On-Demand Webcast 60 min

Hybrid-Converged Infrastructure: Deploying HCI 2.0 with Existing SANs

Webcast Transcript

Enrico Signoretti: Good morning or good evening, everyone. This is Enrico Signoretti and welcome to this GigaOm webinar. The title of this webinar is “Hybrid-Converged Infrastructure: Deploying HCI 2.0 with Existing SANs.” The goal is to talk about hybrid convergence and why enterprise is adopted. But also discuss some of the limitation imposed by this type of architecture and how to overcome them with different approaches. At the end of this webinar we will also take a look at the technology and market trend and what to expect from the next generation of products available in the market.

I’d like to make this webinar as interactive as possible. We will run three polls during this webinar for which we will share the results before the Q&A session, and also you will be able to ask questions by using the panel that you have on the right of your screen. We will try to answer to these questions at the end or if they directly during the presentation.

So the webinar is sponsored by DataCore and today with me I have Manish Chacko, senior product manager at DataCore. He will be with us to bring the point of view of his company regarding HCI as well as feedbacks that he receives usually from the field. Hi, Manish, how are you today?

Manish Chacko: Hi, Enrico. I’m good, thank you for asking and welcome. Good morning, good afternoon, good evening to everybody that’s joined the conference. Thank you for joining. My name is Manish Chacko and I’m the senior product manager at DataCore. I’ve spent quite a few years in IT systems, engineering, and mostly, nearly five years at the Cloud [Plan 0:02:04] Computing Group at Dell. And now I’m at DataCore. And DataCore, as some of you know, is a software defined storage vendor, and we also offer hyperconverged infrastructure. We have over 10,000, been in business since ’99, and are a pretty stable company.

And we are excited to be here to talk to all of you.

Enrico: Great. So, and just to give a little bit of information about myself also, since probably not everybody knows me already. So I’m data storage analyst for GigaOm. I started more than 25 years ago with assembly, and then moved from there in several different roles, both in the technical and in the marketing product strategy positions. So I provided consulting services to enterprises and vendors of all sizes across Europe and the US.

And you can find my latest reports on GigaOm but also comments and other news on my personal blog. It is [juku.it 0:03:18]. So I’m pretty active on social media and you can also find me on Twitter and my handle is esignoretti. So let’s start the presentation then. And we have a pretty interesting agenda. We’ll talk about the good, the bad and the of ugly of hybrid-convergence, then we will look into the tradeoff that tradition HCI imposes. And also to the evolution of HCI and the benefits for end users regarding new technology that vendors are developing. So at the end we’ll also take a few questions. We will discuss about the new option that allow smooth migration to HCI, as well as, again, features that these vendors are talking about right now, and we will see in the next 12, 18 months.

So let’s start with the first poll. Do you have already adopted an HCI solution? So probably you will have a screen popping back, and here you go. And you have the ability to select one of these. We will share all of these answers at the end of the session, of course. Okay?

Let’s proceed with the rest of the – of the webinar. So what is the good of HCI? Okay. Of course there are several benefits in adopting an HCI solution, and the first one is the simplification of the infrastructure. Moving from standard infrastructure that are made of several components – storage, servers, hypervisors, and networking. Okay, HCI provides a simplification collapsing resources in the same node. And usually in a scale-out fashion you an add more node to have your infrastructure. It’s easier then to design and display than any other type of infrastructure.

The other good thing is that usually you find a very familiar environment to operate. Okay, meaning that nothing changes from your old [unintelligible 0:06:28] infrastructure to the new still [unintelligible 0:06:32] infrastructure. So if you were using [VM-ware 0:06:36], that remains your main point of contact and you manage the system through its console and so on. And the same goes for hyper-V or even, if you are using [unintelligible 0:06:50] and so on.

The other thing that is really good is TCO. So when you compare the TCO with traditional systems, okay, especially obsolete ones, you find out that the TCO is very good, just because the other points that we already discussed. So you have a great improvement in the total cost of ownership of your infrastructure. And last but not least, as I said, this kind of infrastructure are scalable and in a very easy fashion. So you keep adding new node when you need more [unintelligible 0:07:36] from us, and this is very predictable, because each node that you add to the infrastructure usually brings resources to manage, fixed on average, fixed amount of [unintelligible 0:07:51] machines. You can carry more [E-ops 0:07:53]. And so on.

So I don’t know, Manish, if you want to add something to this slide, if you see the same things on the market or there is something different that DataCore finds.

Manish: Yeah, yeah. I’ll address a couple of those. I think when it comes to TCO figures, we had a customer out of Poland called LERG, I’m not going to attempt to pronounce the exact name. But they went from having a bunch of – they had a SAN and a few VM-ware servers, and Linux server, and they replaced it with two DataCore HCI nodes and they got a 50 percent cost savings, right?

As far as performance, we’ve had the ECSO from [Mettard 0:08:42], Oregon, which is a 911 communications system. They dropped down their [unintelligible 0:08:49] latency 200 milliseconds down to five. So definitely good performance there. As far as – and also NASA, the [unintelligible 0:08:57] Space Center, they got a roughly 10X performance gain. So definitely I’m seeing very similar things happening in the field, especially with our customers.

Enrico: Very good. But you know, if the first one is the good, then I have also to talk about the bad things that happen when you adopt HCI.

Manish: Yes.

Enrico: So not – okay. We said that scalability is an advantage usually. But there are also drawbacks with fixed scalability [unintelligible 0:09:36] the scalability that you have. In fact I’m talking about you keep adding nodes, and usually these nodes are identical, okay. Sometimes you need to scale more on storage. Other times you need to scale more on CPU, because your application are, you know, data hungry or CPU hungry or you need nodes of different types because, you know, not all the storage is the same. Sometimes you have a lot of [unintelligible 0:10:03] sometimes you need just throughput, sometimes you just need capacity, okay?

So in this case it’s not easy, not all the HCI that you find on the market are very flexible. So you uses a little bit of flexibility. And the other things that happens, more or less for the same reason, is this solution are not designed to cover all the workload types. Okay. Can happen that, you know, you have a big data workload, and in the same cluster you may have, you may want a high transactional database. Okay.
They fight to each other for the same resources, and sometimes it’s really tough to get it well designed, okay, so you have to do some over-provisionings, okay, QS mechanism always present in this solution. So it’s very tough to cover all the workloads that your company usually have. And last but not least, you can’t have it all. It says that most of these solutions are designed to cover one, maybe two of the major architecture. And when I talk about architecture I mean Core data center design or Edge computing, or you know, and connect to the Cloud.

In most of the cases it’s the solution is designed to address Core kind of challenges. It’s very tough to find it at the edge, because it’s not really competitive. There are a lot of drawbacks, it’s expensive and so on. And when I talk about Edge, for example, I talk about remote offices, I talk about especially in the retail, remote warehouses or shops. Okay. So it’s not always the best solution for everybody. Okay? Manish, do you want to add something on this too? Because I think you have a lot of installation across several verticals so maybe you’ve found something that could help our audience to understand better.

Manish: Yeah, yeah, absolutely. And these are valid points, right. There’s always, there’s something really good about a solution, there’s going to be something bad about it. In this case it’s HCI. And a lot of vendors have some of the problems that you talked about, the predetermined scalability. But the advantage of having DataCore being software, is you can have different types of hardware, as long as it’s [unintelligible 0:12:51] server.

So for example, if you have multiple workloads, you can deploy our software on different hardware. So for example, if you have multiple workloads, you can deploy our software on different hardware. One may be more CPU intensive and one may be more storage, then so that can address multiple different workloads and systems. I’ll point you to LERG, the chemical manufacturer in Poland, they have business critical ERP and Oracle solutions, but they also have exchange in others that are important but not maybe as critical.

And so they use different servers for different workloads. And that’s how they solve the problems, since we are software based. So yeah.

Enrico: Yeah. Really interesting. So you’re – just to dig a little bit deeper. Your architecture is designed to provide not just the resources to the HCI environment but actually because you can deploy it also in a hybrid fashion, you can not tune the system a little bit better than you can usually do with other type of system. Did I get it right?

Manish: Yeah, yeah, exactly. So their software, as long as you have an X86 server, if you have workloads that need more Cores or more Quad frequency, for example, or maybe you just need a 4U node with a bunch of SSDs or hard drives or whatever the case might be, we don’t have identical appliances because we are software only. So you can install that on any hardware that meets your needs. But on the hybrid side also, that’s a good point you bring up. Traditional HCI solutions typically only access the local storage on the X86 servers.

But with our software, we not only let you do that, but we also let you connect your existing SANs and things like that. So if you do have existing storage [unintelligible 0:14:46] networks, you don’t have to abandon them; you can use us in a hybrid fashion. Yeah, it’s a good point. Thanks for bringing that up, Enrico.

Enrico: Okay, very good. And what is – even worse than bad, we have some ugly points about HCI. Okay. There are a couple of things that are really important to note. And one is that HCI is usually a kid of rip and replace approach, meaning that when you decide to move to HCIs, you are forced to build the entire solution and migrate [unintelligible 0:15:29] okay.

Yes, you can start small, but actually there is an initial step that sometimes is quite important. So you have to buy a bunch of nodes. But you know, because there are also financial kind of problematics, maybe you have the depreciation of your storage system, which is different from the server system. So you find yourself in a situation where your service are pretty new and you don’t want to give them back. But your storage on a network is pretty old and needs a replacement, or the contrary, of course.

And when you buy an HCI, the idea is that you replace all, the entire infrastructure. So initial investment could be quite expensive, especially for small, medium businesses, okay? And it forces you to make a decision and to do immigration that the day that you decide to implement this. Okay? Knowing a little bit about your architecture, maybe you can have a different opinion on this, Manish.

Manish: Yeah, so this I think kind of dovetails into what I said earlier. We definitely have some customers that are like, you know what? We don’t mind replace – we are getting, seeing the huge benefits of HCI. But we also do have some customers that are like, hey, we like your hybrid approach to HCI. Hence [unintelligible 0:17:12] HCI 2.0, right? Because what they can do is, they can keep their existing SANs and utilize them still, because even if they are aging gear and they may not be as high performance, you can, you know, you can access that data and then everything moving forward can be on the HCI server that you deployed with DataCore, so you can manage them both.

A good example is, I was speaking at the Texas Tech summit last month in Houston, Texas, and we had a customer who has an architecture firm in Houston. They were having this exact same problem. They were seeing the benefits of HCI for their particular [unintelligible 0:17:59] don’t remember the exact application. But their IT man [unintelligible 0:18:03] is like, well, we spent a lot of money on expensive SAN, how can we, you know, justify going to HCI?

And they liked the fact that we are not just saying rip and replace. We are saying just add and use it in conjunction with your SAN. So that helps a lot of customers, especially who are like, well, we don’t want to consider it because we’ve invested in SANs and we are like that’s fine and we are going to let you use that in addition to deploying HCI with the DataCore software. So that’s a huge benefit that a lot of our customers have seen.

Enrico: Very good. What I want to add here, is so I have another slide, which is all about the trade-offs of HCI, okay. I want to be very clear on that, is that I, I think HCI can cover easily 80 percent or more of the workload that you find in a traditional enterprise. Okay. So all these [unintelligible 0:19:07] all the things that you are already virtualizing, okay, are easy to move an HCI. This is the same, at least at the top layers, the same identical platform. And what you get is that it’s, at the end of the day you can get the same identical environment.

Unfortunately, okay, there are a few kind of workloads. They’re not specifically in the right range, okay, in terms of capacity [unintelligible 0:19:40] latency that you want from your application. So when you start to add an application that potentially need a very low, a very small capacity, okay, but at the same time very, very low latency, like high transactional workloads, okay. So [unintelligible 0:19:59] very, very quickly for IOT, for example.

So you keep adding data at a very high pace. Okay. For this solution, for several reasons, okay, sometimes it’s because the scale out mechanism is not efficient, 100 percent, so you add latency and you are in consistency on this latency. Okay. So it’s not perfectly linear and flat, your latency measurement. Or you have this, as we talked about before, Edge, Core application, okay. You would like to have the same identical architecture both on the Edge and the Core. But to keep it that way, you have to spend a lot of money to get the level of [unintelligible 0:20:51] from us that you usually get on the Edge.

Maybe it’s not necessary or maybe it’s not enough. Okay. On the other side of the spectrum there are other applications, okay, that are not really important for many traditional applications, especially in the small, medium business. These application are usually deployed on the Cloud. Okay. But as you grow bigger and bigger, okay, the Cloud is very expensive and there are a lot of repatriation projects. For example, there are a lot of applications that are better to, on your premises, okay.

And even though for some of these workloads, the size today of the cluster, or the size of the storage, is still relatively small. These are the kind of applications that are growing the most in the storage industry at the moment, okay. Everything that is structured as analytics, everything is even structured data, which I didn’t put in this, but it’s still in the very high capacity, even if it’s high latency, okay. So there are a lot of applications that you would like to put in the same infrastructure, because the first thing that we said is simplification of the infrastructure. Exactly first thing.

But if you need to add different components to your infrastructure to make this happen, so bare metal Linux, with [unintelligible 0:22:31] storage and so on, these becomes, again, a new heterogeneous infrastructure, and complexity grows again. So you miss the point of HCI with the simplification and with the benefit that we describe at the beginning. Okay? Manish, what do you think about this slide?

Manish: Yeah. I generally agree with this. I think HCI definitely does cover 80 percent. I think in addition, if you look at the left side, the high transactional workloads of in-memory databases, that’s also something that we cover with DataCore, [Sun Symphony 0:23:16] and also with our HCI product. So the examples I’ll use again is the 911 system in Mettard, Oregon. They were having a [sequel 0:23:26] server that was, you know, having their 911 system data documented, and they went from a 200 millisecond latency down to 5 milliseconds. So we different do play in the field where you have high speed transactional databases.

Because basically they block their storage software. I would say similarly to the LERG chemical manufacturer in Poland, they saw a 5X performance improvement. Similarly with databases, and then of course on the right hand side, perhaps with HBC we do have a customer which I can’t get in the details, but it’s the NASA [unintelligible 0:24:08] Space Center, and they have noticed a 10X performance increase.

So definitely the 80 percent is a good rule, but we do have stuff from outside the range on both sides, where we have customers that have successfully implemented it and seen the benefits. So yeah, definitely.

Enrico: Okay, great. Before continuing with the presentation, I have another poll for you. And again, I will share this, the results later, before the Q&A session. So what are the major pain points of your current infrastructure solution? So I will give you a few seconds to answer. Just a few more seconds.

So it looks like, that 50 percent of that, and these already answered. Just a few more seconds. And done. I think we are okay. Almost everybody answer it, to the poll. So we can continue. So, and let’s go to the next session and take a look at what is going to happen in the next 12, 18 months, okay. Before starting, I want to tell you that not all the vendors are working on the same features, and for some of them a few of these are already available, and for others are a work in progress. While there are others that are still at the very early stage for everybody.

Okay. So if you already have an infrastructure and you see that you are already – HCI infrastructure, I mean, and you see that you already have this kind of feature, okay, it’s possible. But consider that it’s not usually for everybody. So let’s start with the first feature that you usually that a user asks for [unintelligible 0:26:24].

And again, a better, more flexible architecture. Okay. As I said, one of the things that happened to the – to HCI vendors is that they fix on a few certified nodes, of the few configuration. This is good to cover a large number of workloads, but actually not all of them. All the HCI vendors, some in a way, some in another, they are working to simplify the architecture and to bring more option to the table. Okay. And this first point comes with the second, which is improved versatility, which means a lot of things, but actually it’s connected to the first one.

So I mean having the ability to reconfigure the system, to give more, more options to the customer, to make the system as flexible as possible. I don’t know if you have a similar feedback from your customers, Manish, and how they feel about these points.

Manish: Yeah, so like I mentioned earlier, they do like flexibility of having software so that they can pick exactly the hardware that they want, depending on their workloads, and the performance that they’re looking for. So they definitely – I’m seeing a lot of customers happy about that, and being software we can literally, as long as the X86 server, we can be deployed anywhere. And then also, that also removes the, some of the issues with just going with similar appliances where they are locked into maybe a certain configuration or things like that.

So I think from a hybrid-converging sector, or more accurately the hybrid-converged infrastructure that we are offering here at DataCore, I think they are definitely appreciatively of that fact, that we give them flexibility and versatility.

Enrico: When you talk about this flexibility [unintelligible 0:28:49] do you also mean adding more different types of storage, for example? I mean now we have [unintelligible 0:28:58] Obtain, for example, okay, which is a small device but actually really, really fast, okay. So are you planning to deploy this kind of support to your customers?

Manish: Actually, they’re still trying to get feedback from our customers and maybe this – from our partners as well, to see what’s the interest right now. We hear about [Interlopta 0:29:26] and we hear about NBME, and NBME fabrics, so we’re still trying to determine, you know, what’s the best way we can help our customers. So we still are in the process of collecting that information. So yeah, definitely hearing all those things.

Enrico: Very good. And to keep on the same discussion, there are other a few things that are really interesting for our end users, and support for, concurrent support for hypervisor and not just the hypervisors, and we have the machines. Okay. So there are more and more end users, okay, asking to support [unintelligible 0:30:10] and containers in general, okay? Is this something that you see in the field, from your customers?

Manish: Yeah, yeah. Definitely. So what we’re seeing right now is that, you know, it’s not just the VM ESX hypervisor. We have customers using the Microsoft Hyper-V as well. And then as far as containers, we have definitely heard from customers about Dockers, and we actually are looking at how containers can help our customers. In fact, I think we did have a blog about the container persistent storage that some of our customers are – the challenges they’re facing. And so we’re actively working on that as well.

In fact, if you go to the DataCore site, we do have a article on that as well.

Enrico: Okay. Another thing that all the [unintelligible 0:31:17] are looking to develop, or develop activity, are developing, is more Cloud integration. Okay. I mean tiering to the Cloud for some data or [coughing], sorry, disaster recovery to the Cloud, these kind of things. I have to say that the end user in Europe at totally different from what we see in the US. Okay? So I come from Europe and I can tell you that in Europe the Cloud is still behind – adoption of Cloud is still behind, when compared to the US. And the Cloud market is much more fragmented, so a smaller provider.

So everything is a little bit more complicated. People still like to have control over their data, the infrastructure, so it’s not always easy to say, okay, I give my data, part of my data to a Cloud provider. But in the US, it’s a stronger movement, okay. And there is more request about these kind of features. Another thing that maybe somehow is connected to the Cloud is – but also to improve the versatility that we already talked about, is the interest for end users to – having support for more protocols, including file and object storage. Okay.

End users want to concentrate more workloads, more types of data on the same infrastructure. Again, they wanted simple, they wanted easy to deploy, they want easy to manage. And having, with the same kind of methodology deployed, file an object as they do for the VMs is something that they really like. Okay.

Do you have something on this to add, Manish?

Manish: Yeah. I would be really interested to hear, especially from the people in this webinar, what the interest level is in file and object storage or if it’s more of a – I mean just in performance and functionality. And I’m not really interesting in block file or object or if it’s the other way around. So definitely would like to get more feedback on this, on this topic, yeah.

Enrico: Okay. The other point is analytics, okay. So knowing better what the system does is a great thing. There are some companies that lead to market. They are not actually the most interesting ones, unfortunately; or not in the HCI space, but actually these kind of [unintelligible 0:34:27] are really impressive because it gives, you know, you can get the steps of the system on the first side; you can have capacity planning simplified. You can have errors before they actually happen, and so on. This is something that everybody’s working on and it’s getting more and more interesting, especially on the analysis of performance and again, capacity planning and stuff.

And I don’t know if Manish, you have your customer asking for analytics and how [Sun 0:35:12] Symphony and your HCI solution implement this.

Manish: Yeah, absolutely, Enrico. A lot of our customer are saying, hey, we want more analytics, more telemetry information from usage, from the storage, from whether it the capacity planning, whether it’s identifying, you know, how the problems are bottlenecks in the storage, are occurring and things like that. So the customers are very interested and they are really actively working on improving the analytics engine so that we can provide this data to our customers. One thing that I did forget to mention on the containers, the DataCore CSI plug is available on [unintelligible 0:35:56] that’s the one with [unintelligible 0:35:58].

And then we also have the DataCore SDS Docker [unintelligible 0:36:01] plug in available on Docker Hub. So if the people [unintelligible 0:36:07] the webinar today, are listening in later on [unintelligible 0:36:12] they can definitely go to the [Github 0:36:14] or Docker Hub and get our plug in. So –

Enrico: Very good. And there is another thing that is maybe a little bit more far out in real product environment. But actually there in an increase in interest in [unintelligible 0:36:35], especially for Edge applications. Okay, and I don’t even ask you if you are a real demand, Manish, on this; but actually on VM-Ware now, having [ASXI 0:36:49], and a lot of projects starting at the edge, and at the edge for several reasons you want power consumption very lower, you are not really interested in the very [unintelligible 0:37:01] and so on, is getting a lot of traction, okay.

Most of the time with Linux and, you know, these are in better systems, so they are very simple kinds of systems. But actually because the need of share storage, but of the need of cluster mechanism, because and because and because ARM [A60 0:37:23] on ARM is, you know, getting some interest, and you know, sooner or later we will see more and more solution adopting ARM for the Edge, and of course HCI solutions will follow.

Okay, so before getting to the last part of our webinar, I just want to ask you the last poll. The last question is, what would you like to see from your HCI vendor in the next 12, 18 months? Let’s start with the poll. Okay, I think we are, we are good. So practically everybody answered to this one. Okay. So let me go directly to the answers first. And then we will do the Q&A session. The first one – okay, the first one was do you have, already adopted an HCI solution?

Okay, we have 17 percent of the [FNDs 0:39:29] saying yes, 40 percent no. And 40 percent are actually looking into it. Only 3 percent are not really planning anything with HCI, maybe not yet. Okay. And this is an interesting result, that more or less is in, aligned with we are seeing in – lately. So there is a lot of interest but not everybody already deployed this solution.

The second poll was what are the major pain points of your current infrastructure solution? Thirty-three percent is cost; 38 percent is lack of flexibility; 29 talks about scalability issues; 17 doesn’t cover all my use cases. And 29 percent none of the above. Consider that the total is more than 100 percent because it was a multiple answer kind of question. And the third poll, which was what do you expect to see in the next 12, 18 months from your infrastructure provider, 36 answered more Cloud integration, tiering, and/or disaster recovery; 21 percent want a file solution on top of their HCI; 7 percent [unintelligible 0:41:17] object storage; 25 better support for containers. Which is quite interesting because containers are just entering in the enterprise space.

And 11 percent, that’s none of the above. Okay. So wow, a lot of questions are coming in, so let me start with the first question that we have here. The first question is, does it make sense to run more than a single hypervisor at a time? I think this is a tough question. It says that usually no. Often in the past I saw a lot of installation with multiple hypervisors, especially in very large environments, but in the end you need two different teams. There are two different skill sets, and then in end, the advantage is you get maybe from a different licensing model or a different way to deploy it, it vanish immediately when you have to, you know, just pay the training for all these people.

I don’t know, Manish, if you have a different opinion on this.

Manish: No, I think that that’s a very fair statement. When you have a large organization and you need to have multiple team and multiple skill sets, that definitely causes some issues. So I would say you [unintelligible 0:42:56] a hypervisor and then, you know, build the teams and capabilities around that. Of course each one has their own advantages and disadvantages. So definitely makes sense to test it out with your workloads before deploying, but I think you’re right.

Once you’ve picked your hypervisor, I’m not going to recommend one or the other. Each one has their benefits. You should stick to that, absolutely.

Enrico: Okay. The other question is, why I should I rank [unintelligible 0:43:30] nested in VMs, or on top of an [unintelligible 0:43:35] I would think hybrid but actually want to say – I talk about it, infrastructure. Isn’t it better to run it on bare metal Linux? So my honest opinion is yes, it’s better to run it on bare metal Linux. But only if you have enough applications and workloads that need it. If you’re just starting and you are in a development phase, or you are, you know, and you have just a few applications using this, then it’s better to run into VM, because it’s just, you know, you’re taking advantage of something that is existing already in your infrastructure, and doesn’t cost too much to manage or, you know, or to maintain anyway. So I don’t know if you have a different opinion on this, Manish.

Manish: Yeah, yeah. No, I generally agree with that. There are some companies that can benefit from the microservices offered by Kubernetes or [Locker 0:44:36] in that case. Nesting it in VMs. But generally, you are always going to get better performance from bare metal Linux. So yeah, yeah, agreed.

So I think this one is for you. So how is licensed DataCore software?

Manish: So DataCore software, like I said, you know, we’re software only. You can deploy us on X86 servers. So we license per terabyte, right? So you just look at your capacity and go, okay, they have, you know, 10 terabytes or 100, whatever it is, and then simply purchase the [unintelligible 0:45:16] software license that you need and that’s it, nothing complex at all.

Enrico: Okay, there is another question for you, actually. What is the [unintelligible 0:45:25] DataCore configuration? If I want to install it in a remote site, and I can I manage many sites concurrently?

Manish: Okay. Yeah, so definitely I can’t make a specific recommendation without knowing what the customer needs are, obviously. But we would typically, what we have seen our customers do is, they’ll have two servers. If it’s a small remote site, they will have maybe one new server or something like that, and then they’ll deploy two of them for [HCI 0:45:56] and they’ll have a remote office, branch office kind of deployment. And they’re able to manage it successfully. And yes, you can manage many of them remotely.

Typically, you know, depending on your workloads, you might go for CPUs with more number of cores, especially for ATI. And then if you’re just having single [unintelligible 0:46:24] applications, they can’t really take advantage of multiple Cores or you’re not doing HCI, then in that case you would go for a higher [unintelligible 0:46:30] frequency. So it really depends on what they’re trying to do. But typically two 1U nodes running DataCore software.

Enrico: Okay. Sorry. So there is another question. This is a little bit tough. I don’t know how to answer to this one. But Cloud deployed should be more of an open source side, using KBM hypervisor. Easy to deploy and integration is – especially the network inside. And neither of our [unintelligible 0:47:05] read, write of [unintelligible 0:47:08] so from this point of view, maybe I can answer this if I understood it.

So if you want to use KBM, and VME or SSD, anyway, you can use the hypervisor layer or a [Sun Symphony 0:47:30] layer, the same way, okay. It doesn’t change it that much, I think, from this point of view. You provide resources to the hypervisor, the resources are provisional, as standard component, and you can use them whatever the – if you are in Linux or ESXI or whatever. Am I correct, Manish, with this?

Manish: Yeah, yeah. Absolutely. We definitely have some customers using KBM. I’ll have to double check on the tech support in group. We primary focus on ESX and Hyper-V. But we definitely have customers that use KBM and we try to help them as much as possible. But they have definitely done this successfully. So I can understand why that comment is there. So that’s a good point.

Enrico: Okay, there is a question here, and what is an ACL, okay? And ACL hybrid connectivity list. So most providers offer a list of companies that are supported, tested on the platform, and on which you can, you know, trust that everything will work fine. Okay? Do you have an ACL in this regard, Manish?

Manish: So Enrico, so we do have, we do publish the hardware compatible list. But a real, honestly, any X86 server can definitely be installed. I’m looking at the question and it says the difference between Cloud computing and ACL. So I don’t know if he meant HCI. If he meant HCI, I would say briefly, you know, hyper-convergence, or a hybrid-converged infrastructure, simply putting service, storage and compute on one single node, usually on PRAM, so that you can quickly scale out or scale up as needed, and then deploy workloads.

But also just consume it only when you need it. Because you know how much capacity a single server provides you and you can quickly calculate, depending on your needs, if you need two or a five or 10 nodes or whatever the case is. And then Cloud computing, I’m assuming, is just having a workload in the Cloud. Each methodology has benefits, ups and downs, I guess. But if the question was ACL, I’m yet to find any hardware, whether it’s I-Scuzzy or fiber channel cards; we support all the major vendors. Same with servers. Dell, HP, Lenovo, Fujitsu, or IBM – I mean just about everybody that has X86 architecture.

Enrico: Okay, maybe so you answered it right in the sense that the question was [unintelligible 0:50:29] HCI and – not HCI, yeah. So I tried to make it easier because it was really confusing. Maybe it was just a type of – so another question. How would you say [unintelligible 0:50:43] think of composable infrastructure vs. HCI?

So I jut wrote a paper on this. And what I find during my research is that if you need a thousand or dozens of thousands of [unintelligible 0:51:00] a year, probably you are good with composable infrastructure. If you need hundreds or even more service, probably you are good with HCI. Meaning that composable infrastructure organizes resources in the center, in a large of pool of resources. So you have NDME, and you have compute nodes. And there is an orchestrator that reconfigures the networks and storage resources and CPUs to make it, you know, configured, to make it usable from your application. Usually these applications are bare metal Linux space, and one application for node and so.

The good is that you can reconfigure everything very quickly. So if you have 1,000 nodes for a big data cluster, and in the morning you need to run a job for, I don’t know, a dupe, you just have the image with the dupe, you distribute it, you have your cluster running, and maybe a day later or in the afternoon you can run a totally different kind of application, a different stack, even a different operating system for a different thing. Okay.

And it was – but actually if you don’t these kind of numbers, HCI is much more efficient, okay? So you have [unintelligible 0:52:36] –ability that is brought to you by the [unintelligible 0:52:40] layer. Okay. You don’t have the same flexibility but actually you don’t really need it. And the cost of implementing this kind of infrastructure is very expensive. So again, composable infrastructure are usually pretty big, while HCI are usually smaller. Okay. In size.

And the benefit of composable infrastructure is not really visible if you don’t have a large amount of notes. Wow. The questions keep coming. So let’s answer two more. Is it a recommended best practice to have [unintelligible 0:53:27] configuration for HCI, meaning similar node configurations? Maybe you can answer this one. Okay. It depends on the architecture of the product.

Manish: Yeah. Yeah, definitely. I mean if you have a workload that’s pretty consistent, then it makes sense from a management perspective. But honestly, with DataCore being software, you can actually have, you know, nodes that are not similar, but it’s the CPU, the storage or the 1U, 2U or 4U. As far as we are concerned, a node is a node. We don’t care if it’s different in compute or RAM or storage or anything like that. We see a node as just another logical node. And we say do what’s best for your environment, right?

Typically HCI [unintelligible 0:54:20] talk about having similar nodes, but I thin from a DataCore perspective we are like, we give you the flexibility and use what’s best for you. Yeah.

Enrico: Okay. And then I have a tough one. Still for you. How serious is DataCore looking at offering a Cloud connector or Cloud connect, [DRQ 0:54:46] managed by their partners? This is really complex.

Manish: Yeah, it is. And I’ll definitely – I would like to take that offline. So I don’t know, Enrico, if you can provide my email address and things like that, so we can definitely have a conversation offline about that. So –

Enrico: Okay. I will – at the end I will let you give your coordinates. And last one, and then we will have to wrap up, how can integration, [AIML 0:55:18] for predictive analytics, and also integration, help in NCI, and block chain for preservation of [unintelligible 0:55:28]? Is it possible? So I think this is two questions in one. Why is – how can it be good having [unintelligible 0:55:39] machine learning in predictive analytics? And the other one is block chain for [unintelligible 0:55:46].

I don’t think it really applied, the block chain question [unintelligible 0:55:52] with the HCI in this case. So, and maybe it should be, the question should be formulated differently. But of course analytics is very important and AIML could be really interesting. I don’t know if you are doing something in this regard, Manish.

Manish: We’re definitely investigating that. I think the question is a very interesting one because using artificial intelligence and machine learning, you can analyze the work loads that the HCI nodes are handling, and look at the fluctuations and then, you know, once you form a baseline, you go, okay, are these seasonal fluctuations based on, for example, maybe a large shopping retailer during Christmas-time? Or perhaps they are geographic, as in your US office comes online after your European office.

So there’s different factors. Those are some of the ones that are easy for me to like retail. But using AI [unintelligible 0:56:54] can definitely give you a lot more insight into how HCI can be modified to further benefit the organization. So yeah, we definitely investigating into that, but that’s all I have to say about that. And then I guess the block chain, that’s a good point. It’s not really HCI, but generally being interesting in technology, I’ve seen a few vendors using AI with block chain to prevent [Didas 0:57:23] attacks. But I’m no expert on that. I just read up once in a while on it. So yeah.

Enrico: Okay. We are reaching the top of the hour, so this is time to wrap up this session. And thank you very much, everybody, for joining us today. And thank you, Manish, for your time, to talk about HCI and also DataCore. Maybe you want to share a few links about DataCore, and that we can reach out to you to keep the conversation going, especially because there are a couple of questions that the audience asked, and maybe you can answer online.

Manish: Yeah, yeah, absolutely. And thank you, Enrico, for setting this up [unintelligible 0:58:08] and thanks for everybody listening in and asking your questions. This has been very helpful. I’ll just put my information in the chat window. I’ll provide email address. And my LinkedIn is the one that I’m usually active on. Twitter, every once in a while I’ll post some product announcement. And I do once in a while look in there. But mostly on LinkedIn. And then of course our company website.

But you can go to the resources section and look at the customer case studies have deployed both HCI as well as traditional software-defined storage and understand why they pick us. So I’m going to put that in the chat window and hopefully that will show up. Yup.

Enrico: I suggest everybody to look for Manish Chacko on LinkedIn and DataCore.com on the Web. Our team is putting it up, so probably you will receive it in a second. And I think with this we can call it a day. Thank you again, Manish, for your time. Thank you, everybody, again, for joining us today. Thank you, GigaOm, for hosting the webinar. And have a nice rest of the day or a good evening, everybody. Bye bye.

Manish: Thanks, Enrico. Thanks, everybody. Have a good evening.

Read Full Transcript