Intel does not even have an OpenStack distribution. So why are they so involved in the project, and what are they doing to accelerate its adoption? Ruchi Bhargava is Director of Datacenter and Cloud Software Engineering at Intel's Open Source Technology Center, and she was our twenty-third guest on the OpenStack Podcast. During our interview she answered those questions and also talked to us about:

• How she got into tech
• Her opinion on "leaning in"
• How you move VMware users to OpenStack
• The value of hackathons
• What Intel's Open Source Technology Center is working on
• Why her group contributes everything upstream
• How Intel helps women be successful and keeps them at work
• Why Intel's OpenStack environment was very much like free bananas

For a full transcript of the  interview, click read more below.

Niki Acosta:               Hi Jeff.

Jeff Dickey:                Good morning everyone. I'm Jeff Dickey from Redapt.

Niki Acosta:               Jeff's computer is slightly delayed it seems like, but the show must go on. This is Niki Acosta with Cisco. I think what Jeff was trying to say is, "Hi, I'm Jeff Dickey from Redapt." Right Jeff? We have an awesome guest with us today who has great internet, Ruchi Bhargava, introduce yourself.

Ruchi Bhargava:       Hi, I'm Ruchi. I work for Intel. I've been working there since 1991, and I'm glad to be here, Niki and Jeff.

Niki Acosta:               We're are so glad to have you. It's always a special day for me when we get to interview women in the tech industry, so super stoked to have you, and we typically start out the show by letting you tell us your journey through tech up until what you do now at Intel.

Ruchi Bhargava:       Wow. That can be long, but I'll try to make it short. I grew up in India. I have two elder brothers who are engineers. Typically, my mom never said that that girls should be doctors, but she says, "Oh, I wish one of my kids was a doctor" and that kind of ... In those teenage years, your rebellious behavior comes onboard, and I said you know what, "I think you're just saying this for me because I'm a girl and I want to be an engineer."

Sure enough, maybe she was right. I may have made a better doctor, but I went into engineering school, did an electrical engineering undergraduate degree, and I thought I was going to be a power engineer till I did my internship in the junior year. Guess what? The guy who went with me from school was placed on the shop floor doing switch gear and I was put into the backrooms of the computer data center or the computer room. It was really not a data center.

Writing reports, there were no spreadsheets at that time, I'm dating myself, but I had to create reports. And I may not have learned much about computer science in that internship, but I definitely learned one thing: that I think computer science is the field for me, so that got me focused on looking...

I thought I would go to grad school, get a degree in computer science, but then I got a job offer with an OEM manufacturer in India, and at that time they were the Unix wave, mini-computers wave was on. They basically used to sell computers to enterprises and large organizations and also sell the services of software engineers, so I was one of those field application engineers.

My first job, first project which I think I learned the most so far on both in terms of computer science as well as people skills, etc., was to develop a payroll system for 70,000 employees of a government organization. I think things which you learn to deploy something like that at scale was just brilliant, and that's what got me into technology.

Then, I joined Intel in '91 doing factory automation software, which was more to run the Intel fabs, but develop statistical process control which was in-line so that you could make decision on material whether it was of good quality or not.

In those days, there was no cloud. People had to wait forever for a development and test systems, so that's when I got into infrastructure a little bit. I was more of a databases person, but being part of the middleware team I had to work hand in glove with the people who set up the infrastructure and how to optimize the database is working well, working with the middleware, the message bus and stuff.

That's where I first got into infrastructure and then I moved around a lot at Intel both in location. I lived in all places beginning with P, Phoenix, Portland, Penang, Malaysia. That's one good thing about Intel. You get to learn. If you get bored with one job, you can definitely go and be extremely successful with another job as if you were applying to another company.

I got really diverse experiences in the field of information technology both from infrastructure to application software to operations. When I came back to the United States in 2011, I don't know if you know Das Kamhout was the architect of the open cloud program which he was driving, and they needed somebody to drive the execution for that and that's when I joined him. For the last two years till November, I was part of that team, leading the deployment of OpenStack based cloud, private cloud for us.

Niki Acosta:               Have you seen Das' band play? That's the question.

Ruchi Bhargava:       Yes, I did.

Niki Acosta:               Yeah?

Ruchi Bhargava:       Yes.

Niki Acosta:               Awesome.

Ruchi Bhargava:       Yeah.

Niki Acosta:               Take us back a little bit because I definitely just out of my own curiosity, how was your upbringing in terms of getting technical skills? How was that different in India than it is in the United States? I know it's been a while since you've been in school, but just out of my own curiosity.

Ruchi Bhargava:       At that time, there was nothing like a science fair or it was all about whatever you could do yourself. I grew up in lot of cities in India, and I did my high school in the city called Bombay, which is fairly large metro city. They used to do what they called aptitude tests in eighth grade, ninth grade, and tenth grade, which helped kids go and determine which field you wanted to go into.

I always came out with ... You'll be going into doing something technical, technology focus. I was known at home and in my friends; there is this term called jugaad, which is in Hindi it basically means to make it work. You know if something broke down, I will never read a manual, but I'll just ... I won't even open out things but I'll press a lot of buttons to make things work.

My family always thought she will just make it work. At work too, I may not have the right technical answer, but I kind of always find out. I can break something and then try to fix it back.

Personally, that was what ... I don't think that was what you were asking. You were asking more about how do they guide students into technology, right. Now, from what I hear, there is the same kind of initiatives like the science fairs, the talent searches, science talent searches, which are also happening in India.

I spent some time in Malaysia where my kids grew up and did high school, same thing there. I think the STEM initiative. It's not as rooted as it is in the United States, but it's definitely getting up there.

Niki Acosta:               When you go through these sort of aptitude tests, do you have even option to say like, "Yeah, that test is not accurate. I don't want to do that," and completely switch directions or is it kind of not useful?

Ruchi Bhargava:       Oh yeah. Yeah, yeah. It's basically ... Yeah, lot of people will do what their parents tell them to do because of those stereotypes. If you're a daughter, you are a doctor. If you're son, you can do pretty much anything, but I think those times have changed.

Niki Acosta:               Sure. Growing up as a woman in tech, I'm assuming just like anywhere else that you were probably a minority there. What is that being like for you?

Ruchi Bhargava:       Fortunately, it never felt like a minority. I went to an engineering school in Delhi, Delhi University, and the years prior to us there were like two girls in every batch, and the batch which I joined out of the 260, there were like 30 to 40 girls and that was a lot. I mean that really felt a lot, even though by percentage it is not that much.

We had a group of friends. All the girls generally hung out together and then they were certain ... Of course, we had guys as part of our group, but generally it never felt that way. I'm thinking that was a movement which was going across the country, and so when I joined the workplace, I never felt that.

Even at Intel, I don't feel that we are a minority. There are certain groups. I have been in certain organizations where I'm the only woman on the team, but once you get that part off your head, I think it doesn't matter.

Niki Acosta:               What do you think of all this lean-in stuff?

Ruchi Bhargava:       Lean-in is important. You lean in on everybody. The guys have to lean in too I think, but there is some hard facts about data. There is some hard data there that as women progress through their careers, they tend to drop off, especially at senior grade levels in any company you'll see fewer women. That's something as my daughter who is a high school senior right now, she says, "Hey, we have to smash the patriarchy."

When you first hear it, you'll say what you're saying. She says, you ... In order for the balance at the senior levels to be equal, initially you have to tilt it towards the other direction where men might feel that they are being undermined that they are not getting equal rights but it's basically to keep the balance, and then once the balance has been tilted slightly towards women, ultimately it'll balance and make it equal. I think it is more driven towards equality rather than lean-in towards the women's rights.

Niki Acosta:               Great perspective. Fascinating.

Jeff, I know your WiFi is less than spectacular, and sounds like you had a good night and a good morning; however, I'd like to see if you have any questions to my wonderful co-host.

Jeff Dickey:                I'd like to ... Can you hear me okay?

Niki Acosta:               Yes.

Jeff Dickey:                Okay. I'd like to hear more about kind of what your involvement was with Intel's large OpenStack deployment and some of the issues you guys have had and overcome.

Ruchi Bhargava:       Wonderful. I joined the OpenStack or the private cloud team in 2012, and at that time, the team was starting off with their first external focused ... I mean it was an internal cloud, but for our services business which was exposing those applications to it on the internet rather than just the intranet. The reason why we were targeting on that use case was not so much on the enterprise use case was because they were supposed to be designed for failure and perfect applications for cloud deployment.

However, it turned out that the business needs were more that they needed to be located at different locations where we couldn't service, and the business model itself didn't work out, so we actually lost that customer within Intel.

Since we had made that investment and things were working good, we said okay, let's put to test the enterprise use case, and that's where I think the largest amount of learnings came in. And as you probably are hearing, we have a huge investment as most of the enterprises on virtualization and on VMware, right, so there is a lot of ESX underlying investments which have been made over the period of years and where we had about 20,000 VMs running on that.

Now if I have to ... Either I do a greenfield deployment which is using OpenStack and only put certain use cases or I integrate everything, so our strategy was to put a control plane which was on top of existing infrastructure which will orchestrate the existing infrastructure as well as deploy any new instances; whether it be using VMware or using any greenfield products.

I think what you're seeing now in the marketplace there are several products which are basically doing just the same, but at that time, there was nothing available, so it was a lot of DIY.

Niki Acosta:               What technologies are you using to do that? That's fascinating by the way.

Ruchi Bhargava:       We are using pretty much pure OpenStack.

Niki Acosta:               You have ESX as a hypervisor within your OpenStack control plane?

Ruchi Bhargava:       Right.

Niki Acosta:               Then, do you have any sort of like policies that dictate where your instances get deployed, like how do you make the determination if it goes to an ESX hypervisor or KVM hypervisor for example?

Ruchi Bhargava:       Our KVM hypervisor situation ... The roadmap says we will have those policies. Today, we don't have those deployed. We have a separate instance where we are deploying it to KVM and there is separate instance where we're deploying it to ESX which basically goes into the vCenter. vCenter manages after that. Ultimately, the goal for the IT team was before I left was depending on what the customer needs in terms of if there is a latency requirement, if there is what kind of images that they need and location, we'll basically go and provision in the right underlying infrastructure. The customer doesn't really care whether it's ESX or KVM or Xen or a container.

Niki Acosta:               Are you finding it easy to help transition people who are traditionally VMware users to OpenStack? How are you kind of bridging that gap from sort of one technology set to another inside of Intel?

Ruchi Bhargava:       I think that was a huge challenge, and I'm sure that's a challenge for everybody. Because when people started using VMware, prior to that a lot of people were just using ... At Intel, we still use in our design, where the silicon design teams they use a Grid and that's very DevOps focused, that's very automation focused.

Then when you use VMware, VMware does a lot of it for you, but it also provides a pretty GUI. We'll then put together a front end for it which took us a long time, when did we start doing our VMware journey, so it took us a long time to get to the place where our operations team was much more GUI focused. It was not automation focused.

Getting them to move towards ... When we move to OpenStack, required us to re-skill a whole lot of people from an operations perspective and also changed their mindset because it's very easy to go away from the automation focus to click button and then choose things and somebody else does it for you.

Very common example which we use internally was if the hard disk gets filled up, every night or every three weeks an operations person will get up, they'll do the clean up and then go back to sleep instead of automating that solution, right.

That's the behavior which we have been trying to change. We've been fairly successful, but I think every enterprise who is using VMware will probably or something similar will have to migrate away from that mindset.

Niki Acosta:               What is your automation tool or configuration management tool of choice? Have you guys picked one or are you using multiple?

Ruchi Bhargava:       Well, it's again historical. People initially in the Grid, we were using something homegrown and then they used CFEngine, and then when we started with the OpenStack effort, we were using Puppet not because it was the best or anything but that was something which we just picked.

We did look at over the period of time in different teams who were also looking Chef, Ansible, and also at CFEngine, so a combination thereof.

Niki Acosta:               Yeah, certainly I've seen that large scale teams especially when I was at Rackspace use everything. I mean depending on what component of the public cloud they were working on they certainly had some strong opinions on those. As far as the re-skilling people, how are you re-skilling people? Is it training? Is it just doing it? Is it hiring people? Is it all the above, none of the above?

Ruchi Bhargava:       All of the above. Within IT operations, it's a huge outfit. Everybody I think went through Linux training, everybody went through Puppet. A lot of those operations also went with Puppet training. Everybody went through OpenStack training. They all have been through VMware training and they are learning. Now, they also are being exposed to, how do you design an operate cloud-aware applications.

Then, we drive a lot of hackathons within the company, within the IT organization where if you go through the training and don't use it, that's not going to be of any worth. Having those hackathons has definitely helped the organization.

Niki Acosta:               Real life use cases, hands-on experience, budding up with people who know what they're doing kind of thing?

Ruchi Bhargava:       That's correct.

Niki Acosta:               That's good. I think there is a lot of people out there that are trying to figure out how they make that transition, especially looking at VMware over to OpenStack just because the technologies are so fundamentally different the way that you architect and scale applications.

Jeff, you have any questions?

Jeff Dickey:                Yeah, I want to hear more about what you're doing now. Like what the group's doing and what you're doing.

Ruchi Bhargava:       I moved away from the group after we deployed the product instance of the OpenStack control plane orchestrating the VMware infrastructure. I have now joined the Open Source Technology Center which is within the software services group at Intel, and what that group does as I was telling earlier, it's a 1200-people organization and has been traditionally focused on Linux. When they first started when they were very small, Intel was very contributing at noise levels on Linux. Now, it's one of the largest contributors to Linux.

Our goal in that space was mainly to drive Intel so that to ensure that Linux runs best on Intel architecture. Now, that is I guess at a lower level in the stack, but from an OpenStack perspective or in the private cloud perspective, our goal is to drive higher levels of adoption in the private cloud, be it OpenStack, be it VMware, be it Microsoft as well as so that people who are ... because public cloud is not for everybody. There are requirements for security and some people from cost perspective or existing investments perspective that they'll still be in private cloud scenario.

We have started growing our OpenStack team, cloud software engineering team. I have been growing for the last two months and I know will be doing so in the near future. Our focus has been multifold. One of them is to help meet the needs of the Win The Enterprise efforts of the Foundation.

A lot of the inputs have come from shops like the Intel IT shops and several other members of the Win the Enterprise initiative and things like live migration work, having HA at scale for all the services. Those are standard things which are still not there, the problems of which every enterprise faces; How do you upgrade from one release to the other with zero downtime to existing tenants?

Those are things which we know are problems, and those are the problems which my team has been working on.

Niki Acosta:               How do you determine what Intel contributes back and what sort of Intel keeps as secret sauce? Is there anything [inaudible 00:21:25]?

Ruchi Bhargava:       We don't have any secret sauce in this. My group does everything upstream.

Niki Acosta:               That's great. Are you finding it easy to contribute back up through OpenStack? What is that process like for Intel?

Ruchi Bhargava:       That's definitely has been a struggle for, especially when you hire people who are developers and not existing cores or existing PTLs. To get that influence into the community, one takes time as well as the second part is the review process.

Every now and then I say, "Oh my patch or my proposal just got blocked because there are so many other important things which they have to work on." I think that that is one of the focus areas which the community needs to start looking at how to do differently.

Niki Acosta:               Are you finding it difficult to wield influence inside of OpenStack given that Intel's focus is probably a little bit lower in the stack?

Ruchi Bhargava:       I don't think so. That's not because of the why. I think it's also ... When I'll do the analysis, it's also a matter of numbers; how many people you have in which team. If you have a lot of people focusing on the same, say for example the top three or four companies are HP, Red Hat, Mirantis, IBM, Rackspace; If you look at this, then you can easily correlate the number of people who are there in each of these teams working on it. That's the number of amount of influence.

We've started growing, but we're growing in a very focused manner with the focus on Win the Enterprise and Win the Telco so that there is more adoption. The goal is to drive higher adoption for OpenStack, and by doing that we don't have a solution to sell, so that probably gets us little leeway with the community because we get a lot more support than probably some other vendors if they are proposing something.

Niki Acosta:               Yeah. Definitely, definitely see that. I was doing a write up for this woman in OpenStack Open Mic Series, and one of the questions is which OpenStack debate gets you most fired up?

I think there is a lot of companies. Intel is probably an exception just because as a company you guys have been contributing to Linux for some time, but I think larger companies who have built businesses from intellectual property probably find it a little bit difficult to shift their thinking into this open source model way of doing things and trying to figure out how you monetize, where you monetize, where in the stack you add value, how to be a good community member, how you deal with coopetition. Those are definitely things that I think a lot of companies are struggling with now.

Ruchi Bhargava:       Right. We are definitely ... A lot of our team members across the globe are driving a lot of community efforts. For example, I've got a team in China. They're doing a meetup in Shanghai next week for all of OpenStack. It's not just focused on a particular project, and I think a lot of attention been given to operations and even the enterprise at the operator summit which is happening this week in Philadelphia.

I think with that kind of focus and fairly neutral focus on driving up where the focus is on driving up the adoption of OpenStack rather than selling a particular product, I think that helps a whole lot.

Niki Acosta:               Somebody says to you why Intel for my OpenStack deployment, how do you answer that?

Ruchi Bhargava:       It's not why ... When you say why Intel, are you talking about Intel architect in terms of silicon?

Niki Acosta:               Yes.

Ruchi Bhargava:       There are certain ... We don't have a distribution of OpenStack, so there is no Intel software distribution. Whatever we contribute goes straight to each of the projects within OpenStack, so it will be there in Red Hat's, it will be there in HP's, or any other distribution.

If you want a architecture, definitely what we're trying to do is ... There are several features of Intel architecture, Intel silicon, which we also expose through OpenStack. For example, PCI, SR-IOV, some network optimization features which if you deploy on Intel architecture, Intel servers, then whether you use any vendors OpenStack distribution, you should be able to consume that and expose those.

Niki Acosta:               That's great. It probably has much more mass appeal than some other solutions.

Jeff, you like you want to say something.

Jeff Dickey:                Yeah, you were talking about kind of the accelerating adoption. What are the big things kind of keeping people from adopting OpenStack?

Ruchi Bhargava:       Again, those same three things right. One is operations at scale; things like Live Migration, things which VMware has historically provided a lot of great capabilities in the ecosystem. There is this thing called Evacuate on host failure where you ... There is no such thing which is working today. People can make it work by workarounds or by providing glue, but it's not automatic in OpenStack as yet. It's not a priority right now because people want a stable system right now, right, so that is one aspect of it.

Second is HA of services. Like for example, if you deploy collections using Heat, Heat itself is not resilient and there are several other Heat. If Heat service becomes HA, then it's going to be a great thing for the enterprise.

Then, of course the third part is lot of times people have that myth and rightly so it was the situation earlier that OpenStack is one not mature. You really need to have a large IT organization to deploy OpenStack and all that was fairly true earlier, but now with all these solutions which a lot of vendors are providing where you can have a control plane which will orchestrate existing infrastructure with a new infrastructure, things are I think beginning to shape up really well for OpenStack.

Whether it is you're deploying containers or you're deploying VMware or ESX hypervisors or KVM, I think even if you look at all your podcast recently, there have been several solutions in the marketplace which will meet those needs. I think the future is bright.

Niki Acosta:               What are containers? It seems like every show what we host ultimately containers becoming a topic of discussion. In fact, noting on some of the OpenStack submissions, it look like there were quite a few talks about containers just in general. Where do you see the future of containers and how is that going to change people's ideas and methodologies in how they deploy and migrate applications?

Ruchi Bhargava:       First of all, I'd like to quantify that I'm not an expert on containers, and second from what I have learned of late it definitely provides a lot more, VMs come with little more overhead than containers.

If I have to deploy 200 containers and 200 VMs, it's much much more faster to go and deploy containers today. Now, the other part is it also provides some sort of portability. Because within a container, you don't have any noisy neighbor issue.

Now, the third part which people don't ... I think and I'm not an expert as I said is the security feature. There are security constraints with containers which VMs provide, so definitely this is going to be a mix in the future.

From an enterprise use case, I'm still not convinced that containers can meet each and every use case. Definitely in the web app space, definitely it is viable solution, but maybe in the enterprise use case, it's not quite there, but I'm sure there is somebody working on it.

Niki Acosta:               Are you seeing a lot of these deployments? Most of the deployments that you're learning about are they still primarily greenfield deployments or we starting to see brownfield deployments or attempts to convert applications from their traditional sort of states into a more cloudy state?

Ruchi Bhargava:       I think it's a combination. A lot more. Even this within Intel, Intel IT and our customers, there is a lot more effort going towards any new applications when they develop them. They are doing more resilient application development, and there are definitely groups within Intel who are hosting on Intel cloud who are developing resilient apps, but however there are 80% or 90% of the traditional enterprise apps. I don't think we should force them to redevelop because let's leverage the investment what has already been made. In doing that, there are things, tools which will help you cloudify the existing applications and deploy them in an cloud environment. I think that's where on the application layer and lot of focus is going to be and should be.

Niki Acosta:               I agree. I definitely agree. I mean whoever can figure out how to take a traditional application and seamlessly convert it to make it cloudy, is going to be onto something I think.

Ruchi Bhargava:       True. Business opportunity.

Niki Acosta:               Jeff, you're smiling.

Jef