
And in fact, over provisioning
[233]
is one of the most
common pitfalls to avoid.
[235]
So for example, when
provisioning on-premises,
[238]
you'll likely use virtual
machines with more memory
[240]
or cores than you need
[242]
in anticipation of peak traffic and load.
[244]
But in the cloud, it's best
to start with right-sized VMs,
[247]
and scale as needed.
[248]
Now, let's look in more detail
[249]
at our two cost models and your options,
[251]
starting with variable costs.
[253]
Once you've determined the baseline
[255]
for your fixed cost predictable workloads,
[257]
you can align your variable cost workloads
[259]
directly to fluctuations
in your business demand.
[262]
Purposefully architecting for elasticity
[264]
can help save costs
[265]
and set yourself up for
future peaks in demand,
[267]
all without giving up performance,
[269]
and it's important to choose
the right tool for the job
[272]
and reconfigure as needed.
[273]
For example, take your database backends
[275]
that run in virtual machines.
[277]
Here you can use Constrained
vCPU capable VMs,
[279]
while maintaining the same memory,
[281]
storage, and I/O bandwidth.
[283]
This is like a car with a V8 engine
[285]
that can disable four of its
cylinders at freeway speeds
[287]
to save fuel.
[289]
Now because many database
workloads are not CPU-intensive,
[292]
not only can you save on compute,
[294]
but because licensing is
based on the number of cores,
[296]
you can save on software licensing too.
[299]
Then for storage, Azure offers
multiple storage options
[302]
and a variety of access
tiers for blob storage,
[304]
like Hot, Cool and Archive,
[306]
and here you're paying for the speed
[307]
and frequency of access.
[309]
So, matching storage
types to your requirements
[311]
can help save considerable costs,
[313]
while meeting your required
performance levels.
[315]
Now let's go into more
detail on your options
[317]
for fixed cost predictable workloads,
[319]
where in return for a
longer-term commitment,
[321]
you've got a few options to save costs.
[323]
The first, and lesser-known
option, is Azure Dedicated Host.
[327]
This service provides the
physical servers you need
[330]
dedicated to your Azure subscription,
[332]
which means that capacity
[333]
isn't shared with other organizations.
[335]
You can start with as little as one VM,
[337]
or of course, once you've
right sized the VMs you need,
[340]
you can even bring
[340]
your whole data center
infrastructure to Azure.
[342]
And in fact, some
organizations use this option
[345]
just to get a footing in the cloud.
[347]
And because you can bring over
[348]
your on-premises Windows
Server and SQL Server licenses
[351]
with Software Assurance benefits,
[352]
there's little change
[353]
to your existing processes and procedures.
[355]
It's managed just like
you would be on-premises.
[358]
Now the trick for best efficiency
[360]
is to make sure that
your utilization level
[362]
is at around 90%,
[363]
as you would if you owned the
infrastructure, to save costs.
[366]
Now a stronger, more commonly used option
[368]
is Azure Reservations.
[370]
This allows you to commit resources
[371]
for one-year or three-year
plans for multiple services.
[375]
Now if you're worried about lock-in,
[377]
in Azure that's not an
issue, this model's flexible.
[379]
You can pick the same
premium compute resources
[382]
that you're familiar with,
[383]
mix VM sizes and trade them over time.
[385]
Costs are then evenly spread out
[387]
over the duration of your reservation.
[389]
So let's put these
economic tips to the test.
[391]
A good idea is to leverage the
free Azure Pricing Calculator
[394]
available at
aka.ms/azurepricingcalculator.
[398]
Now, assuming that you
decide to bring over
[400]
your existing software licensing assets
[401]
with Software Assurance,
let's walk through an example.
[404]
Now bear with me in this,
[405]
because it's going to involve some maths.
[407]
Here I've got an E16 VM
with 16 virtual cores
[410]
running Windows server
in the West U.S. region.
[413]
Now if I do nothing at
all, it's going to cost me
[415]
around $1354 per month.
[419]
Now if I take advantage
[420]
of my existing Windows server licenses,
[422]
and apply the Azure Hybrid Benefit,
[424]
that price is going to be
reduced to $817 per month.
[429]
But, to get a more accurate view of costs,
[431]
we need to add back the
cost of Software Assurance
[433]
of your Windows Server
2019 Standard license,
[436]
which at 25% of $972 in open pricing,
[440]
equates to $243 per year
for Software Assurance,
[444]
or about $20 per month.
[446]
Now I add that $20 to my $817 total
[449]
from the Azure Hybrid Benefit earlier,
[452]
and this brings my monthly
total to around $837.
[457]
And that's a saving of $517 per month,
[460]
and you can check these calculations
[461]
at aka.ms/WSOpenLicensing
and aka.ms/SAMechanics.
[468]
Now let's see how this affects SQL Server,
[470]
starting with constrained vCPUs.
[473]
In many cases, your database
might need the memory
[475]
and throughput of a larger VM,
[477]
but not the processing power
[478]
that the additional cores give you.
[480]
So we'll use the same VM with 16 cores,
[482]
this time running SQL Server Enterprise
[484]
and Windows Server Standard.
[486]
And it's $5734 per month.
[489]
Now if we change this to a
constrained vCPU instance,
[492]
of this same VM with four active cores,
[495]
you'll see the price
change to around $2450.
[499]
Now in addition to the constrained vCPUs,
[501]
in the same way we applied
existing Windows Server licenses
[504]
and Software Assurance before,
[505]
we can also do this with SQL Server.
[507]
So let's continue with our same VM,
[509]
now I'm going to add Azure
Hybrid Benefit for SQL only,
[513]
and that gets us back to $1354,
[516]
and then adding the Hybrid
benefit for Windows Server,
[518]
I'm back down to $817.
[521]
So I've effectively brought
over my license assets
[523]
for Windows and SQL, and
I'm only paying for compute.
[527]
But now we'll go one step further,
[528]
and add an Azure Reservation.
[530]
Let's see what happens if I
apply a one-year reservation.
[532]
You'll see the cost goes down
to around $482 per month.
[536]
And with a three-year reservation,
it goes down even further
[539]
to around $310 per month.
[541]
Again, if we add the required
Software Assurance costs
[543]
at 25% for Windows Server
and 25% for SQL Server,
[547]
that adds around $20 per
month for Windows Server,
[549]
as we saw earlier.
[551]
And then using similar calculations
[552]
for SQL Server Enterprise,
[554]
this ends up at around $573 per month
[556]
for Software Assurance on four cores.
[558]
Now you can find all the pricing details
[560]
for SQL Server at aka.ms/SQLOpenLicensing.
[564]
So hopefully you're still following along,
[565]
and that's $593 per month total
[568]
for Software Assurance costs
[569]
across Windows Server and SQL Server
[571]
that we need to add back
[573]
to our last running total of $310,
[576]
which is with a three-year reservation
[577]
and the Azure Hybrid Benefit applied.
[579]
So that brings us to
$903 per month in total.
[582]
So slightly higher than earlier,
[584]
but when you compare
this to the original cost
[587]
of around $5734 per month without
[590]
the three-year reservation
and hybrid benefit applied,
[592]
you can see how bringing
your licenses with you
[595]
is a significant advantage.
[597]
So that was just one example,
[598]
but whatever levers you use,
[599]
again you're not locked into
[600]
any specific resource configuration.
[603]
You can tweak your selections over time,
[604]
and in fact, once you're in Azure,
[606]
the Azure Cost Management tool
[608]
allows you to set target budgets
[609]
and track against your spending.
[611]
And of course, beyond the
costs associated with resources
[614]
and software licenses that
we've gone through today,
[616]
there are some additional
benefits with Azure
[618]
that accrue to business value and savings
[620]
that are worth taking into consideration
[622]
as you build your economic plan.
[624]
For example, if you're trying to reduce
[625]
your carbon footprint,
[626]
studies show that Microsoft cloud services
[628]
are up to 93% more energy efficient
[631]
and up to 98% more carbon efficient
[634]
than data centers run on-premises.
[636]
Moving to Azure also helps you
[637]
to take advantage of built-in security
[639]
and platform resilience
to offset overall costs.
[642]
And as an open platform,
you can bring the skills
[644]
and the tools that you've got today
[645]
to reduce your management overhead
[647]
and focus on building new apps
and services to get ahead.
[650]
So that was a quick overview
of Cloud Economics in Azure.
[653]
Keep checking back to Microsoft Mechanics
[655]
for more tips on Cloud Economics,
[656]
and to learn more, check out
aka.ms/AzureCloudEconomics.
[661]
Thanks for watching.
Most Recent Videos:
You can go back to the homepage right here: Homepage