Public cloud, private cloud, hybrid cloud, multi cloud, infrastructure as a service, cloud compute, hyper-scale, mumbo-jumbo.,.. Which one of these is right for you? And how do we match you to the right environment for your business?
A massively common discussion on the internet is cloud and cloud costs, and what you can do to lower your cloud costs. There’s plenty of papers and white papers and one-person-talking-ads and some people on Twitter, and God knows what else going on right now talking about your cloud costs. So let’s talk about it!
The first thing is: YES, you are spending too much money on your cloud. I don’t know you but I will tell you that you’re probably spending 30 to 40% more on your cloud than you should be. I don’t care how big you are, we have customers ranging on the public clouds from $10,000 a month to $2,000,000 a month and plus, and the universal fact of life across every single one of them is: there is a 30 to 40% number within that they can reduce.
That’s not appropriate for most companies and so we don’t walk into these conversations saying “hey look you know you should do all these different things because we know you’re going to say 30 or 40%,”right.
The number one recommendation that comes out is going to be user reserved instances. And a reserved instance may or may not be appropriate for business.
I mean do you want to lock yourself into a three year contract for a specific compute type when you may want to get rid of that compute type in the next 12 or 18 months?
So there’s a lot of business decisions that go into cloud and cloud spend more so than just: are you spending too much and can you spend less?
So proper instance sizing you know is the other one of “oh you know you’re on a fill in the blank word acronym instance type and you should be on this instance type” or “you’re on classic and you should go to VPC, or you should do this you should go do that…”
And all of these things have real costs to you: also in terms of do you even have the technical staff that even knows how this application is running well enough to be able to migrate that application to another thing? and the chances are probably not… or if you do they’re so busy doing other things there’s no way that you can distract them for months in order to make this migration happen.
So coming after you and saying “hey look you should do all these other things…”
Or the other “favorite one” is: “you should automatically turn off your instances at night every night.”
Now, that sounds great on the surface EXCEPT now you have to make sure you’re properly tagging and identifying the instances when you create them or you’re going to shut things down that you don’t need.
There’s a lot of low hanging fruit that you should be looking at with clouds. We’ll just use Amazon as an example because it’s the most popular now: do you have unattached EBS volumes? I guarantee you that if you have any size cloud and you have EC2 instances starting and stopping, you have a ton of EBS volumes that are just sitting there unattached that are costing you money every month that you just don’t even know exist.
Dynamo DB TTL: we’ve seen lots of some environments where they’re spending a quarter of a million dollars a month on DynamoDB and just by enabling TTLs they can lower that cost in HALF!
Enabling the TTL has a huge cost component too because of how Amazon actually writes to Dynamo, and these become really specific conversations and something that we can help you with, that are actually specific to you. but are you spending too much money in your cloud? Absolutely you’re spending too much money on your cloud!
Next version of this conversation gets into this great chat on multi cloud. so everybody in the market is talking multi cloud non-stop.
It’s all BS! multi cloud doesn’t exist!
Why does multi cloud not really exist? Because of data. Somewhere you have your data. your data is in a data lake, it’s in a data warehouse, it’s in a database it in a pocket, it’s somewhere. You data is somewhere.
And there is real costs with taking that data out of a region within one cloud provider (let alone taking the data out of 1 cloud provider and shifting it to another cloud provider). So this whole concept around multi cloud is really really really specific to how your data actually sits, where it and what is used.
What’s really common? what’s really common is we see a lot of companies that are using Amazon that are moving over to GCP because they want to use big query big table machine learning pipelines available on Google, and then start evaluating whether or not they want to bring the computer over to Google as well. And there’s lots of cases where wholesale migration from Amazon to GCP or Amazon to AWS makes a lot of sense.
but to do that effectively you need a partner that can help you through this process. You also want a partner involved because the partner can bring along market development funds to help pay for that migration for you between one cloud vendor another cloud vendor. So that’s a big secret there: don’t try to do this yourself. Get a partner involved, a partner will actually help pay for that migration for you between your different clouds.
Multi cloud is also a marketing terminology and it’s a thing that “kind of exists” but if you’re spending $1,000,000 a month on Amazon maybe you’re spending $100,000 a month on GCP. So you’re multi cloud, but are you really multi cloud? Probably not.
what is real? What’s real is what now is called hybrid cloud. So now this whole thing of hybrid cloud just means you have a public cloud provider and you still have the physical data center somewhere and this is where the math gets really exciting. Because the economics on a physical data center is massively in your advantage. we’re talking 80% reduction numbers off of a public cloud spend. And that sounds crazy to say out loud, and you’re probably listening to this and saying “Max you’re completely just making numbers up.” But I kid you not: we’re seen reductions of upwards of 80% a month on moving an application out of public cloud into an actual data center environment.
Why is this? Well first off, the cost of the actual compute and your server equipment and your compute infrastructure is so much lower when you’re actually purchasing it or leasing it yourself. You’re going to see massive savings ’cause you also virtualizing your stack: you’re running containers, you’re running doc, you’re running Kubernetes, you’re doing something along those lines, you get huge efficiency out of your physical equipment.
The other thing that people don’t really talk about is your data egress cost.
Data egress is an absolute killer in clouds! They price based on dollars per gigabyte. And we’ve had customers get excited, like, “Oh my goodness we got this PPA from this cloud vendor cause we’re putting out like 20 pentabytes a month of egress costs, and we got our costs down to 2 cents per gig, with this private vendor, and ‘high-fives’ and we’re so excited…”
And then you come back and you say “well now you’re paying about $8.00 per megabit on 95th percentile for that egress traffic, and list rate at your volume from an actual network provider is like $0.15 per megabit. So you’re paying $8 in the cloud vendor, and you can be paying $0.15.”
So if you have a heavy bandwidth egress application, if you’re a heavy CDN usage, the data transfer cost will absolutely cream you in terms of what you’re paying that cloud vendor.
So now we’re seeing hybrid cloud as a real thing and we’re seeing a lot of companies that were originally resistant to this idea are now my adopting this. And once they see the economics of this, they become big fans of hybrid cloud.
We get it: You don’t have data center expertise, you don’t have people in the ground in you know the Ashland VA area, or you know, Singapur… and guess what? you don’t need it.
There are plenty of service providers that specialize in logistics related to purchasing server equipment, installing the basic operating system on it, and putting it into a data center for you.
So you can still leverage almost everything you get out of your public cloud in terms of this provisioning cycle and physical management, without actually having to do it yourself!
You still get the semblance of elasticity if you negotiate your contract properly, in terms of being able to turn things down in one place and bringing them up in another place, and guess what? Your costs are so much better!
What we see now that’s become very common is: customers figuring out what their actual cloud spend is, let’s say it’s a million bucks, and looking at that from the standpoint of reducing that $1,000,000 spend, to maybe $300,000 or $400,000 a month from that cloud provider. They can still have elasticity, they can still have peak capacity, they can still deal with all the unknowns, but it’s not putting $700,000 a month into a data center. It turns out they have $300,000 or $400,000 running with the public cloud and maybe $300,000 running with the data center all in.
so you look at that cost structure and you’re gonna say, “no they, didn’t reduce 80%,” Yes in this case they were just 40% in that a evaluation, but they have a lot of elasticity left, they still have a public cloud environment, they a lot of flexibility going forward.
When you’re in a conversation with somebody and they wanna talk to you about saving money in AWS, and your eyes roll back in your head and you completely tune them out and you glaze over, I do not blame you because public cloud cost management is like the dark arts of IT spend management control.
It really requires a lot of specific things related to what you’re doing in that public cloud.
Same thing with multi cloud: everybody wants to talk about their multi cloud strategy and some are really good but it really also is so specific to your business and how your application is actually running, and does your application even support this idea in the first place? or should you skip the whole multi cloud nonsense and just go back to a hybrid environment where you have a physical data center and you have physical data center that exists to run your consistent workload that you know and love, and then have your public cloud environment to deal with that elastic workload that you know is coming down the pipeline as well. This is the best of both worlds, and this is what you should probably be, and this is a place that we take a lot of our customers.
At ITBroker.com we will help you understand this, we will help you understand what’s actually going on with your workload, what makes your application actually tick, what you can actually pull off, what is just hype nonsense marketing mumbo jumbo garbage and what you can actually achieve, and how do you move from a API driven web interface public cloud environment to a data center environment, and how it actually works for you.