The cloud is emerging as having different architectures for different users and workloads.
The cloud is getting increased attention worldwide as a viable resource for IT, and, while there are still security concerns, companies and individuals are beginning to migrate workloads to the cloud. Learning about the new cloud infrastructures is important, and IBM is one of the early adopters, proponents, and solution providers of cloud services. Last month, MC Press Online ran the first half of an interview with Bruce Otte, IBM senior marketing manager, Enterprise Initiatives, Cloud Computing, who explained the different types of IBM cloud models devised to date in the article, The Many Facets of Cloud Computing. Below is the rest of the interview with Otte conducted last August in which he discusses various considerations in adopting cloud services, what is meant by shared services, desktop and "hybrid cloud" models, and what types of organizations have successfully implemented cloud services.
Bruce Otte: A desktop cloud is basically a virtualized desktop environment, and you can have private desktop clouds. It's called a cloud because once you build out the image, it’s the end user who is in charge of what happens—when, where, how—and all you do is maintain that "form" in the background. That’s what makes it a cloud delivery model. Where we’re starting to see some nice traction is with our desktop images on the IBM cloud where it actually is residing in our datacenter, and we’re seeing a mixture—and the reason we call this a shared-services model and not the fully public model—is that in the shared services model, I can also dedicate some physical resources to you.
So, for example, with the recent win with an electronics industry company that we had, they wanted to do a desktop on the IBM cloud, but they needed to be able to segregate the information—segregate their desktops into their environment. So they’re paying for the desktops on a per user basis, but we’re providing them—as part of our contract—that their virtual desktops will only reside, or sit, on a certain set of physical servers in a certain datacenter.
So, we basically can peel off resources that way for certain customers in that shared environment. It's still our assets, and if they’re not utilizing those assets, we can take them back and reallocate them to somebody else—so it’s that cloud environment—but in this case, they’re signing up and saying, we want this many users. As long as they have that many users, those resources will be dedicated to them.
Smith: IBM is creating cloud computing centers everywhere it seems—how many do you have so far?
Otte: We have three main cloud datacenters from a delivery model…that are going to involve production workloads or development and test—two primaries and then a secondary. Our Lotus Live is today running in 27 datacenters around the world; our Information Protection Services, which is a resiliency backup service from the cloud, is today running in just over 50 datacenters around the world. The key is how much business do we have, and how prolific is that getting, and where’s [that coming from]. We also have to date 18 cloud computing centers that are designed purely for cloud research and cloud prototyping.
Smith: I think those were the ones I was thinking of. The backup centers, the 50 datacenters that allow for disaster recovery and so forth—have you had those for some time?
Otte: We have not. They were actually an acquisition [that] we did three years ago, and the company we acquired had two datacenters, and we today have 157 backup and recovery datacenters around the world. So far we’ve built out about 50 of them that are doing the cloud backup and recovery. The reason for that, by the way, is a lot of that data can’t leave a country, and so therefore we have to be able to support in-country data where it’s required and mandated.
Smith: What is the difference between the ones doing cloud backup recovery and the others that are doing traditional backup recovery?
Otte: Well the cloud backup and recovery, you fully configure and enable yourself and manage which servers you’re backing up, and how you’re backing up, where you’re backing up—you can choose physical to physical, you can choose virtual to virtual—you decide how that’s going to be set up; we just provide the raw resources and the portal….It’s completely self contained and self managed, and all we do is charge you on a per-storage-byte how much are you storing on our site. [With] the… traditional, we manage the entire backup recovery process for you. The resources we put there are dedicated to you so you’re paying on a longer term contract basis for those resources that we’ve set up and configured for you…. And it’s a fully managed environment on our …[side].
Smith: Do you assign any variation in security levels between the two?
Otte: We do not. We have the same levels of security between the two. The difference being in the former, with our Information Protection Services, it can be more of a shared environment where you pay a little bit less per byte, or it can be a dedicated environment where you pay a little bit more. That’s up to you then to be able to segregate your information.
Smith: Dedicated meaning physical assets?
Otte: Meaning physical resources dedicated to you.
Otte: Another thing that… I don’t know if you wanted to explore here s this whole concept of hybrid…. The hybrid model is really focused about the end-user service. Cloud computing is about enabling services for the end user, but the service can be something like what I call a business process layer. If I’m talking in consumer terms, it could be booking travel online, it could be doing your banking online or through your mobile phone device. That’s an end user service that business process service.
There’s a company in Japan that is doing a bank payment process service through the cloud and we're doing it for their banks that they have supported directly, but they’ve now got an additional 15–18 banks that are buying that bank payment processing service. They don’t really care about the underlying application, what they care about is that process that you’re getting from the service. But it’s about focusing on the end user, be it raw resources—I want a computer, or I want storage, or be it a process or an application—I want a specific application, or be it what we call "workload"—it’s a collaboration where I want to do analytics or things like that.
The idea of a hybrid cloud is that it delivers that service to the end user. I’m going to use a combination of resources [that] I have in house plus resources from a third party, and I’m going to integrate those two so that the end user has access to both of those capabilities in a single service [that] I’m delivering, and that creates this hybrid model—it’s a hybrid of cloud and traditional.
An example would be a customer resource management application (CRM). Say I’m going to have my traditional CRM that I have today from Oracle or SAP, for example, but I really want to take advantage of accessibility to information. …[I want] my field reps to be able to access it quickly, easily, update it quickly [and] easily and migrate it to no matter where they’re located. So I’m doing to do, [say], Salesforce.com. But I want that to integrate with my internal CRM so that to my sales people, my end users, it gives them a common single experience for managing their customers and how they do it. So I’m going to build a hybrid model that’s going to integrate those two together.
Smith: It's easier than transferring huge amounts of data from one to the other…
Otte: Exactly; it also then keeps you from having to potentially having to transfer huge amounts of data from your CRM to Salesforce.com so your sellers can access it. So by pulling these two things together….That type of model, that hybrid model, is really where IBM has a big focus, and is why IBM acquired Cast Iron Systems, because we think they have a very strong engine that’s all about integrating traditional with cloud-delivered services to pull together these hybrid models.
Smith: I can see why that could be huge.
Otte: In fact, IDC says that they think that between now and 2015 that’s going to be the fastest growing area of the cloud market, this hybrid.
Smith: What about security for that; I would think that would be even harder to secure.
Otte: Exactly. That becomes much harder to secure, and so therefore you have to pay attention to the key elements around cloud security, which are going to be…the obvious one is your vulnerability element. You have to make sure that between you and your cloud and the network in between that you’ve got strong vulnerability management in place….you avoid those attacks from outside, but you also have to be careful about the attacks from inside. You’ve got to have a much stronger role in place for your identity and your access management, and you’re also going to have to make sure that you maintain that auditability—particularly when you’re talking about customer data, because for a lot of companies, that is the crown jewel.
Smith: Are there hardware preferences in terms of the various cloud models?
Otte: It depends on the vender you talk to. From an IBM perspective, cloud is a delivery philosophy so the hardware shouldn’t matter. The idea behind the hardware should be if you want to leverage the utilization of the hardware, either the cloud service has to consume that hardware or you’re going to want to be able to virtualize that hardware effectively and efficiently so that you can share it across multiple services, or multiple users. From an IBM perspective, all of the compute platforms that we work with, be it the mainframe, System z, …Power, which is a UNIX-based system, or…our System x, your x86 type of compute, all of those can very easily be virtualized effectively and doing so in a very secure way.
[The same is true] for storage—you want to make sure that the storage platform you’re working with accommodates virtualization, whether it's solid-state storage, disk, tape, or even to some degree your archive. What are your systems that we're archiving? Can they also be virtualized to leverage the utilization and accommodate that shared environment?
There are other venders out there pushing, and…leaving the impression that cloud is going to be an x86 platform only. Well, that’s because it’s a very common platform and it’s very inexpensive to get into, but it may not be the best platform for the workload that you want to run. If you’ve got certain analytics, for example, that you want to drive, it can take you up to two weeks to do something on an x86 system that you can get accomplished in a matter of a half a day on a UNIX system. So, yes that x86 is being [fully] utilized, but it’s also being tied up that entire time so you’re not… able to share it with other…other jobs, other projects…. If you migrated [the job] to a UNIX platform, yes it’s going to use those resources for half a day, but then you can reallocate them. So our feeling is it should be platform-independent and the platform used should be based upon the workload and volume for that workload.
Smith: Is cloud computing creating certain needs within the storage hardware environment?
Otte: Absolutely. The biggest need it’s creating within the storage environment is… for… what I would call self-managed, self-control in virtualizing storage. There’s been a lot of work over the past two or three years to be able to virtualize storage, and that is typically by putting it in front of a software appliance that basically does the virtualization for you. It maximizes the utilization of the physical assets, and you, as the end user, may or may not have visibility to where [your] data is being stored. Because, in fact, your pieces of data get scattered over a number of devices based on how the software desegregates it, and then reintegrates it when it’s bringing it back to you.
Cloud is introducing the fact that as an end user you may need to have visibility to where it’s going to get stored, so this software now has to be able to not only virtualize the data but map to where it’s going and allow you to self-guide whether or not it can map across those several devices, or if you need to have it go to a specific device.
So, we’re seeing a lot of change in that virtualization layer...The second piece that’s being driven is a number of companies—(and IBM being one of them) within their service management software where we are managing the environment for the customers—are starting to have to build in stronger links between the service management of the servers with the service management of the storage such that some of this virtualization of the storage is not only driven by the end user but can be automated based upon the types of data an end user is using.
So a company can say yes, we can have virtualized storage, but for these specific types of data it can’t be virtualized, or it can but needs to go to physical server A, B, or C. So you want to have that stronger integration between your server management software and your traditional storage management software.
Smith: Is that having a domino effect on data types?
Otte: It’s definitely having a domino effect on data types, and it’s driving up the importance of data type tagging.
Smith: I would imagine that this is also creating some kind of career opportunities or definitions that trickle down to the university level?
Otte: Absolutely, that's very big, and in fact we just awarded a project to the University of Missouri …that's [in] this specific area. If you want an example of a privately built shared services cloud it would be the University of Bari in Italy that literally built out a shared or community services cloud specific to fishermen and farmers and trucking companies, the distributors—in and around the city of Bari.
Smith: You wouldn’t expect those industries to jump into this first off
Otte: You would not, but what they’re trying to do, and what they’ve created, is an e-marketplace for the fishermen such that when they’re out on their boats they literally can log right into the cloud and see if there is a market for the fish they are getting ready to catch or that they have just caught. And by the time they pull into dock, the trucks can be ready to take the catch right from them, and [meanwhile] they’ve already sold it to the seller, and the truck driver already knows exactly where he’s headed.
Since this interview was conducted, IBM has issued an announcement about some new security measures it has taken to address users' concerns in this area. That announcement can be read here.
LATEST COMMENTS
MC Press Online