Azure SQL Managed Instance just got WAY faster - Meet Next gen GP | Data Exposed

Key Takeaways

Azure SQL manage instance just got way faster from a memory I apps configurability and more. Learn all about it this week on data exposed. >> Hi, I'm Anna Hoffman and welcome back to Data Exposed. Today I'm joined by Arush from the Azure SQL team. Thanks so much for coming back on Data Exposed. >> Thanks Anna. Happy to be here today. Awesome to have you on the show and today we're going to be talking about how Azure SQL manage instance got way faster and I'm excited to kind of dive into some of those details. Uh but before we get into it, can you share with us a little bit about like these different tiers or what people need to know and then what we've done to make it so much faster? >> U yeah definitely. So today in SQL manage instance uh customers can choose between two different service tier business critical and general purpose and the main delta between those tiers is that business critical is based on the local storage architecture while general purpose tier is based on the remote storage architecture. So with nextG GP release what we actually did is we brought the improved version of general purpose tier by doing the storage replatforming. So what what sometimes confuses customers when we see using like the new tier uh term in this case is actually not the case. This is just the upgrade of the existing g purpose tier which comes for free for the customers. >> Awesome. Okay. So we love to hear that it's not like really a new thing or costs any more money. It's just better and they just get it. Awesome. So tell me more like what's what what's in it that makes it so much different or better or what sort of performance uh impacts does that have? >> Uh thanks. So uh yeah so when we talk about improvements that are happening uh with this release uh we actually bucketize them into three different pillars let's say that way. So first pillar is the performance and higher baseline performance uh that comes with this uh replplatforming which means for the same price you're getting the upgraded uh skill. >> Nice. >> Uh yeah that's that's that's the say the fundamental thing and then on top of that we have two different things. One is scalability with higher limits related to the storage with density and vores and then flexibility uh within those with those options with uh more options for customers to control their hardware configuration of their instances. >> Okay, awesome. So what types of scenarios does this kind of enable for customers? >> Mhm. So uh we can start with with the performance. So let's imagine that you have like already your general purpose uh instance uh and uh you're running your workload and like all of a sudden you get new option where you can do where you can aggregate. So all what you have to do is with a single button click switch from the classic to the next gen general purpose. Uh in the background the operation will switch you to the new architecture and this is nothing new because the experience is the same like for example where when you today scale your request for example. You had some performance issues. You tried uh you tried to solve it with weaker scaling. You got that experience of updating the underlying infrastructure. Well, here the experience is the same. Just we are not touching your hardware configuration. We are bringing you to the upgraded platform and you get better performance out of the box. Uh so lower latency, higher amount of IOPS, higher throughput, all of that at your uh at your uh disposal. And to to bring this closer to to to users, we our team did some uh weightlifting uh and did some perform uh performance comparisons. So what we used is like the industry standard hammer DB test connected to the let's say the classic the old version of the general purpose and the nextG general purpose and put uh put that side by side we see that next gen immediately brings higher number of transactions. So with a like one click and one operation you end up with more performance for for the same price and to to see like uh slightly under the hood uh and compare some additional metrics we did the setup of database watcher. Uh so we monitored the transaction log or the user databases and uh what we can see is that the latency immediately was cut to uh for half like from 4 milliseconds to 2 milliseconds. On the right side we see the NG instance. Uh then the IOPS went three times up from 73 to 250. Uh and the same applies for throughput from 3.5 to almost 11 megabytes per per second. >> Wow. Wow. Those are like not small performance gains. Those are like [laughter] pretty substantial. Uh yeah like the expectation is that you immediately like the it's like these things are workload dependent but definitely uh you must feel the improvement when you switch to to next gen. >> Cool. Um one question for you. Uh in the previous slide you talked about how it's kind of a uh just a toggle on enabled disabled if you want the nextG if you're currently using SQL MII. Is it like is that upgrade process you were saying similar to like a a slow upgrade? So is it a size of data operation related or is it faster? Is it slower? Like what should people expect when they click that enabled button? >> Yeah, it's is it is the size of the data operation because as I mentioned in the beginning we did the storage platform. So we switched from one type of the storage to another. So we need to do this data copy but it is oneoff operation one when you're on nextG then all these operations are much more elastic and faster. >> Got it. Makes sense. And it's just like all the other skew or slow upgrades where you know we keep you online until the moment we over. >> Exactly. Exactly. >> Amazing. Awesome. And then for new databases is this on by default or do you still have to go through the enabling process? uh they uh uh at the moment they need to manually opt in but the experience is the same like uh you have this toggle you just say hey I want with with to go with nextg GP and and that's it >> awesome cool thanks for answering those questions I'm sure there are other things uh people want to know what else [laughter] >> um yeah as they mentioned we have three buckets this was the first one about baseline performance improvements and better better limits uh the second thing is scalability like now when you're more performant You might have more data. You might have more databases. And all these limits have been uh doubled like the maximum amount of storage has been increased from 16 terabytes for to 32 terabytes. And you can host up to 500 databases now on your instance which is like 5x uh improvements. So for example, you can now consolidate multiple instances into one uh and do additional cost saving of for your uh environment. Same applies for number of files and on the this these things are like the scalability of the storage and the storage rep that enable us to achieve this but we also have increased the number of vehicles max amount of vehicles that you can go so if you need more compute resources you now have more options at uh at your disposal >> ah that's great >> and I also mentioned that it's not only about the scalability but also about the flexibility that we have within these limits Now, so uh speaking about vor previously we have like only eight options at your disposal. So if you know you need more compute power you had to jump from four to eight or to 16 which means not only more uh more vehicles that you need to pay for but more licenses to cover those vehicles. Now with next gen we have this concept of more quick stops so you can for example go to six week course no need to jump to to to eight. Uh but probably the biggest the biggest change that is coming with this release is that you have two additional sliders. We call them sliders because in portal it's like drag and drop slider capability uh where you can independently scale your memory independently of vicors and you can also further tune your IOPS. So you get some amount of built-in IOPS but you can add more and add more power to your to your instance. >> Awesome. I feel like the last two sliders those are things that if we were in an audience people would start clapping. Yeah. [laughter] Uh yeah, we we had this experience uh for a couple of conferences where you presented this but yeah definitely um I would say the true past experience for controlling your your resources cloud. >> Yeah. Amazing. Awesome. What else should people know as they uh they move towards nextG GP? >> Mhm. Uh yeah. So these sliders are not immediately visible. So you need to switch to nextG. This is exclusively nextg uh nextG capability. So once you enable uh this for your instance first you will observe that v core and memory now have higher limits the thing that that we spoke about in scalability concept. So you can go up to 128 and all the way up to 32 terabytes and then you will see also see two new sliders. So one for memory and uh one uh for IOPS. So in previous example maybe you did not scale your vehicles uh because you need more CPUs for your workloads but because you uh required more memory now this is not the case you don't have to increase your vehicles you can just add more uh more memory and again this is also costsaving opportunity because when you add more vehicles you need more to pay for more licenses on top of that those vehicles now you just add more memory and that's it >> so it's indep completely independent like I can change the memory I can change IOPS without touching the vores. >> Correct. Correct. >> Amazing. Awesome. >> Uh those those values are still kept at some point uh and depend on the vehicles but within one vehicle selection you have the you have the flexibility. >> Awesome. >> Um and yeah I mentioned like during our conversation I mentioned this like you can do cost saving for that you can do cost co cost saving for that uh with with these options. So the the actual value for the users is actually the improved PCO because now for less money you can get more performance and actually improved the total cost of ownership of your project or the service whatever is the scope that you do these measurements for. >> Amazing. Awesome. It seems like it's honestly it's a no-brainer. Like if you're using manage instance today, you should change or upgrade to nextg GP to get better baseline performance and all these other options. If you're not using manage instance today, you should maybe re-evaluate that decision because it just got a whole lot more configurable and cost-effective. But >> uh yep. Yeah, definitely. As you mentioned, it's like >> Oh, sorry. From your side, what like call to action or tips and tricks do you have for folks who are maybe considering this or getting started? >> Um yeah, uh two things that I would call out. uh definitely we we still see lack uh lack of awareness for the free SQL manage instance. So you can try free SQL manage instance for free uh for 12 months. So full year uh with 70 720 vehicles per month which is like equal to the full working week. So uh with when we introduced an aent GP we wanted to immediately enable it for the customers to try it out. So it's part of the free offering as well. You also have the toggle. You opt in, choose between plastic and nextG. But definitely the recommendation try out the nextG uh immediately. Uh so if you're starting the project for scratch or uh doing like the certification of the SQL MI for your project, this is a good starting point starting point where you can again achieve these things for free. In case that you already have some existing environment or uh or or or server that you want or consider migrating to SQL MII then definitely uh SQL server migration as arc that was recently G8 is is a way to go. So you're guided from the entire uh process from the discovery uh to the to the migration itself and nextgen GP is a valid destination in this in the in this post. So you can migrate directly to to next genp. >> Awesome. Great. I learned so much during this episode and quite frankly I'm pretty excited about these performance improvements. I think you're right. Azure SQL manage just got way faster. So to our viewers, if you like this episode, go ahead, give it a like, leave us a comment and let us know what the most exciting part about this announcement is to you. We'll put some of those links in the description for you to learn more. And we hope to see you next time on Data Exposed.