57. Maintaining Efficient Operations Module Introduction
We encounter this again and again in Cloud+, don’t we? Where there is a technology and there’s an approach that we take in the cloud, and it’s one of the reasons we move to the cloud, but it’s not automatic. What am I talking about? I’m talking about efficiency in our overall operations. Just because you make a move to cloud technology doesn’t mean that you automatically get all this automation, and all this orchestration, and all this great stuff. So, in this module, we’re gonna discuss how we can really be efficient in the cloud. And then in the next module, I’m gonna break down automation and orchestration in more detail for you.
58. Right-sizing and Placement
You’ve seen me do several demonstrations in this course where we really saw how important right sizing workloads was in the cloud. Remember in Google Cloud platform where we were looking at the cost of the various virtual machines? And then we used a pricing calculator for AWS and we saw the projected cost, just like we saw in Google, the projected cost for the equipment. When we were looking at virtual machine workloads that we might deploy from AWS Marketplace, we saw that there was a listing of various virtual machine sizes we could do and the various costs. We haven’t talked really anything about placement though, so we better give that some particular attention in this video.
Now, I do want to give the discussion of right sizing a little more love here before we jump into just a brief discussion on placement. But understand that you can see what’s happening here. This really demonstrates something. Look at this. I have these Splunk systems that I was testing. This is a Splunk search head. This is a Splunk forwarder. And this is a Splunk indexer. So, there’s three Splunk components that I was testing with, and you can see the instance types that these are. Remember what I just said. I was testing with these systems. So, sure enough, these are little t2.microsystem. Look at microinstance types. Look at our little cloud_plus_test over here. It’s a t2.microinstance type. That is very little, you know, hardware resources. This is very little RAM. It is very little virtual memory. It is perfect for testing things.
And why we do this free-tier eligible stuff when we’re testing is, let’s go over here and do a search, AWS Free Tier. When you’re dealing with things like the AWS Free Tier, you have to remember that, sure, it’s 12 months of free services, but look at this, there are portions that are always free. See these designations, and let me blow this up so we can make sure that we can see it with ease. And yeah, look at this. 25 gigs of DynamoDB storage, always free. 1 million requests per month of the AWS serverless compute service, free always. 1 million Amazon SNS publishes, free always. My gosh, I’m gonna look these all over because I’m getting excited right now about some things I’m gonna actually build just to have fun with. CloudWatch, 10 metrics. So what’s this all about? This is intense monitoring of some of your architectures. So, notice we get to set up 10 custom metrics and alarms in Amazon CloudWatch for watching our own solution. One terabyte of data transfers to CloudFront, the caching solution to make your solution globally faster. 50,000 Amazon Cognito. You know what? There’s so many of these always free services that, oh my gosh, look at this, it just goes on and on and on. I’m not gonna have time to read this all. But obviously, this is something I’m pointing out to you, because when you’re thinking about right sizing, stay within the free tier if you can. And how does that make itself evident?
Well, watch this. Notice that these Amazon Linux instances here are free-tier eligible. If I go to spin up something like this macOS Monterey, notice there’s no free tier mention at all. And if you go in and look at some of these costs, look, no free tier, I cannot do the free tier. In fact, the only machines that I can do this on, wow, look at this, all of these are not eligible. So where are the eligible machines? My goodness gracious, what is going on here? I guess what they’re saying is all instance families, no, it’s special ones for this… Well anyways, I think I’m making my point, right? These new instances that are running the macOS, they are very-very specific to only certain supported hardware. Notice many of these Linuxes and even the Windows Server 2019 is free-tier eligible. That’s just the base OS. So, stick to things that are free-tier eligible, and stick to things that are free-tier sized, look at that. And a lot of these resources are always going to be free for you.
Now, what about placement of instances? Well, notice when we go to spin up one of these virtual machines, we have a lot of decisions, and I don’t care what clouds you’re working with. Remember, we’re using AWS here. Just for illustrative purposes; I could have picked any of the clouds to show you this. You get to be very specific about placement. Notice, I get to select the virtual private cloud. I get to select if I want to, the particular subnet. I get to select the IP address settings. I get to affect the host name. Now notice we can add this to a placement group now in AWS. So, if you wanted to constrain your virtual machines into the same actual servers up in the AWS cloud and reduce the latency between those machines, this is something that we can do now. You have to understand though whenever you start going in and saying, ‘I want everything to be close together,’ and look at this, ‘I want to run on my own dedicated hardware instance in the cloud,’ when you start making choices like this where ‘I don’t wanna share my stuff in the cloud’ and I want all my stuff close together,’ you are most definitely gonna go outside the free tier. So, these types of placement decisions where you get to dictate exactly where stuff goes in the cloud, you are responsible for understanding that costs are going to increase. Something else that I’ve run into before, and I’ve done videos about this on YouTube, it’s really hilarious, is even when testing some solutions, I will run some AWS regions out of resources. I’ve done this myself. Yep. So I’ll go to spin up my solution and AWS will turn to me and say, ‘Oh, you’re gonna have to try again a little later. And we’re so sorry, but the region that you are in doesn’t have any virtual machines left of that particular size.’ So, that’s just remarkable, isn’t it? That you know, here’s AWS running out of machines of a certain size that little old me needed just in doing some testing. So, it really goes to show you that you have to be careful about forcing placement of resources, because sometimes the cloud provider will just turn right around and say, ‘Sorry, I can’t do what specifically you’re asking me to do as far as placing these resources.’ Well, thank you so much for watching.
59. Device Drivers and Firmware
Well, think about it. This is a little weird. What about things like device drivers and firmware? I mean, my goodness, we are in this heavily virtualized environment, and now, we’re talking about hardware that, of course, might need some handy, dandy firmware updates. The hardware is virtualized, but it still needs the firmware updates, and how about device drivers for all these things that are virtualized? Well, it’s an interesting discussion. Let’s talk about it.
So the big thing that you’re gonna want to really research as part of your design is you’re gonna wanna research what kind of virtual device drivers do exist for your various equipment. You see, oftentimes, Cisco and HP and all these now will offer hardware that you can put in your cloud, in your private cloud, let’s say, and that hardware is filled with things like virtual NIC-capable interfaces. They’re amazing. One physical NIC can represent dozens of virtual network interface cards, so the hardware is specialized to be virtualized, and the virtual device drivers needed in everything are baked right in, but if you’re building your own hardware to put in your data center’s cloud, and you’re gonna be piecing it together, you really need to be aware of virtual device drivers.
Now, let’s talk about firmware. This gets really fun and interesting. Think about it: when, let’s take Cisco as an example. When we put their cloud-based hardware in our cloud, it can reach out via the internet, and it can get the firmware updates just like it would if it were physical hardware. Cisco, the updating service doesn’t care. It just sees that it’s calling up to get its firmware updates, and it’s delivered down into the virtualized environment. So, it’s just amazing. Whether it’s physical, or whether it’s virtual, it doesn’t matter. It’s gonna be reaching out to the internet nowadays to get its firmware update, and that firmware update will be delivered to it whether it’s hardware or it’s virtualized hardware, and that firmware update will be implemented.
Now, we need to keep this in mind any time we’re talking about the cloud, and that is for things like this, what is going to be the bandwidth requirements? You know, we need to make sure the bandwidth requirements are always there. Now, the great news is with firmware updates, those are typically ridiculously small, but I’m just kinda thinking about this. Expand this discussion out for a moment, if you would, in your mind. Just think about this as far as other things we’re gonna need to get in and out to our virtual machines from the internet. Sometimes that can make for some large-large files. Think about if we’re moving iOS images for the actual Cisco devices themselves, or here’s a great one: what if we’re taking Windows 11 images, and we’re moving those up and down? Oftentimes, those are gonna be like 50 gigs in size, and maybe we need to move a whole bunch of them, so these are the types of things you have to be careful with when it comes to my gosh, what’s the bandwidth between us and the cloud?
Now, something else that I wanted to impress upon you with this discussion is firmware updates and all that kinda stuff, I don’t know about you, but it stinks, right. It drives me crazy, checking in on that stuff and making sure that firmware updates didn’t break stuff, and that type of things, and that’s another time where public cloud comes in hugely handy. Think about public cloud for a minute. Think about all of the stuff in public cloud that someone like Amazon is… How about that? That’s how I write Amazon on our whiteboard. That’s crazy, and I just wrote on our wood. Yikes, I gotta stop doing that. All right, anyways, Amazon, okay, think about everything they’re doing for us we don’t have to worry about. Think about that. Think about all of them running around, worrying about the firmware updates, right? We take our solution. We put it on their equipment, and we don’t have to worry about firmware updates. They’re the ones that have to worry about all the firmware updates, so once again, we’ve seen another compelling reason why we should at least go hybrid cloud, so some services in the public cloud, and we’ll let them worry about all the work. Thanks so much for watching.
