Parte 1 | Caixas de ferramentas MathWorks: Como otimizar em qualquer modelo de licenciamento

[0:00] Mae: Hello everyone and welcome to the first session of our webinar series entitled MathWorks toolboxes, how to optimize in any licensing model. 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, just drop your questions and Malou will answer them in the Q&A session of her presentation. If we don’t have the opportunity to answer all the questions, rest assured that we will send them to Malou and we will reach out via email or through LinkedIn to answer those inquiries.

[0:39] 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.

[1:17] Malou: Thank you Mae for that introduction and thank you everyone for joining us today and for allotting your time with us. This is the first installment of our webinar series about MathWorks license optimization and our agenda is as follows. First, introduction to MathWorks licensing, and then we’ll talk about the optimization strategies for license type and also some customer examples.

[1:51] So let’s start. In MathWorks they offer these license types to their commercial customers: individual, designated computer, network named user, concurrent user, and pro-rated user. Individual license is locked to a specific user and must not be shared with others. So if you have seven users like I have here, you will need seven individual licenses. The users can activate the license on up to four computers but they may not use a product on more than two computers simultaneously. So when we do license usage reporting for this type of license we will need to do a count of the unique number of users.

[2:41] The next type of license is called designated computer. Now designated computer is locked to the machine, right. So unlike the individual license from a while ago which is locked to the user, this one is locked to the computer. That means in my seven users here they can share the license installed on a single computer but not at the same time. You may redesignate the license to a different machine whether temporarily or permanently, up to four times in any 12-month period. And when it comes to reporting what we’ll need to do is count the unique number of machines, right, because this is designated computer. In my case here even though I have seven users using my MathWorks licenses, I only have four designated computer licenses, so that’s what we are going to be counting when it comes to reporting on this type of license.

[3:48] The next one is called network named user. This is like the individual license in the sense that we are counting the unique number of users as well. The difference is that here we are using a license manager to manage the licenses, hence the network in the name. The license manager has the control of allowing the users to check out a license and then manage it when it is checked back in again. You define who should have access through the use of license or option files. In our case I will need to enumerate in my option files these seven users that I have if I want them to access the licenses using the license manager. Also, the license manager removes the need for activation. A while ago we talked about the individual license, you can activate it on up to four computers. Here you don’t have to do the activation. We can install the application on as many computers or machines that our users are using, but we just have to point the application to the license manager. A user is also allowed to use the product on up to two machines simultaneously. So if you have a user that needs to use the application on a third machine they need to close the application on the second machine they’re already using, they must check in the license before they can check out another one, else they will be denied. You can also redesignate the license to different users, you can modify it in option files whether temporarily or not, up to four times in any 12-month period.

[5:31] The next type is called concurrent licensing. It’s also network licensing, it is utilizing the license manager. But instead of counting the unique users we’ll need to count the number of users that are using the licenses simultaneously. So in this case we will need fewer licenses than the number of users. My seven users here for example can share only one license if they never use it at the same time. If the available licenses have been maxed out, the next person that tries to check out the license will get a denial event. So the license manager is handling all that limiting for us, right. You cannot exceed the number of licenses available for you and you will get a denial event.

[6:25] The next type here is called pro-rated user, and this is available to enterprise customers of MathWorks. Essentially it is a named user license but instead of automatically counting a user as a full user, you know if you remember we talked about individual license, you count a user as one user right away, named user you count one user right away, this one is pro-rated. So we will be using the full user equivalent based on the elapsed time they accumulated within 12 months. So if you direct your eyes to the small table here, for example if a user used MATLAB for 33 hours, it will be in this bracket within the 12-month period, that user will only be counted as 38% of a full user. If that user is using toolboxes for 33 hours then that user will be counted as 75% of a full user. And for MATLAB and Simulink if you use less than seven hours it will not be counted, and it will take 128 hours or more to be counted as a full user.

[8:00] This pro-rated user licensing can also use the license manager. I said can because we have some customers that have a mixture of license-manager-enabled MathWorks, they also have standalone licenses, network individual, and designated computer, but overall they have this enterprise agreement with MathWorks with the pro-rated user licensing. So even though it’s using the license manager there’s no enforced limit to the number of licenses available for usage. The available licenses will just be arbitrary, usually just a very high number. You are allowed to use as many MathWorks licenses as you need. I think that’s the idea, and then later on after that 12-month period you will just receive a bill saying how much you’ve consumed. So this requires a lot of reporting and making sure that you are using just the right amount.

[9:15] Now combining license types is very common and I actually suggest that we do that. It is very cost-effective if done right. So a real quick few rules of thumb. One, named user must be reserved for power users. Even though named users are cheaper than concurrent users, if they are unused, if you have bought a bunch of named users and they’re just sitting there idle, then you are spending so much it becomes more expensive than just sharing a concurrent license. Second, power users should have named user licenses. Say for example all your licenses are concurrent and you have a lot of power users who need them most of the time, several days a week, several hours a day. So what they do to make sure that they have those licenses available is they will start reserving the license. What happens is that you end up having a named user right because they’re using it as a named user, but you are paying for the price of a concurrent license. Third on my list, concurrent licenses must be shared amongst infrequent users. If you have many users that need licenses but not often, then they should have concurrent licenses, they should share those concurrent licenses. So ideally you would have a combination of named user for power users and then concurrent users for infrequent users. There are of course other rules, and you can break these depending upon different factors such as internal policies, SLA, and other things. Feel free to contact some of our business consultants, you’ll find them on our website, they’re very easy to talk to.

[11:09] Optimization process. First is that we have to set goals. Is it right-sizing, consolidation, user productivity? Most cases it’s a combination of different things. And then do we want to optimize based on runtime usage, right, check-in and check-out or open and close of the applications? Or do we want to base it on true active usage? So we can imagine having the available licenses, this is what you own, and then you have the usage. Is it based on just users opening the application you count that, or do you want to know if they’re truly actively using it or not? So we have to set that as a goal because the following activities that we need to do would need to answer the goals that we set. Second, understanding your license agreement and its nuances. Are we using the correct metric for reporting? Are there optimization actions that we are not allowed to do? We have worked with a customer where in their agreement with MathWorks it’s specifically mentioned that they cannot use another application to suspend or terminate inactive sessions. We can’t do that. We know that there’s a lot of inactivity but we cannot do any of the termination or suspension of inactive applications. So what we did is we implemented another procedure we call user capping, because that one is allowed, and I can show you later what we did. But that allows us to optimize around the inactive applications.

[12:54] Third, of course we need to collect data, and this is why I mentioned we have to set goals from the beginning. Because for example if we would like to collect or to base our optimization on active data, then we need to collect data from each of the workstations. We need to know if they’re truly active or not and we do that by looking into what’s really happening on the workstation. But if we have set the goal and say well we would like to optimize it based on runtime usage, then maybe especially if you only have network applications we can collect data on license servers only. Fourth here is consolidate the data. We would like to put that into one central view, just having everything. If you have several license servers we would like to put that in one view, if you have workstation data we will have to combine that, we have to consolidate that. We need to implement certain mappings or certain normalization. And then analyze the data in relation to your agreement and goals. So if you have network named user we need to report on the distinct users, if we have concurrent we need to report on the maximum concurrent users. And then for the optimization goals maybe we need to simulate moving to another agreement, moving users from your named user to concurrent and doing some simulations. And last of course is to implement the action points from our analysis. We have worked with some customers and they know what to do but they failed the last part, to actually implement it. It’s really important to be able to implement the action points for our optimization process.

[15:07] Now collecting data from both license servers and workstations depends on, number one again, active versus inactive or just runtime. Number two, what type of licenses do you have. So if you have network licenses, network named users, or concurrent, we just need to collect from license managers, right, license servers. But if we have standalone licenses like the designated computer or the individual, then we also need to collect data from the workstations. And we also have to make sure that we are able to collect the correct data. In MathWorks they have a lot of toolboxes and usually those are the expensive ones. For the license servers it’s easy to collect both the application, MATLAB and Simulink, and all the toolboxes, because FlexNet just outputs the usage on their license status. But for standalone it’s not very straightforward. But at Open iT we’re able to do that. I think we are the only provider that can give you detail of toolbox usage for standalone applications. It’s easy of course to just meter matlab.exe, everyone can see that you are running MATLAB for this long. But MATLAB has many toolboxes inside, maybe you’re running Control System Toolbox, maybe you’re running Deep Learning Toolbox. We would like to know if you are actually using that, especially if you have the standalone licenses, you’ve allotted the license to a named user for instance or you’ve installed it on a designated computer. You would like to know if these toolboxes are really being used or not, because they’re really expensive. You need the license for the toolbox, you’re paying for that, you need the license for MATLAB and maybe Simulink and other toolboxes as well. So it’s really important to be able to drill down on this level of detail. So we are able to do that at Open iT, providing you that level of information both for license-manager-enabled MathWorks applications and also for standalone MathWorks.

[17:49] Now challenges in collecting data. Collecting data is important but it’s not always easy. A while ago we talked about maybe you can’t always collect that level of detail although you need it. Another challenge is a segregated network. Of course we need to first have the visibility of all the usage, everything that you have and how you use that. But then you can be blindsided. Segregated network for example, we still need to collect data from those segregated networks even though you can’t send it automatically to a central server, just collect the data and then we’ll need to transfer the data manually through a USB stick or something, that’s the only way. Because it’s really important to do this, we need to have as much visibility as possible. Work from home setup, we need to decide how to collect data and maybe you have to implement certain policies. Do we require our users to do a remote access to their machines in the office? That would be easier to collect data that way because we just need to install the clients on the workstations. Or do we allow them to use their personal laptops, install clients on that and then require them to connect to our VPN at least once a week so the clients can send data to the server? If all your licenses are network then the users will have to connect to the server anyway, so we can get the usage from the license server. It will be easier for us because we just need to collect data from the server. But you might need to collect data from the workstation as well if you have individual or designated computer licenses, or when we start collecting active versus inactive data, we need that. And another one is when users start borrowing licenses because of the possibility of disconnecting from the server. We might want to really know if they’re actively using the applications or not, and that means collecting data from the workstations.

[20:16] Consolidating data. Data from several license servers, including the segregated ones, we need to merge them, we need to consolidate them. For named users it would be easy, you just need to count the unique users. But say for example you have concurrent users, we need to process the data in a way that we’re able to do a true concurrency. Because if you have server one and you have server two and you have a concurrent license agreement, it will not be correct to just add up the licenses. We need to calculate, are they really using the licenses at the same time? Then we need to implement certain mapping mechanisms. Data from workstations and license servers have to be consolidated as well, because they’re coming from a different data source. There’s a possibility that they won’t match right away. So we need to do labeling, we need to do normalization of the data. And so at Open iT we are able to handle that for you.

[21:34] Okay now let’s go to the details on how we can optimize concurrent users. First of course is that we need to collect data from all the license managers that we have. Visibility is very important. And at Open iT we’re able to collect the data, we are able to consolidate them into a single server, like MathWorks combined here as you can see on the screen. You can drill down from the application level, could be MATLAB or the toolboxes, and then you can drill down on the user level. The data will be available both in real time and in historical, and that gives you several ways to optimize the licenses. Real time can be used to correct certain behavior, and I will talk more about that later on. And then you can do analysis on the historical data. For example here you know you have 200 available licenses and looking at the maximum peak concurrent users we have used 176. And you might say well I can cut down the 200 available licenses to 176. But that’s just one measure. If we add another data point into a report and say but how long did they use that, well 176 licenses are used concurrently only for 0.08 hours, 0% compared to the total time. Within 99% of use we only use 156 licenses. So depending upon the SLA you have internally, if you can allow certain wait time or denial, 156 licenses can be the number of available licenses, or 144 if your SLA is within 95% of the time. I’ll just mention that of course this would require renegotiation, or maybe you would stop renewing licenses for those that are not used, or you can also use this to set SLA internally.

[24:04] Now involving users in the optimization initiative can have a very good impact. So here we’re looking at the user details. We are able to see the date they last used the application, the days since last use, elapsed time, and we are able to see how many licenses they use in an hour, in a day. There are certain user behaviors that should be managed, like for example users checking out a license for an extended period of time longer than needed, 48 hours or more, they’re not using that, or they’re using multiple licenses simultaneously. We should try to notify them. It is of course possible that they have forgotten about this. But by doing this, by involving users, we have to make sure to communicate the purpose of what we’re doing, that this is not about being a big brother, we just want to manage the resources and limit waste. When the user understands, they are very willing to cooperate. We have customers at Open iT who were able to reach 99% efficiency by involving users, setting policies, and alerting and notifying the users.

[25:28] For the individual license and for the designated computer, again looking into the usage of those individual toolboxes and identifying, you know, are you using this? If you don’t use that then maybe we can reassign it. Remember that you can reassign or redesignate those licenses up to four times within a 12-month period.

[25:54] Okay, now I’ll be proceeding to pro-rated users. So again, pro-rated user is based on this table, a user can be a full user or a fraction of a user. Visibility of course is very important, being able to see who used what and when. And then instead of waiting for a report from the vendor, we should have up-to-date information about the current usage that you’ve already consumed. Because this is pay-per-use and at the end of 12 months they will give you a bill, but while you are consuming that you don’t know how much you’ve spent already. So it’s important to be able to have that information right away. At Open iT we have reports out of the box showing you these are the applications, and you can also include cost center or other groups if you would like to see how the usage is distributed. Then we can apply cost to it as well so you can really see how much you’re spending, not only when it comes to usage but in actual financial cost.

[27:12] We can apply the same ideas we talked about a while ago. Avoiding multiple sessions, because we are basing this on time spent, so imagine you have two licenses at the same time, you’re spending two hours of license instead of one hour only. We would like to avoid that. You can notify the users because we have real-time usage, we can notify the users, especially if you communicate to them what we’re doing, they will be very cooperative. We know that by experience. Also avoiding unnecessarily long sessions. Heat maps out of the box as well, here for example you can see that the users are just checking out the license and then they don’t check the license back in. We have to manage these long sessions.

[28:12] Now I’m just going to jump and show you this report here. So this is a side-by-side comparison of what is the effect of excluding long sessions on the cost. On the left here we have just the usage without any filtering. On the right we have simulated what would happen if we prevented long sessions. And at the end we can actually save 28% of cost by filtering out usage outside prime time. So if we can control long sessions, if we can control some of the behaviors like checking out multiple licenses at the same time, we can really save a lot. And this is true not only for pro-rated users but also for concurrent licensing and other license types as well.

[29:10] Double counting is another thing. Since we are looking at combined servers it is possible that a user will be counted twice. So for example here we have Deborah using the Control Toolbox from both of these servers. If we just look at them individually you might end up counting her as 1.19 users instead of just counting her as one full user. So it’s important to be able to combine different sources and different servers into just one central combined license server.

[29:47] Now here we have an optimization I mentioned a while ago. We work with this customer and they’re not allowed to implement any optimization like automatic harvesting of the license, freezing the applications, or closing the applications automatically. So what we end up doing is using the option files for user capping, because in option files it is possible to set a maximum available licenses for a feature. So it will limit how many licenses a user group can check out. We look into the license efficiency report and we use that in order to set the user capping, and we’re able to provide them the scripting that can modify the option files automatically. Of course this requires some professional service during that time because it’s a different optimization procedure.

[30:53] Okay, that’s the end of my presentation. Now I am ready for questions if we still have time.

[30:59] Mae: Yes, thank you so much Malou for your insightful presentation. Yes, we are very excited that our audience is very interested in our webinar today and we have received questions over optimize@openit.com. So given the time that we have I think I’m just going to throw at you one question, and here goes. Can Open iT get the license details from license files including available licenses, expiration, and license type, or is it a manual input?

[31:50] Malou: Hi, okay, sorry. Open iT reads license files, there’s no need to do a manual input. If you have several license types on the same license, network named user, concurrent, we can distinguish between them. We’re also collecting the asset info which can be used to map to your invoices for example. License expirations are collected and then you can also set alerts to notify yourself for upcoming renewals for example.

[32:29] Mae: Thank you, and yes we have time for one more. We have an enterprise agreement with MathWorks, we have several license servers and still have standalone licenses. How does Open iT deal with this type of setup?

[32:43] Malou: So we need to collect data from both license servers and workstations that still have the standalone licenses. And then we need to combine the data using mapping mechanisms from Open iT. After doing that we can have a consolidated view of usage from all the servers and workstations, and then after that we can use the existing reports, the out-of-the-box reports, for analysis.

[33:15] Mae: Thank you very much. So that wraps up our webinar. Thank you very much Malou.

[33:22] Malou: Thanks Mae and thank you everyone for joining.

[33:30] Mae: And also thank you to all our attendees for joining our webinar today. A quick reminder that this webinar is recorded and will be sent to you via email, and we will also upload it on our website together with Malou’s presentation. Since this is a two-part webinar series let me take the chance to invite you to our second session on July 27th, Implement a Cost Reduction Strategy for Your MathWorks Toolboxes. You can register through the link shown on the screen or visit the Open iT LinkedIn page for more information. If you have additional questions for this session or for the next you can simply send them to optimize@openit.com. And 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. If you have any requests, suggestions, or comments please write to us at podcast@openit.com. We will see you again at the next session. Once again this is Mae, your host for today. Thank you and stay safe.