Part 2 | Implement A Cost Reduction Strategy for your MathWorks Toolboxes

[0:01] Mae: Hello everyone, welcome to the final session of our two-part webinar series entitled Implement a Cost Reduction Strategy for Your MathWorks Toolboxes. My name is Mae and I am your host for today. Very shortly our presenter Malou Albendia will be joining us. Before we get started let me encourage everyone to ask questions. Send your inquiries to optimize@openit.com and Malou will address them all in the Q&A session of her presentation. If we don’t have the opportunity to answer all of the questions we will reach out via email or through LinkedIn to answer those inquiries.

[0:40] Equipped with 10 years of experience in the software industry, Malou is currently working as a solutions architect at Open iT’s Norway office. She is highly experienced in software development, business intelligence development, data analysis, and software asset management and optimization. She also enjoys traveling and recently visited the wonders of Spain. Malou is always up for the challenge to help companies achieve their software optimization goals in various industry verticals such as engineering, energy, automotive, manufacturing, aerospace, and more. Ladies and gentlemen, we have Malou Albendia.

[1:20] Malou: Thanks Mae, thank you for that introduction and for everyone joining us today, thank you for allotting the time. This is the second part of our webinar series about MathWorks license optimization. Our agenda: first we’re going to have a recap of the different licensing offered by MathWorks and then we’ll dive into the cost reduction strategy and we’ll have Q&A at the end.

[1:53] Okay, let’s start then with the different licensing offered by MathWorks. A reminder to everybody, we talked about this in detail in the first part of this webinar series, so if you haven’t seen that yet I encourage you to watch that, we have the link I believe in the chat box or in the message sent to you. Anyway, individual, it is locked to the user, it can be activated on up to four machines, but you can only use the applications on up to two machines simultaneously, and then you can reassign the licenses up to four times within any 12-month period. Designated computer is locked to the machine, that means you can have multiple users sharing those licenses but of course not at the same time. Similar to the individual you can also reassign the licenses up to four times in any 12-month period. Network named user is also locked to the user, but instead of activating the licenses on individual computers you have to put the license on a license manager instead and it will handle the check-in and check-out of those licenses. And you will have to use the option files to list the users who have access to the licenses. Similar to the others you can also reassign the licenses up to four times within any 12-month period. Concurrent user is shared across different users, but instead of counting the unique users like we did in the network named user and in the individual, here we are counting the simultaneous number of users. It is also using a license manager, so you don’t have to activate the licenses on individual computers, you just have to install the license on the license manager and it will do the check-in and check-out management for you. If the licenses have been maxed out, say you have 10 licenses and you already have 10 users sharing the licenses at that point in time, then a denial event will be issued. So that’s how the concurrent license works. Pro-rated user is an enterprise agreement offered by MathWorks. Essentially it is named user, the difference is that the users are pro-rated based on elapsed time. So a user can be a full user like you have in a network named user or individual, or a fraction of that user depending on how much time they spent using the application accumulated within a 12-month period. Okay so that’s the recap.

[5:05] These are the four areas where we see savings coming from, and today’s webinar will be focused around these four areas. First is right-sizing, making sure that we have a match between the demand and the availability. That we don’t go above the need where we are over-licensed, and we don’t go under-licensed because then we will be losing productivity and we want to avoid that. We want to be on the sweet spot wherein we just have enough licenses for our user community. Continuous monitoring is part of that, automatic license harvesting, looking at idle licenses. We’ll look into this in detail. Just-in-time deferring of purchasing, if we don’t need to purchase right now it would be best to just buy it when we need it. There are many ways where we can do that. We have to maximize the licenses that we have, maybe there are opportunities for flexible license agreements, pay-per-use for instance. Consolidation, analyzing the application lifecycle, analysis of usage patterns, detecting misuse. User involvement is also providing us cost savings, so proactive alerting and we’ll look into that in detail later. Chargeback or showback, and also management of application adoption.

[6:50] Visibility is the foundation of our cost reduction initiatives. Visibility of our entitlements, you know what you’re allowed to use, what is installed, because sometimes we buy licenses but we don’t install them right away, and out of those installed what is used. Regardless of the license type we should be able to see the usage. We should be able to meter the application regardless of where it is installed. You might have a very complicated environment but that’s not an excuse, we should be able to meter that. And I will just say this right off the bat, we have a product called LicenseAnalyzer™ that can provide the level of detail wherein we can see the usage down to toolbox regardless of how the license is installed or how the application is installed. That’s the only tool in the market that can provide you that level of information. So for example in the screen I have here, you know we have these toolboxes with different types of licensing, designated computer, individual, and concurrent, and we’re able to show the usage information across those different license types. And sometimes those toolboxes are only used with one license type, for example Embedded Coder, they only have concurrent users, but sometimes you might have a combination, like Signal Processing Toolbox where there are designated computer and concurrent licensing. And you want to be able to know that information to start our optimization.

[9:01] Now once we have the visibility, we know what we own, we are able to meter the license usage across all our license managers and all the workstations using it, we can start with the low-hanging fruit. It’s a very easy list, list all the apps and all the toolboxes with no usage. This is applicable to any license type that you have whether it’s concurrent or whether it’s a named user, and it should be a very easy exercise. In my example here the license type is network named user and that’s why we look into the unique users, or in my screen I put distinct user. Aside from the fact that we have several toolboxes that were never used, we can also notice that the MATLAB application here at the bottom is only 64% utilized. MATLAB is a requirement for all toolboxes really, this is an indication that there is over-licensing there. We should just stop paying for all these unused toolboxes especially if you have a year of data or so and we haven’t used that.

[10:11] Now how about those applications with some usage, you use them from time to time but not all the time? Well we can consider other license types. We can drill down into the detail and that’s the beauty of having a lot of data, wherein you can drill down into all the users, the dates, down to the hour even. You can really do a lot of analysis. So to illustrate I have a list of users here, and I’ll just mention you can anonymize, or you should be able to anonymize, the user information with the tool that you are using. Right now in Open iT we have that functionality, you can anonymize users. So anyway, we have the list of the users here and the hours they spent daily for the whole month of April. On April 12 we can see that a few users used the applications and seven of them used the application, and this is the busiest, the peak number of unique users in a day. Most of them only use Stateflow for a few times the whole month, like user 933 here who used it only twice the whole month. Now if we count the users, by the way this is not a complete list of users, but if you count them here there are 37 of them. And they can share a few concurrent licenses instead of having their own named user. Say for example, just say that they are concurrently using the licenses, these 37 would only need seven concurrent user licenses shared amongst 37. And this is not the complete list, there are a long number of users that can share these few concurrent licenses. And that will give you a lot of optimization. Even though the concurrent licenses are usually four times more expensive than the individual or the named user ones, if you have a lot of users sharing those licenses it will give you a lot of cost savings.

[12:36] For concurrent users you can run reports like license efficiency. It tells you how many licenses were used within 100%, 99%, 95% of the time. Like here for example, even though the peak usage is 176 out of the 200 available licenses, we only used 176 for 0.8 hours. That is 5 minutes in total, it’s 0% of the total time because we are looking at an eight-month worth of data. That is a very expensive five minutes of usage. Now depending on the SLA that you have internally, or if you can tolerate certain wait time, you might want to reduce your licenses to 155, which had covered 99% of the time in the past 8 months. Or you might be able to reduce that to 143, and we know some customers who have a policy they set internally that they are only going to cover 95% of the time. And that really gives them a lot of cost savings. But of course it’s a balancing act.

[14:03] And again depending on policies that you have, of course you should be able to run impact analysis reports to show you the hours and the days that will be affected should you reduce the licenses. So for example here if we reduce our licenses based on that eight-month worth of data and for example run it on the month of March, when will I be expecting certain wait time? If you reduce it to the 99% which is 155, you’ll see that in yellow here on my screen. You have a few hours here, it actually happens just in about a week, and then a few times in two weeks as well wherein we have wait time. But of course it might not be that much of a wait time, a few minutes a day of wait time might be okay for your company.

[15:07] Now once we do right-sizing, and of course there are other ways to do the right-sizing and I’m not going to go into detail on each of them, but once you do the right-sizing in your portfolio and when you have just enough, then it becomes very important to have a system for continuous monitoring. Because a little misuse, an extended checkout here and then a multiple checkout there, can cause denial and it will cause productivity loss and we would like to avoid that. So what do we do? We should have a system, a tool in place for real-time monitoring. I will suggest LicenseAnalyzer™ by Open iT. It will give you a single pane of glass for all the license managers that you have. You will have the ability to drill down into details at the user level and feature level. It can combine several license managers into one without changing anything on your infrastructure, which can give you the ability to simulate global versus local agreements as well. So there are many benefits to doing that.

[16:13] But really because we have the real-time data we can use that to alert the users. I mentioned things like extended usage longer than expected. So if you have users checking out an application and leaving it running for 48 hours, that’s an extended use. We can alert the users, notify them. Sometimes they just forget that they have licenses checked out. It’s really good that we notify them about this different behavior. Another one is checking out multiple licenses. Maybe they have several machines, they’re working on several projects, again they forgot about it. We should be able to alert them. We must involve our users, user involvement gives a lot of optimization as well. I mentioned this in part one, we have a customer that reached 99% efficiency with one of their applications by involving users. They set policies, they informed their user community about this initiative and why they have to do it. And then they started to send alerts during a particular period of time, and their users were very willing to participate and to really help. That’s why they were able to achieve 99% efficiency.

[17:48] Now of course the reports will be available not only in real time but also as historical reports. You can make dashboards, identify users with applications running more than one instance, and then you can also analyze the user behavior. When are they using these licenses or these applications? You’d see here for example that we have some users using the licenses outside work hours. This is configurable of course depending on where you are and the kind of work environment that you have. So in my system here I have configured work hours to be between 8:00 AM to 7:00 PM, you can change that, and then outside that will be outside the work hours. So here MATLAB we still have a lot of user usage outside work hours. Especially if you have a pay-per-use agreement like pro-rated user that we talked about in the first session, you want to avoid having this usage outside work hours. So we can alert the users regarding this, inform them about this usage outside work hours because that is costing us to spend more money than needed.

[19:20] Another way to involve users is through chargeback. Aside from being able to distribute the cost back to the business units you can use this to drive user behavior as well. At Open iT we have a very flexible metric for chargeback, you can choose from concurrent, distinct user, elapsed time, distinct host, and many more. So for instance you have concurrent licenses, you have many denial events, you notice that many users are using the licenses concurrently, they’re using them at the same time, and you would like to minimize that. What you do is you can implement chargeback based on maximum concurrent user per group as the metric. That means if they have more users checking out multiple licenses at the same time using the applications in this one hour, they will have to pay more. And you will see the change in the user behavior based on that.

[20:41] I will not dive into detail about chargeback because that will be a whole new topic, but let me just mention that the chargeback model that we have is not only flexible when it comes to the metric but also with the overall model. So you can have a pay-per-use agreement, you can have one with a base pay, or you can have just purely pay-per-use or purely base pay. So for example on the screen here we have MATLAB with fixed charges and usage charges. That means we have implemented, at least for MATLAB, a base pay that they have to pay regardless of how much they use, up to a limit of course, and everything all the excess, the overdraft, will be charged according to a unit that we defined in our cost catalog. And also this invoice, this exact report that you are seeing on the screen, comes out of the box. You can subscribe to it, you can set the sending of the report depending on how often you want it. You can send the report monthly, quarterly, depending on how often you do the chargeback to your business units.

[22:08] Okay now having the idea of how much these licenses cost will be even better for decision-making. It will be easier for communication with stakeholders. Here we show how much licenses will cost based on usage on different license types. So for example Aerospace Blockset will cost this much, say $50,000, based on 10 concurrent users. But with network named user it will be more expensive, and it will be cheaper for pro-rated user because you’re not using a lot of elapsed time per user. By the way this is raw based on the usage, so we didn’t implement any filter. We talked about in the beginning, identifying users that can share licenses and move them to concurrent, identifying power users and then giving them named user licenses. So that kind of combination of different license agreements is not yet applied here, this is just the raw out-of-the-box report. You will need to apply the filtering, the analysis, and all those rules to enable you to get the combination. There are certain tasks to be done in order to get to that, or you can either do that yourself or you can ask one of our analysts in order to get that kind of combined view. But definitely we have a system that can give you the cost information out-of-the-box reports like this.

[24:00] Now optimization of course is not a one-time exercise, it must be an ongoing process. And there are levels of optimization as well. Collecting and analyzing active versus inactive usage is kind of level two of optimization. So far we’ve been analyzing runtime, looking at check-in and check-out if you have FlexNet-enabled applications. If you have standalone we’re looking at open and close, you know is the MATLAB application running or are these toolboxes being used. But here we are looking at a kind of deeper metering because we are looking at true active usage. We know that the application is running because it has the licenses checked out, but is it truly actively used? The end goal of course is to have the right-sizing based on the true active usage. The available licenses must be based on the true activity.

[25:22] So say for example we have already performed the exercise we did a while ago for the right-sizing, we can do some of the same exercise but using this data. The question is why didn’t we do it right away? Actually you can, but many of our customers find it easier to do it step by step. So you do the optimization based on the runtime usage and then we implement metering of true active usage and then we do the right-sizing and optimization for that.

[25:51] Now active versus inactive usage is classified or defined based on different indicators. So in the screen here I have listed CPU activity, I/O, mouse, keyboard. It depends based on the type of application. For example you have design, very interactive applications, and it makes sense to use keyboard and mouse as the indicator for activity. But if you have simulation applications running in the background doing computations it doesn’t make sense, and it will always be inactive if we just use keyboard and mouse activity. So that’s why we can use CPU as the indicator, or sometimes it’s best to have these four as indicators for activity.

[26:47] Now once you have this reporting effective you can decide whether we should do the automatic license harvesting. If there is a lot of inactivity, the automatic license harvesting will be based on the true active usage. So if they’re inactive, depending on the configuration, say after 30 minutes of inactivity we should harvest the license. You can also say after an hour, it’s really very configurable. The license harvesting can be a suspend and resume, it can also be terminate, or save-and-terminate. I’m just thinking about the pro-rated user or pay-per-use applications, especially if you have that it makes it even more important that we have a system that can suspend the session or terminate the session when the applications are inactive.

[28:05] So that’s actually the end of my presentation and now I am open for questions.

[28:15] Mae: Thank you so much Malou for your insightful presentation. We have questions here that came from our inbox sent through optimize@openit.com. First, what comes automatically from the system and what do we need help or professional services to handle?

[28:41] Malou: So all the reports that have been presented here today come either out of the box through pre-set reports, the type of report where you can subscribe to it, or through BI dashboards or Excel. But the analysis, the filtering of the data, you’ll have to do it yourself, or you can also avail vendor analysis services from us. The chargeback requires professional services.

[29:21] Mae: Thank you, and our second question: how much is the average return on investment or ROI your customers have seen so far?

[29:30] Malou: On average, 30% savings in the first 3 to 9 months. But it could vary, we’ve definitely seen higher than that, especially if they don’t have any systems in place. But 30% in the first 3 to 9 months is the usual that we’re seeing.

[29:52] Mae: Thank you very much Malou, and to all our attendees for joining our webinar today. So that completes our two-part webinar series. A quick reminder that this webinar is recorded and will be sent to you via email. We will also upload it on our website together with Malou’s presentation. And also we are providing a copy of the presentation which you can download from the handout section of the GoToWebinar control panel. We appreciate your attendance on this series of webinars. You can book a free one-on-one consulting with one of our Business Solutions Consultants to talk about your specific environment. Please visit our website at openit.com to schedule a demo and learn more about our solutions. Also let me grab this opportunity to invite you to listen to our podcast, Talk IT at SAM, for engineering, on Spotify, Amazon Music, Apple Podcasts, and Google Podcasts. And if you have any requests, suggestions, or comments please write to us at podcast@openit.com. We are glad to assist you on your software optimization journey. Always be updated, follow us on social media. Once again this is Mae, your host for today. Thank you and stay safe. Thank you.