Ambuj Goyal, a 31-year veteran of IBM, became the company's top storage executive in January of 2013 when he took over as general manager (GM) of IBM's System Storage & Networking business. Goyal's previous roles at IBM included GM of global development and manufacturing for the Systems and Technology Group, GM of IBM information management software, GM of workplace, portal and collaboration software, GM of solutions and strategy for software, vice president of services of software, and director of computer sciences. With his first year as the IBM storage GM winding down, we spoke with him about IBM's flash storage, how he sees the
What were your main goals for improving IBM storage when you took over as GM early this year?
Ambuj Goyal: There were a few things I wanted to change. We had a lot of products. We were doing a battle on speeds and feeds and capacity. What I tried to do was say, 'This is not a storage battle, this is a data battle.' Clients are looking at how they manage data. Their different workloads have different data management needs. When people think about data, the first thing they say is, 'Don't lose my data.' The second thing they say is, 'My data should be available to my workload.' And the third thing they say is, 'When I need capacity or performance, give it to me.' So we changed the focus to data management and that has rationalized the product portfolio.
Does that mean you have to get rid of products?
Goyal: When you get into speeds, feeds and capacity, everything feels like it is overlapping.
Let me give you three scenarios of data management. First is, I have business-critical data that I cannot survive without -- I cannot load my ledger, I cannot report earnings, whatever it is. I have to make sure the data is available and secure. The second scenario is, I have lots of data, let me understand the data and leverage data. Let me start quick and add value -- tell me what I want to keep and what I can throw away and what I can put into cost-effective scenarios and where I need real-time analytics. The third scenario is I have a clutter of data, I have lots of products, and I want to start a new project quickly to leverage all my existing data on depreciated capital.
In the first scenario, which is business-critical data, we lead with multi-copy data management. It's about how you access a created copy of the data [and] how fast you can create a copy. In the past it was called batch processing. Now it's called creating an analytics copy so you can do analytics associated with it. Our DS8000 family is useful for copy management of business-critical data.
In the world of start quick and add value, we have the world's most popular virtualization platform. Originally that was SVC [SAN Volume Controller], and we have significantly improved it over the last three or four years to create things like non-disruptive integration in a data center. So when you put it in it takes a short time for applications to run, and it gives you a huge amount of utilization improvement through data reduction. And this product line is now called Storwize.
In the third scenario where you say have lots of data, the cloud, big data, people are looking for amazing capacity and a grid-scale architecture. I want to make sure I have so many applications running, yet it should automatically self-adjust and provision itself so I don't have to get humans involved. But it still has things like encryption of critical data. That is the XIV family. The XIV family is now the most used in the OpenStack, big data, clouds and analytics scenario.
Our strategy has shifted. We start with a workload, understand what the need for that particular data is, and then lead with a solution rather than speeds and feeds. Now many of our clients have shifted from saying, 'Give me the best dollars per gigabyte,' to saying, 'I want to buy the right data architecture.'
Our sales team has really been well-educated about that, and that's why we are taking competitive share now.
Taking share? I haven't seen any numbers that indicate that.
Goyal: I don't know about IDC or [Gartner] Dataquest, they will publish the numbers. But I look at competitive changes. We are starting to get into many clients where the normal answer was, 'We are standardized on something, show me your value before we will even consider you' and we are starting to displace competitors.
In that sense we are starting to move forward. I'll give you an example. We just went into eBay. EBay is using our big data and cloud solutions for huge amounts of grid-scale data, and that's built on XIV.
Storage is more than 10 Gigabit Ethernet versus Fibre Channel versus 6 terabyte DASD [direct access storage device], or eMLC [enterprise multi-level cell] or SLC [single-level cell] flash. Storage is about data management.
Where does flash fit in?
Goyal: Flash is impacting everything -- flash is not a product for us. Yes, we can sell a standalone product, but flash is leveraged behind Storwize and SVC, all-flash technology is leveraged in multi-copy data management scenarios with the DS8000, and flash is being leveraged in XIV for big data and cloud scenarios, as well.
We use flash to get the fastest ROI [return on investment] without any operational change in the data center. We just announced a DS8000 product line that is all-flash. There's a significant improvement in performance, a significant improvement for clients who want extremely consistent response time. An all-flash DS8000 is good because from an application perspective, the mainframe is using DS8000 and no software needs to change. You can get amazing response times and consistent response times with a reduction in floor space without changing a single line of software. If you roll in a new product with different APIs [application programming interfaces] and different management environments, then you will have to disrupt the data center and ROI will take a long time.
What about the FlashSystem all-flash platform acquired from Texas Memory Systems?
Goyal: That goes into scenarios where clients say, 'I'm already doing things like data backup and replication and all the data loss prevention things in my software. All I want is the amazingly fast capability to access data." In those scenarios we are seeing a huge interest in all-flash.
More than 1,000 organizations in six months have purchased FlashSystems, and we have exceeded 100 petabytes of flash.
What are the challenges your customers are seeing in backup?
Goyal: In backup, they want cheap and deep. For the data they want to access, it needs to be quick. They don't want to depend on a particular media. They say, 'Don't sell me tape, don't sell me flash, don't sell me a separate controller, I have a data management problem. I want to put away petabytes and petabytes of data, and then I want to tell you my recovery point and recovery time objectives, and then you decide the media and give me the cost associated with it.'
There are scenarios where we used to say the answer is a virtual tape library for backup if you only want disk, or real tape, or some other backup software scenario. Now we say [that] through a mechanism we are calling long-term file system, we can use any combination of tape and flash and disk. You tell us the capacity you need, we will give you the lowest capacity for dollars per gigabyte for storage and the best performance based on recovery point and recovery time objectives.
When storing data for archive -- think about for an audit or legal hold -- in those scenarios a combination of flash and tape works nicely. In a media asset management scenario where people need to play out movies, a combination of disk and tape is working nicely.
We don't want to have a separate media business associated with tape, disk or flash. I look at what problem I am I trying to solve, and come up with the right software and media.
What is your cloud storage strategy?
Goyal: There are two ways to think about the cloud. One is being an arms supplier to people who are providing the cloud. Those can be MSPs [ managed service providers], and there are lots of people building private clouds based on our XIV with a cloud option. The second way is through our SoftLayer offering.
Are you seeing any other storage trends in the market?
Goyal: Everything that we are doing is going open. Even when we do tape, the drives we put LTFS on can run on somebody else's libraries. Everything we are doing with respect to OpenStack, Cinder or Swift, we are not going to create proprietary APIs. Many clients say, 'We're stuck with proprietary APIs, and now our application is tied to the API associated with a proprietary vendor.' We want to be in a situation where we endorse open standards and win with execution.
So our strategy is to be open so people are not tied. Even with Storwize and SVC, you can put non-IBM storage behind it. Many of the flash vendors and our traditional competitors enable their storage behind our storage virtualization engine. I can change applications and change the storage associated with it because it's open [source]. I'm trying to endorse everything open. Just like I said flash is pervasive behind every architecture we are doing, open is also becoming pervasive. EBay would have never bought our XIV product without OpenStack support.
We want to win with execution, not by controlling your data center.