What Drives VMware CTO Stephen Herrod? Customers, Challenges, The Cloud

A Q&A with the tech chief on what keeps the company looking above the cloud.

Virtualization Review editor in chief Bruce Hoard recently interviewed VMware CTO Stephen Herrod on a wide range of topics, including challenges faced by VMware customers, its growing expertise in application development, and the role of virtualization in VMware product planning.

Virtualization Review: What are the biggest problems that customers face when they virtualize their data centers?

Herrod: Do you mean if they are starting from scratch?


Herrod: We've tracked this for many, many customers for many, many years now, and the challenges have changed a fair amount as the product has matured and people have gotten it in place, but I do think that the single biggest challenge when you go into a big virtualization product is you do need to prepare for this. One gentlemen told me this, and the quotation was pretty funny. He said "If you're bad at IT, you're going to be really bad at virtualization," meaning if you don't have any controls in place, and you don't understand what it means to have things that are faster moving than they were in the past, you can get in trouble, so I do think expecting it not to have you think about what's going on is one of the mistakes people make. And again, you can move a lot faster when you're virtualized, so things can get amplified if you're not putting the controls in place there.

How are you making it easy for customers to use vSphere 5?

Herrod: We announced two big things as part of vSphere 5. One is called VMware Go. VMware go is a hosted service. It lets customers who have never heard about virtualization and don't have an IT staff get up and running as quickly as possible. Some of these companies have the needs of a big enterprise IT department for security to availability, but they don't have the skill sets, the people or the money to do that, so VMware Go is a very simple tool. You literally log into a web page, and with your permission, it finds out what hardware is in your environment. It then will automatically deploy the ESXi hypervisor directly to the hardware that it finds, and it lets you manage and run it all from that same web page. It's not a full suite for management, but if you just want to get up and running for the first time very simply, that's the way that we're seeing more and more of our customers doing it. And over time, we're going to be adding some pretty interesting things to VMware Go, like very simple management tools that let you patch your virtual machines.

Please describe the vSphere Storage Appliance.

Herrod: The vSphere Storage Appliance is very interesting from a technology and a business standpoint. What it does is takes the individual disk drives in your servers and clusters them together to create a virtual SAN in many ways, so it creates a nice little cluster using the individual drives in the server and presents them as if it was a Fibre Channel array or a network attached storage. That's the technology angle, but from a business point of view, that's what enables things like our vMotion technology and high availability. So it's a way to bring to the very small companies these enterprise class features that they haven't had in the past.

Was that a largely customer-driven product? There's such a hue and cry about all the problems associated with storage.

Herrod: It was definitely very obvious from the customer side and from the technology side that you could do this, but if you got to all our small customers and you asked them "What are your challenges moving forward,?" storage is at the top of the list, both from a cost and complexity standpoint. This attacks both of those challenges, using the existing things you have, which is less expensive, and it is dead simple to install and run. You literally put a CD into one of the servers, and it finds everything and just pulls it nicely together.

How do you measure the quality of competing products?

Herrod: Candidly, we don't focus on them too much. We obviously do careful performance analysis, making sure we maintain our lead there, but we're so far ahead on the types of things that we do in our approach that there's no comparison that we would make there. One of my favorite stories is about a set of customers that keeps sending us screenshots of their management tool, vCenter, and it shows the uptime of their ESX servers, and we literally have several that have ran over a thousand days without having to reboot. Another one of my favorite stories is about a company that was sending us updates month after month saying, "It's still up, here is the screen shot," and then they sent me a note that said, "It's down now," and all of our team groaned, but we found out they had to take it down because accounting rules dictated that the server had to be retired if it was more than three years since they bought it.

Back on a serious note, we have to really focus on testing scale, and testing the highest in workloads that no one else is even a candidate to run right now, so we have a lot of innovation in our testing processes, and we have this massive lab for doing it. There is as much there as we have on the development side.

How long have you personally been working on developing vSphere and what lessons have you learned during that process?

Herrod: I've been at VMware almost 10 years fulltime, and have been working in the vSphere area the whole time. I continue to be impressed by the focus on quality. That's easy to say, but until you've see it in person, you can't really appreciate it, and very early on we stopped many, many releases and held them up because we weren't sure if there might be a problem. We've actually done this to this day--we have very strict requirements you have to satisfy, and it has paid off quite well. People just trust that it's going to run, and especially for virtualization as it's used for consolidation, they're saying that if you put all your eggs in one basket, you had better have a very good basket, and from day one the company has been focused on making sure it will run, no matter what, and we would prioritize that over adding new features.

Was View 4.5 an example of that? You just didn't feel like it was ready to go at the time?

Herrod: We have different goals in different product lines, and they all have different things that you have to prioritize. For vSphere specifically, it has to be the absolute gold standard for everything. Nothing can crash there, because that's where you get in trouble. For all the things that we're building around it, quality is extremely high as well, but sometimes people are trying to trade off other capabilities. View 4.5--and we recently launched View 5--had a very, very high quality bar. What you also have to focus on very much in the end user space is security, so I would lump those two together. Particularly with View 4.5 and View 5 we had just a massive investigation looking at every possible security hole there. It's not something you can tout as a product feature, but if it's not there, and someone gets access to your desktop and your data, you definitely hear about it.

Down the line, how has VMware benefited from the acquisition of SpringSource?

Herrod: That's been a fun one. Paul [Maritz] speaks quite a bit about this. Buying SpringSource was a very big move for VMware and it moved us into the discussion about how you write the next generation of applications. Spring is a very simple way of writing Java programs, and it got us in the game of simplifying applications. We've come a very long way since that acquisition. We've made I think at least five other acquistions in that space, and really a lot of this is culminating in Cloud Foundry. So Spring really represents the beginning of us entering into this more general space, and it has been great for modernizing Java applications. It really taught us that there are a lot of people really wanting a simpler way to write the next generation of applications, and we knew that we had to go all-in on open source for this to be successful. This is especially true now, because there is so much change going on, and so many new frameworks.

Micro Clouds is one of my favorite new products, and it is the entire Cloud Foundry Cloud, but it's on a USB stick, and you can have it running on your laptop, and you get a very micro Micro Cloud. You write an application there and it works exactly the same as if you had deployed it to a giant cluster in a giant cloud. So Spring was the launching point from which we really started thinking how applications will work in the future.

How hard will it be for even the most advanced VMware customers to digest VMware's cloud infrastructure suite? VMware customers as a group are very advanced, but even as advanced as they are, this seems like a lot to digest.

Herrod: We've been testing if for quite a while with customers, and certainly we expect a world where you can very easily have a mix of vSphere 4 and vSphere 4 with vSphere 5, and we've designed all our management tools to allow that to happen. We also spend a lot of time looking at the migration path and we are making it very easy to vMotion from one version to the other, and to convert the file system and the other pieces. So we've been through a lot of training on allowing it to be an easy upgrade, and the customers we have talked to so far have seen enough value in the storage pooling and in the new version of vShield, and in the new management tools, that they'll take that upgrade path if it can be non-disruptive, and that is what we've focused on there.

Do you feel like you develop technology that customers grow into?

Herrod: That's a really good question. I would certainly say that our goal is to be very much ahead of where things are headed, and make sure we have a good proven product when we get there As it is now, we have some very, very aggressive customers who would like to see vSphere 7 today, and they have their list of features, so we're appealing to a quarter of a million different customers and companies with this product. Some of them are happy using basic consolidation and many of them want the largest VMs with the highest availability, and everything else you can throw into it.

What do you think about the growing trend of mainframe clouds?

Herrod: I actually haven't seen that trend. Are you seeing that?

Yes, IBM's doing it.

Herrod: There are a lot of things I'm not quite sure about, but one thing I am sure about is that the true clouds that are at scale are going to be based on industry-standard parts, and that the economies and the performance just outweigh everything else in sight. I believe there's a world where obviously there are legacy applications that you're not going to move, but any cloud will be built on x86 processors and based on industry-standard software and hardware, I'm quite confident in that.

At any point do you believe that VMware was putting too much emphasis on virtualization?

Herrod: I believe that virtualization has been a great tool and a great way to get started. It's not the end state, and in fact you should think of virtualization as a how, not a what. It is a means to an end. Our ultimate goal is to provide very efficient and available IT, and this has been one tool in an arsenal to do that. It's a very good tool, and one that we continue to invest in, but we've done so much more beyond the hypervisor. With 4,000 engineers, we probably have 900 on that core layer. The others are working on the management tools, the portals, the ways of accessing that.

The other thing that has become very clear is that there is no cloud without virtualization. Even if it's not a VMware cloud, it all starts with virtualization, because it is that first step where you separate software from the underlying hardware. That is how multi-tenancy comes up, that is how mobility comes up, along with scale and efficiency, so it is a core ingredient and we're doing that, but we're also doing the other pieces.

Will VMware ever offer a Type 1 client side hypervisor?

Herrod: I would never say never, but I can say that we've flirted with that several times, and in fact we even talked about it at VMworld events in the past. What I've found is that the speed at which new types of peripherals and hardware changes--tablets, phones and PCs--is very, very aggressive, which requires aggressive new moves in power management and other things such as docking stations, and all this diversity makes it very hard to keep up with those changes, and support the very latest in desktop hardware. As a result, if you're doing a true bare metal device, you're taking on ownership of ACPI tables and docking stations, the latest graphics cards and everything in there.

What we found is by doing a type two approach, we're able to provide the core value virtualization, which is isolation of an environment, but we can allow the normal development of those device drivers. We started on laptops doing a type one hypervisor, but we found out we don't need that. We can get the security and the portability we're after without it. The mobile phone virtualization that we recently announced actually started as something running bare metal on the phone, but now it's something that installs on top of Android. The thing that ultimately matters isn't the architecture, it's can it provide the security and isolation, and the other capabilities that you want ultimately to deliver there?

Talk to me a little bit about Project Horizon, and the impact you see that having down the line.

Herrod: Horizon Application Manager is the product in the market that lets me as an IT department broker access for my users to software-as-a-service applications--Saleforce.com, g-mail--those sorts of applications. We think it should be a universal broker, and it should be just like the switchboard operator of old. It should be the thing that connects me with all of my applications and my data, so think of a world where there are applications running on a desktop, where some are running in a backend public cloud service, and where some are just normal Windows applications. Horizon sits right in the middle of those, and the user who ultimately wants to access them. So think of it as a big switchboard operator that's going to broker access to all of your apps, and this is going to be the insertion point for some of the critical IT policies--who can access what when, how do I track data?--all those really difficult things that we're dealing with right now.


Subscribe on YouTube