Tuesday, September 01, 2009

XDAS Standard Aims to Empower IT Audit Trails from Across Complex Events

Transcript of a sponsored BriefingsDirect podcast on an emerging standard aimed at easing governance and compliance in heterogeneous IT environments. Recorded at The Open Group's 23rd Enterprise Architecture Practitioners Conference and 3rd Security Practitioners Conference in Toronto.

Listen to the podcast. Find it on iTunes/iPod and Podcast.com. Download the transcript. Learn more. Sponsor: The Open Group.

Dana Gardner: Hi, this is Dana Gardner, principal analyst at Interarbor Solutions, and you’re listening to BriefingsDirect.

Today we present a sponsored podcast discussion, coming to you from The Open Group’s 23rd Enterprise Architecture Practitioners Conference and the associated 3rd Security Practitioners Conference in Toronto.

We're going to take a look at an emerging updated standard called XDAS, which looks at audit trail information from a variety of systems and software across the enterprise IT environment.

This is an emerging standard that’s being orchestrated through The Open Group, but it’s an open-source standard that is hopefully going to help in compliance and regulatory issues and in the automation of heterogeneous environments. This could be increasingly important, as we get deeper into virtualization and cloud computing.

Here to help us drill into XDAS (see a demo now), we're joined by Ian Dobson, director of the Security Forum for The Open Group. Welcome, Ian.

Ian Dobson: Hello.

Gardner: We're also joined by Joël Winteregg, CEO and co-founder of NetGuardians. Welcome, Joel,

Joël Winteregg: Hello.

Gardner: First off, not that many people are familiar with the audit trail issue. We've, of course, heard a lot about log files over the years, and the information from variety of systems in IT. What is the problem set that we're working on and why did The Open Group get involved, Ian?

Dobson: We actually got involved way back in '90s, in 1998, when we published the Distributed Audit Service (XDAS) Standard. It was, in many ways, ahead of its time, but it was a distributed audit services standard. Today’s audit and logging requirements are much more demanding than they were then. There is a heightened awareness of everything to do with audit and logging, and we see a need now to update it to meet today’s needs. So that’s why we've got involved now.

A key part of this is event reporting. Event reports have all sorts of formats today, but that makes them difficult to consume. Of course, we then generate events so that they can be consumed in useful ways. So, we're aiming the new audit standard from XDAS to be something that defines an interoperable event-reporting format, so that they can be consumed equally by everybody who needs to know.

The XDAS standard developers are well aware of, and closely involved in, the related Common Event Expression (CEE) standard development activity in Mitre. Mitre's CEE standard has a broader scope than XDAS, and XDAS will fit very well into the Event Reporting Format part of CEE.

We are therefore also participating in the CEE standard development to achieve this and more, so as to deliver to the audit and logging community an authoritative single open standard that they can adopt with confidence.

Gardner: Joël, tell me a little bit about why you got involved. What was the problem that you identified that needed to be improved?

Single standard is easier

Winteregg: My company is working in the area of audit event management. We saw that it was a big issue to collect all these different audit trails from each different IT environment.

We saw that, if it was possible to have a single and standard way to represent all this information, that would be much easier and relevant for IT user and for a security officer to analyze all this information, in order to find out what the exact issues are, and to troubleshoot issue in the infrastructure, and so on. That’s a good basis for understanding what's going on the whole infrastructure in the company.

Gardner: As it stands now, audit information comes across helter-skelter. There isn’t a single way. It's dependent upon the vendor, the actual device, and/or the software.

Winteregg: Exactly. There is no uniform way to represent this information, and we thought that this initiative would be really good, because it will bring something uniform and universal that will help all the IT users to understand what is going on.

Gardner: Also, there is currently very little emphasis on the analysis of this audit trail information. Most of the solutions that are available are just simply to harness and collect it.

Winteregg: Yes. There is a lot of effort spent on collecting and then normalizing all this information, while the most important effort, the analysis of this audit trails, is left behind, because it takes so much effort to understand these trails.

If you take, for example, logs from Cisco, Nortel, SAP, and so on, each different vendor is using another language. It is like understanding French audit trails, Chinese audit trails, or German audit trails. There is no uniform way to provide this information.

Then, for auditors or administrator, it is really costly to understand this information and use it

You will be able to track the who, the what, and the when in the whole IT infrastructure, which is really important these days . . .

in order to get relevant information for management to have metrics and to understand what's really happening on the IT infrastructure.

Gardner: Why is this different from log information? The audit information is something that tells us about what's going on within an event, for example?

Winteregg: Audit information deals a lot with the accountability of the different transactions in an enterprise IT infrastructure. The real logs, which are modulated to develop strong meaning for debugging applications, may be providing the size of buffers or parameters of an application. Audit trails are much more business oriented. That means that you will have a lot of accountability information. You will be able to track the who, the what, and the when in the whole IT infrastructure, which is really important these days with all these different regulations, like Sarbanes-Oxley (SOX) and the others.

Gardner: So, those folks who have to comply with regulations -- maybe it’s the payment card industry, or specific regulations for specific industries -- need to create this audit trail. Right now, it’s expensive, and the XDAS standard is designed to simplify and automate that.

Complying with regulations

Winteregg: Exactly, because each IT user has to define how they will collect this information in order to comply with all these regulations. For example, the banking industry has Basel II or SOX, which have a big impact on auditing and accountability management. Each company, each bank, has to deal with its own defined strategy to analyze these trails, to collect them, or to store them.

With a standard like XDAS, it will be much easier for a company to be in compliance with regulations, because there will be really clear and specific interfaces from all the different vendors to these generated audit trails.

Gardner: And this is an open-source standard, so it’s under the Lesser General Public License (LGPL). Is that correct?

Winteregg: Yes. The standard will be open, but there is a Java implementation of that standard called XDAS for J, which is a Java Library. This implementation is open source and business friendly. That means that you can use it in some proprietary software without having to then provide your software as an open-source software. So, it is available for business software too, and all the code is open. You can modify it, look at it, and so on.

Gardner: This is available for examination and download at Codehaus. Is that correct?

Winteregg: Yes. It’s on the Codehaus platform.

Gardner: Why is this important, as we move toward heterogeneity that spans not just systems

In distributed environment, it's really hard to track a transaction, because it starts on a specific component, then it goes through another one, and to a cloud. You don’t know exactly where everything is happening.

but sourcing, for example, cloud, a supply chain, or software as a service (SaaS)? Compliance still needs to be adhered to and regulations need to be complied with. Yet, many of these systems are no longer under your roof.

Winteregg: In distributed environment, it's really hard to track a transaction, because it starts on a specific component, then it goes through another one, and to a cloud. You don’t know exactly where everything is happening. So, the only way to track these transactions or to track the accountability in such an environment would be through some transaction identifiers, and so on.

Collecting all the different logs from all the different components of a cloud is really useful, because you collect everything in a single point and then you have all this information available for analysis and correlation. So, you can correlate maybe a transaction ID between all the different transactions.

Then, you can drill down into this information to track the whole transaction without having to connect to each different component of the cloud. So, it's really useful to remotely collect this information in order to enhance all the accountability aspects of this computing method.

Gardner: Of course, it's going to grow more important. What about in a virtualized environment, where perhaps you're still inside of your own IT organization, but you've got virtualized instances of applications and services? Sometimes, those come and go, depending on the elasticity and efficiency that you're seeking. Logging and auditing also perhaps would disappear. Is this something that can be useful in the context of a highly virtualized environment?

Similar to cloud

Winteregg: Yes, that’s a similar context to the cloud-computing environment. We had an example like this at Geneva State in Switzerland, where the SAP system was moving around to several different instances. Sometimes, the service is on specific machine and a minute later, it's on another machine.

All the different instances will be sending this information to a place where you can analyze it through, maybe, user names. You don’t really care at the end exactly where the transaction or the processing happens. You only care about collecting the information and then analyzing all of this in a single point. So, there's less effort spent on collecting each different point of this information, because everything is already into a single box, a single place.

Gardner: Please tell me where are we in terms of the maturity of this XDAS standard? Is this something people can use already? What additional work and/or acceptance does this need to go through before it’s enterprise ready?

Winteregg: The standard was mainly done by people from Novell, like David Corlette or John Calcote, who are involved into defining the standard. It is at a draft stage right now. It is available for consultation and for feedback as a draft, but as we think that pragmatic approach is much more efficient in the definition of such a standard.

That’s why, even if it’s only a draft, we've started to already develop an open-source library, like

We believe that having such a tool before the standard is strongly defined will help in enhancing all the different aspects of the standard.

XDAS for J, which enables IT users and developers to try to include this library into their testing program or business application, in order to get audit trails in a good and understandable format. We believe that having such a tool before the standard is strongly defined will help in enhancing all the different aspects of the standard.

Gardner: What about the role of the vendors, the suppliers of these devices and software and appliances? What do they need to do in order to make this standard more pervasive?

Winteregg: The best thing would be to have some feedback about how easy it is to use and how easy it is to understand or if there are some use cases that use the standard. We started another pragmatic approach, based on the Agile development process of software development, which is made up of use cases and test-driven development.

Through these different iterations, we’ll bring a more efficient standard. So, we're waiting for some feedback from vendors and users about how it is easy to use, how helpful it is, and if there are maybe some use cases -- if the scope is too wide, too narrow, etc. We're open to every comment about the current standard.

Gardner: Well, great. We've been learning about an audit trail standard that’s emerging. It's called XDAS, and we certainly encourage people to take a look at it as a way of adhering to compliance in complex environments and across virtualized and cloud and extended enterprise activities.

We've been joined in our discussion here by Ian Dobson. He is the director of the Security Forum for The Open Group. We've also been joined by Joël Winteregg, CEO and co-founder of NetGuardians. Thank you, Joël.

Winteregg: Thank you.

Gardner: This is Dana Gardner, principal analyst at Interarbor Solutions, and you've been listening to a sponsored BriefingsDirect podcast from The Open Group’s 23rd Enterprise Architecture Practitioners Conference and the associated 3rd Security Practitioners Conference here in Toronto. Thanks for listening, and come back next time.

Listen to the podcast. Find it on iTunes/iPod and Podcast.com. Download the transcript. Learn more. Sponsor: The Open Group.

Transcript of a sponsored BriefingsDirect podcast on an emerging standard aimed at easing governance and compliance in heterogeneous IT environments. Recorded at The Open Group's 23rd Enterprise Architecture Practitioners Conference and 3rd Security Practitioners Conference in Toronto. Copyright Interarbor Solutions, LLC, 2005-2009. All rights reserved.

Monday, August 31, 2009

Cloud Adoption: Security is Key as Enterprises Contemplate Moves to Cloud Computing Models

Transcript of a sponsored BriefingsDirect podcast on the state of security in cloud computing and what companies need to do to overcome fear, reduce risk and still enjoy new-found productivity.

Listen to the podcast. Find it on iTunes/iPod and Podcast.com. Download the transcript. Learn more. Sponsor: Hewlett-Packard.

Free Offer: Get a complimentary copy of the new book Cloud Computing For Dummies courtesy of Hewlett-Packard at www.hp.com/go/cloudpodcastoffer.

Dana Gardner: Hi, this is Dana Gardner, principal analyst at Interarbor Solutions, and you’re listening to BriefingsDirect.

Today, we present a sponsored podcast discussion on caution, overcoming fear, and the need for risk reduction on the road to successful cloud computing.

In order to ramp up cloud-computing use and practices, a number of potential security pitfalls need to be identified and mastered. Security, in general, takes on a different emphasis, as services are mixed and matched and come from a variety of internal and external sources.

So, will applying conventional security approaches and best practices be enough for low risk, high-reward cloud computing adoption? Is there such a significant cost and productivity benefit to cloud computing that being late or being unable to manage the risk means being overtaken by competitors that can do cloud successfully? More importantly, how do companies know whether they are prepared to begin adopting cloud practices without undo risks?

To help us better understand the perils and promises of adopting cloud approaches securely, we're joined by three security experts from Hewlett-Packard (HP). Please join me in welcoming Archie Reed, HP Distinguished Technologist and Chief Technologist for Cloud Security. Welcome, Archie.

Archie Reed: Hello, Dana. Thanks.

Gardner: We're also joined by Tim Van Ash, director of software-as-a-service (SaaS) products at HP Software and Solutions. Welcome, Tim.

Tim Van Ash: Good morning, Dana.

Gardner: Also, David Spinks, security support expert at HP IT Outsourcing. Welcome, David.

David Spinks: Good morning.

Gardner: Of course, any discussion nowadays that involve cloud computing really deserves a definition. It's a very amorphous subject these days. We're talking about cloud computing in terms of security and HP. How do you put a box around this? What are the boundaries?

Van Ash: It's a great question, Dana, because anything associated with the Internet today tends to be described as cloud in an interchangeable way. There's huge confusion in the marketplace, in general, as to what cloud computing is, what benefits it represents, and how to unlock those benefits.

Over the last two years, we've really seen three key categories of services emerge that we would define as cloud services. The first one is infrastructure as a service (IaaS). Amazon's EC2 or S3 services are probably some of the best known. They're there to provide an infrastructure utility that you can access across the Internet, and run your applications or store your data in the cloud, and do it on a utility-based model. So, it's a pay-per-use type model.

If we look at platform as a service (PaaS), this is an area that is still emerging. It's all about building applications in the cloud and providing those application-development platforms in the cloud that are multi-tenant and designed to support multiple customers on the same platform, delivering cost efficiencies around development, but also reducing the amount of development required. Many of the traditional tiers from data persistency and other things are already taken care of by the platform.

The last area, which is actually the most mature area, which started to emerge about 10 years ago, is SaaS. Great examples of this are Salesforce.com, HP's partner NetSuite, and, obviously, HP's own Software-as-a-Service Group, which delivers IT management as a service.

Gardner: When we're talking about applying security to these definitions, are we talking about something very specific in terms of crossing the wire? Are we talking about best practices? Are we talking about taking a different approach in terms of a holistic and methodological understanding of security vis-à-vis a variety of different sources? Help us better understand what we mean when we apply security to cloud.

Different characteristics

Van Ash: Once again, it's a great question, because you see very different characteristics, depending on the category of the service. If it's IaaS, where it's really a compute fabric being provided to you, you're responsible for the security from the operating system, all the way out.

You're responsible for your network security, the basic operating system security, application security, and the data security. All of those aspects are within your domain and your control, and there really is a large difference between the responsibility of the consumer and the responsibility of the provider. The provider is really committing to providing a compute fabric, but they're not committing, for the most part, to provide security, although there are IaaS offerings emerging today that do wrap aspects of security in there.

For PaaS, the data persistency and all those elements, for the most part, are black box. You don't see that, but you're still responsible for the application-level security, and ensuring that you're not building vulnerabilities in your code that would allow things like SQL injection attacks to actually mine the data from the back-end. You see more responsibility put on the provider in that environment, but all the classic application security vulnerabilities, very much lie in the hands of the consumer or the customer who is building applications on the cloud platform.

With SaaS, more of the responsibility lies with the provider, because SaaS is really delivering capabilities or business processes from the cloud. But, there are a number of areas that you're still responsible for, i.e., user management in ensuring that there are perfect security models in place, and that you're managing entry and exit of users, as they may enter a business or leave a business.

You're responsible for all the integration points that could introduce security vulnerabilities, and you're also responsible for the actual testing of those business processes to ensure that the configurations that you're using don't introduce potential vulnerabilities as well.

Gardner: Archie Reed, it sounds as if there is a bigger task here. We had to evaluate whether the provider has instituted sufficient security on their end. We have to be concerned about what we do internally. It sounds like there is a larger security wall to deal with here. Is that the case when we look at cloud?

Reed: Absolutely. One of the key things here is, if you take the traditional IT department perspective of whether it's appropriate and valuable to use the cloud, and then you take the cloud security's perspective -- which is, "Are we trusting our provider as much as we need to? Are they able to provide within the scope of whatever service they're providing enough security?" -- then we start to see the comparisons between what a traditional IT department puts in play and what the provider offers.

For a small company, you generally find that the service providers who offer cloud services can generally offer -- not always, but generally -- a much more secure platform for small companies, because they staff up on IT security and they staff up on being able to respond to the customer requirements. They also stay ahead, because they see the trends on a much broader scale than a single company. So there are huge benefits for a small company.

But, if you're a large company, where you've got a very large IT department and a very large security practice inside, then you start to think about whether you can enforce firewalls and get down into very specific security implementations that perhaps the provider, the cloud provider, isn't able to do or won't be able to do, because of the model that they've chosen.

That's part of the decision process as to whether it's appropriate to put things into the cloud. Can the provider meet enough or the level of security that you're expecting from them?

Suitable for cloud?

The flip side of that is from the business side. Are you able to define whether the service value that's being provided is appropriate, and is the data going into the cloud suitable for that cloud service?

By that, I mean, have we classified our data that is going to be used in this cloud service regardless of whether it's sitting in a PaaS or SaaS? Is it adequately protected when it goes into the cloud, such that we can meet our compliance objectives, our governance, and the risk objectives? That ultimately is the crux of the decision about whether the cloud is secure enough.

Gardner: Let's go to David Spinks. It sounds as if we almost fundamentally need to rethink security, because we have these different abstractions now of sourcing. We have to look at access and management control, what should be permeable and perhaps governed at a policy level across the boundaries.

I suppose there are also going to be issues around dynamic shifting, when processes and suppliers change or you want to move from a certain cloud provider to another over time. Do you think it's fair that we have to take on something as dramatic as rethinking security?

Spinks: That's absolutely right. We've just been reviewing a large energy client's policies and procedures. While those policies, procedures, and controls that they apply on their own systems are relevant to their own systems, as you move out into an outsourcing model, where we're managing their technology for them, there are some changes required in the policies and procedures. When you get to a cloud services model, some of those policies, procedures, and controls need to change quite radically.

Areas such as audit compliance, security assurance, forensic investigations, the whole concept of service-level agreements (SLAs) in terms of specifying how long things take have to change. Companies have to understand that they're buying a very standard service with standard terms and conditions.

Before they were saying, "Our systems have to comply with this policy, and you have to roll out patches." In a cloud services environment, those requirements no longer apply. They have very standard terms and conditions imposed on them by the cloud providers.

Gardner: So, while we need to think out how we approach cloud, particularly when we want a high level of security and a low level of risk, the rewards for doing this correctly can be rather substantial.

Tim Van Ash, what are the balances here? Who is in the role of doing the cost-benefit analysis that can justify moving to the cloud, and therefore recognize the proper degree of security required?

Pressure to adopt

Van Ash: It's a very interesting question, because it talks to where the pressures to the adoption of cloud are really coming from. Obviously, the current economic environment is putting a lot of pressure on budgets, and people are looking at ways in which they can continue to move their projects forward on investments that are substantially reduced from what they were previously doing.

But, the other reason that people are looking at it is just agility, and both these aspects – cost and agility -- are being driven by the business. Going back to the earlier point, these two factors coming from the business are forcing IT to rethink how they look at security and how they approach security when it comes to cloud, because you're now in a position where many of your intellectual property and your physical data and information assets are no longer within your direct control.

So what are the capabilities that you need to mature in terms of governance, visibility, and audit controls that we were talking about, how do you ramp those up? How do you assess partners in those situations to be able to sit down and say that you can actually put trust into the cloud, so that you've got confidence that the assets you're putting in the cloud are safeguarded, and that you're not potentially threatening the overall organization to achieve quick wins?

The challenge is that the quick wins that the business is driving for could put the business at much longer-term risk, until we work out how to evolve our security practices across the board.

Gardner: We've been dealing with security issues for many years. Most people have been doing

When we start to look at what the cloud providers offer in terms of security, and whether our traditional security approaches are going to meet the need, we find a lot of flaws.

wide area networking and using the Internet for decades. Archie Reed, are the current technologies sufficient? Is the conventional approach to security all right? Or, do we need to recognize that we, one, either need new types of technologies, or two, primarily need to look at this from a process, people, and methodology perspective?

Reed: That's a long question. Tying into that question, and what Tim was just alluding to, most customers identify cost and speed to market as being the primary drivers for going or looking at cloud solutions.

Just to clarify one other point, in this discussion so far, we've been primarily talking about cloud providers as being external to the company. We haven't specifically looked at whether IT inside a large organization may be a cloud provider themselves to the organization and partners.

So, sticking with that model, alongside the cost and speed to market, when customers are asked what their biggest concerns are, security is far and away the number one concern when they think about cloud services.

The challenge is that security, as a term, is arguably a very broad, all-encompassing thing that we need to consider. When we start to look at what the cloud providers offer in terms of security, and whether our traditional security approaches are going to meet the need, we find a lot of flaws.

What we need to do is take some of that traditional security-analysis approach, which ultimately we describe as just a basic risk analysis. We need to identify the value of this data -- what are the implications if it gets out and what's the value of the service -- and come back with a very simple risk equation that says, "Okay, this makes sense to go outside."

If it goes outside, are the processes in place to say who can have access to this system, who can perform actions on the service that's providing access to that data, and so on.

Traditional approaches

Our traditional approaches lead us to the point where we can then decide what the appropriate actions are that we need to put in play, whether they be training for people, which is very important and often forgotten when you're using cloud services. Then decide the right processes that need to be used, whether they be implemented by people or automated in any way. Then ultimately, down to the actual infrastructure that needs to be updated, modified, or added, in order to get to the level of security that we're looking for. Does that make sense?

Gardner: Yes. It sounds as if it's not so much a technological issue, as something for the architects and the operational management folks to consider, a fairly higher-level perspective is needed.

Reed: Arguably, yes. Again, it depends what you're putting into the cloud. There are certain things where you may say, "This data, in and of itself, is not important, should a breach occur. Therefore, I'm quite happy for it to go out into the cloud."

An example may be if you have a huge image database, for example, a real estate company. The images of the properties, in and of themselves, hold little value, but the amount of storage and bandwidth that you as a company have got to put into play to deliver that to your customers is actually quite costly and may not be something that your IT department has expertise in.

A cloud provider may be able to not only host those images and deliver those images on a

Generally, when we talk to people, we come back to the risk equation, which includes, how much is that data worth, what are the implications of a bridge, and what is the value of the services being provided.

worldwide basis, but also provide extra image editing tools, and so on, such that you can incorporate that into an application that you actually house internally, and you end up with this hybrid model. In that way, you get the best of both worlds.

Generally, when we talk to people, we come back to the risk equation, which includes, how much is that data worth, what are the implications of a bridge, and what is the value of the services being provided. That helps you understand what the security risk will be.

Gardner: So, if you start to "componentize" your workloads and understand more about what can be put on a scale of risk, you can probably reduce your costs dramatically, if you do it thoughtfully, and therefore gain quite a competitive advantage.

Reed: Absolutely. We have a vision at HP. It's generally recognized out there as "Everything-as-a-Service." An IT department can look at that and take things down to those componentized levels, be it based on a bit of data that needs to be accessed, or we need to provide this very broad service. In that way, they can also help define what is appropriate to go into the cloud and what security mechanisms are necessary around that. Does the provider offer those security mechanisms?

Gardner: Is it important to get started now, even for companies that may not be using cloud approaches very much, to fully engage on this? Is it important and beneficial for them to start thinking about the processes, the security, and the risk issues? Let me pass that to David Spinks.

Next big areas

Spinks: The big areas that I believe will be developed over the next few years, in terms of ensuring we take advantage of these cloud services, are twofold. First, more sophisticated means in data classification. That's not just the conventional, restricted, confidential-type markings, but really understanding, as Archie said, the value of assets.

But, we need to be more dynamic about that, because, if we take a simple piece of data associated with the company's annual accounts and annual performance, prior to release of those figures, that data is some of the most sensitive data in an organization. However, once that report is published, that data is moved into the public domain and then should be unclassified.

What we're finding is that many organizations, once they classify a piece of data as confidential or secret, it stays at that marking, and therefore is prohibited from moving into a more open environment.

We need not just management processes and data-classification processes, but these need to be much more responsive and proactive, rather than simply reacting to the latest security breach. As we move this forward, there will be an increased tension to more sophisticated risk management tools and risk-management methodologies and processes, in order to make sure that we take maximum advantage of cloud services.

Gardner: Tim Van Ash, as companies start to think about this and want that holistic perspective, does adopting SaaS and consuming those applications as services provide a stepping-stone? Is this a good validation point?

Van Ash: Going back to the point that David was just making, it comes down to which

The level of data being held within an organization like Salesforce is extremely sensitive. Salesforce has had to invest tremendous amounts of time and energy in protecting their systems over the years.

processes you're putting into the cloud and the value tied to those processes.

For example, Salesforce.com has been very successful in the SaaS market. Clearly, they're the leader in customer relationship management (CRM) in the cloud today. The interesting thing about that is, the information they store on behalf of customers are customer data and prospect data, things that organizations guard very carefully, because it represents revenue and bookings to the organization.

If you look at how the adoption has occurred, it started out with small to medium companies for whom speed was often more important than the financial security, but it has now very much moved into the enterprise. The level of data being held within an organization like Salesforce is extremely sensitive. Salesforce has had to invest tremendous amounts of time and energy in protecting their systems over the years.

Likewise, if we look at our own SaaS business within HP, not only do we go through external audit on a regular basis, but we're applying a level of security discipline. It could be SAS 70 Type II around the data centers and practices, or being certified to an ISO standard, whether it be 27001 or one of the earlier variations of that. Cloud providers are now having to adhere to a very rigorous set of guidelines that, arguably, customers don't apply to the same level around their information internally.

The big reason for that is that when you run element as a service, you have to build supporting elements around that service. It's not a generic capability that exists across the entire business. So, there's a lot more focus placed on security from the SaaS model than maybe would have been applied to some of those elements within smaller to medium organizations, and, certainly, in some of the non-core functions in the enterprise.

Gardner: I assume that the ways in which an organization starts to consume SaaS and the experiences they have there does set them up to become a bit more confident in how to move forward toward the larger type of cloud activity.

Fear, uncertainty, doubt

Van Ash: That's a great point, Dana. Typically, what we see is that organizations often have concerns. They go through the fear, uncertainty, and doubt. They'll often put data out there in the cloud in a small department or team. The comfort level grows, and they start to put more information out there.

At the same time, going back to the point that both Dave and Archie were making, you need to evolve your processes, and those processes need to include the evaluation of the risk and the value of the information and the intellectual property that you're placing out there.

Spinks: One of the observations I've had talking with a lot of customers about so far, some big customers and small, is they're experiencing this situation where the business units are pushing internally to get to use some cloud service that they've seen out there. A lot of companies are finding that their IT organizations are not responding fast enough such that business units are just going out there directly to a cloud services provider.

They're in a situation where the advice is either ride the wave or get dumped, if you want an analogy. The business wants to utilize these environments, the fast development testing and launch of new services, new software-related solutions, whatever they may be, and cloud offers them an opportunity to do that quickly, at low cost, unlike the traditional IT processes.

But, all of these security concerns often get lost, because these things that they want to work on

Many enterprises today looking for quick wins are leveraging elements like IaaS to reduce their costs around testing and development.

are very arguably entrepreneurial in nature and move very quickly to try to capture business opportunities. They also may require partners to engage quickly and easily, and getting holes through firewalls and getting approvals can take months, if not quarters, in the traditional model. So, there is a gap in the existing IT architectural processes to implement and support these solutions.

That's what IT has got to deal with, if we focus on their needs for a minute. If they don't have a policy, if they don't have a process and advertise that within an organization, they will find that the business units will get up on that wave and just ride away without them.

Van Ash: We do see enterprises are being somewhat cautious, when they're applying it. As Archie was saying right upfront, you see a different level of adoption, a different level of concern, depending on the nature of the business and the size of the business. Many enterprises today looking for quick wins are leveraging elements like IaaS to reduce their costs around testing and development. These are areas that allow them to get benefit, but doing it in a way that is managing their risk.

Gardner: It sounds as if we need to get this just right. If we drag our feet as an organization, some of the business units and developers will perhaps take this upon themselves and open up the larger organization to some risk. On the other hand, if we don't adopt at a significant pace, we risk a competitive downfall or downside. If we adopt too quickly and we don't put in the holistic processes and think it through, then we're faced with an unnecessary risk.

I wonder, is there a third-party, some sort of a neutral certification, someone or some place an organization can go to in order to try to get this just right and understand from lessons that have been learned elsewhere?

Efforts underway

Reed: We would hope so. There are efforts underway. There are things, such as the Jericho Forum, which is now part of The Open Group. A group of CIOs and the like got together and said, "We need to deal with this and we need to have a way of understanding, communicating, and describing this to our constituents."

They created their definition of what cloud is and what some of the best practices are, but they didn't provide full guidelines on how, why, and when to use the cloud, that I would really call a standard.

There are other efforts that are put out by or are being worked on today by The National Institute of Standards and Technology, primarily focused on the U.S. public sector, but are generally available once they publish. But, again, that's something that's in progress.

The closest thing we've got, if we want to think about the security aspects of the cloud, are coming from the Cloud Security Alliance, a group that was formed by interested parties. HP supported founding this, and actually contributed to their initial guidelines.

Essentially, it lays out 15 focus areas that need to be concentrated on in terms of ensuring a level

So, my suggestion for companies is to take a look at the things that are underway and start to draw out what works for them, but also get involved in these sorts of things.

of security, when you start to look at cloud solutions. They include things like information lifecycle management, governance, enterprise risk management, and so on. But, the guidelines today, knowing of course that these will evolve, primarily focus on, "Here is the best practice, but make sure you look at it under your own lens."

If we're looking for standards, they're still in the early days, they're still being worked on, and there are no, what I would call, formal standards that specifically address the cloud. So, my suggestion for companies is to take a look at the things that are under way and start to draw out what works for them, but also get involved in these sorts of things.

Gardner: I just want to make sure I understood the name. Was it Jericho, the project that's being done by The Open Group?

Reed: Jericho Forum was the group of CIOs who essentially put together their thoughts, and then they've moved it under The Open Group auspices.

The Jericho Forum and the Cloud Security Alliance, earlier this year, signed an agreement to work together. While the Jericho Forum focused more on the business and the policy side of things, the Cloud Security Alliance focused on the security aspects thereof.

Gardner: What is HP specifically doing to advance the safe and practical use of cloud services, working I would imagine in concert with some of these standards, but also looking to provide good commercial services?

HP's efforts

Reed: There are many things going on to try and help with this. As I said, we were involved in the formation of the CSA, and we were involved, and are still involved, in helping write the guidance for critical areas, a focus in cloud computing, and the next generation. We are, through our EDS folks, directly involved with the Jericho Forum, and bringing those together.

We also have a number of tools and processes based on standards initiatives, such as Information Security Service Management (ISSM) modeling tools, which incorporate inputs from standards such as the ISO 27001 and SAS 70 audit requirements -- things like the payment card industry (PCI), Sarbanes-Oxley (SOX), European Data Privacy, or any national or international data privacy requirements.

We put that into a model, which also takes inputs from the infrastructure that's being used, as well as input based on interviews with stakeholders to produce a current state and a desired or required state model. That will help our customers decide, from a security perspective at least, what do I need to move in what order, or what do I need to have in place?

That is all based on models, standards, and things that are out there, regardless of the fact that cloud security itself and the standards around it are still evolving as we speak.

Gardner: Tim Van Ash, did you have anything further to offer in terms of where HP fits into

Cloud Assure is really designed to deal with the top three concerns the enterprise has in moving into the cloud.

this at this early stage in the secure cloud approach?

Van Ash: Yeah. In addition to the standards and participation that Archie has talked about, we do provide a comprehensive set of consulting services to help organizations assess and model where they are, and build out roadmaps and plans to get them to where they want to be.

One of the offerings that we've launched recently is Cloud Assure. Cloud Assure is really designed to deal with the top three concerns the enterprise has in moving into the cloud.

Security, obviously, is the number one concern, but the number two and three concerns are performance and availability of the services that you're either consuming or putting into the cloud.

Cloud Assure is designed and delivered through the HP Software-as-a-Service Group, so that its a way that organizations can assess potential cloud services that they want to consume for those security issues, so that they know about it before they go in. This can help them to choose who is the right provider for them. Then, it's designed to provide ongoing assessment of the provider over the life of the contract, to ensure that they continue to be as secure as required for the type of information and the risk level associated with it.

The reason we do it through SaaS is to enable that agility and flexibility of those organizations, because speed is critical here. Often, the organizations aren't in a position to put up those sorts of capabilities in the timeframe the business is looking to adopt them. So, we're leveraging cloud to enable businesses to leverage cloud.

Gardner: David Spinks, are there areas where success is being meaningfully engaged now? Are there early adopters? Where are they? And, are they really getting quite a bit of productivity from moving certain aspects or maybe entire sets of IT functions or business functions to the cloud?

Moving toward cloud

Spinks: We're seeing some of the largest companies in the world move towards cloud services. You've got the likes of Glaxo and Coca-Cola, who are already adopting cloud services and, in effect, learning by actual practical experience. I think we'll see other large corporations in the world move towards the adoption of cloud, because obviously they spend the most on IT and, therefore, have got the most to gain from incremental savings.

The other key technology that we'll see emerge from one of the issues in cloud computing in the whole area of personal authentication, authorization, and federated access is this concept called Role-Based Access Control (RBAC).

There are a number of clients who are talking to us about how we might use our experiences with some of the largest corporations and government agencies in the world in terms of putting more robust authentication processes in place, allowing our largest clients to collaborate with their customers and their partners.

One of the key technologies there, and obviously one of the key technologies that Jericho have been pushing for years, is much more robust identity management and authentication, including technologies such as two-factor authentication and managed public key infrastructure (PKI). I would prophesize that we're going to see an explosion in the use of those technologies, as we move further and further into the cloud.

Gardner: Well, very good, I'm afraid we're about out of time. We've been having a discussion about overcoming fear -- caution and the need for risk reduction on the road to successful cloud computing. Our panelists have been Archie Reed, HP Distinguished Technologist and Chief Technologist for cloud security. I certainly appreciate your input Archie.

Reed: Thank you very much, Dana.

Gardner: Tim Van Ash, director of SaaS products at HP Software and Solutions. Thank you, Tim.

Van Ash: Thanks very much, Dana.

Gardner: And David Spinks, security support expert at HP IT Outsourcing. Thank you, David.

Spinks: You're very welcome.

Gardner: This is Dana Gardner, principal analyst at Interarbor Solutions. You've been listening to a sponsored BriefingsDirect podcast. Thanks, and come back next time.

Listen to the podcast. Find it on iTunes/iPod and Podcast.com. Download the transcript. Learn more. Sponsor: Hewlett-Packard.

Free Offer: Get a complimentary copy of the new book Cloud Computing For Dummies courtesy of Hewlett-Packard at www.hp.com/go/cloudpodcastoffer.

Transcript of a sponsored BriefingsDirect podcast on the state of security in cloud computing and what companies need to do to overcome fear, reduce risk and still enjoy new-found productivity. Copyright Interarbor Solutions, LLC, 2005-2009. All rights reserved.

Harnessing 'Virtualization Sprawl' Requires Managing Your Ecosystem of Technologies

Transcript of a sponsored BriefingsDirect Podcast on how companies need to deal with the complexity that comes from the increasing use of virtualization.

Listen to the podcast. Find it on iTunes/iPod and Podcast.com. Download the transcript. Learn more. Sponsor: Hewlett Packard.

Free Offer: Get a complimentary copy of the new book Cloud Computing For Dummies courtesy of Hewlett-Packard at www.hp.com/go/cloudpodcastoffer.

Dana Gardner: Hi, this is Dana Gardner, principal analyst at Interarbor Solutions, and you’re listening to BriefingsDirect.

Today, we present a sponsored podcast discussion on better managing server virtualization expansion across enterprises. We’ll look at ways that IT organizations can adopt virtualization at deeper levels, or across more systems, data and applications, at lower risk.

As more enterprises use virtualization for more workloads to engender productivity from higher server utilization, we often see what can be called virtualization sprawl, spreading a mixture of hypervisors, which leads to complexity and management concerns.

In order to ramp up to more, but advantageous, use of virtualization, pitfalls from heterogeneity need to be managed. Yet, no one of the hypervisor suppliers is likely to deeply support any of the others. So, how do companies gain a top-down perspective of virtualization to encompass and manage the entire ecosystem, rather than just corralling the individual technologies?

Here to help us understand the risks of hypervisor sprawl and how to mitigate the pitfalls to preserve the economic benefits of virtualization is Doug Strain, manager of Partner Virtualization Marketing at HP.

Doug Strain: Thanks, Dana.

Gardner: Help us out. What is the current state of virtualization adoption? Are we seeing significant pickup as a result of the economy? What’s driving all the interest in this?

Strain: Virtualization has been growing very steeply in the last few years anyway, but with the economy, the economic reasons for it are really changing. Initially, companies were using it to do consolidation. They continue to do that, but now the big deal with economy is the consolidation to lower cost -- not only capital cost, but also operating expenses.

Gardner: I imagine the underutilization of servers is like a many-headed dragon. You’ve got footprint, skills, and labor being used up. You’ve got energy consumption. You’ve got the applications and data that might be sitting there that have no real purpose anymore, or all of the above. Is this is a big issue?

Underutilized capacity

Strain: It definitely is. There’s a lot of underutilized capacity out there, and, particularly as companies are having more difficulty getting funding for more capital expenses, they’ve got to figure out how to maximize the utilizations they’ve already bought.

Gardner: And, of course the market around virtualization has been long in building, but we’ve had a number of players, and some dominant players. Do you see any trends about adoption in terms of the hypervisor providers?

Strain: Probably, we’re seeing a little bit of a consolidation in the market, as we get to a handful of large players. Certainly, VMware has been early on in the market, has continued to grow, and has continued to add new capabilities. It's really the vendor to beat.

Of course, Microsoft is investing very heavily in this, and we’ve seen with Hyper-V, fairly good demand from the customers on that. And, with some of the things that Microsoft has already announced in their R2 version, they’re going to continue to catch up.

We’ve also got some players like Citrix, who really leverage their dominance in what’s called Presentation Server, now XenApp, market and use that as a great foot in the door for virtualization.

Gardner: That’s a good point. Now, we introduced this as a server virtualization discussion, but virtualization is creeping into a variety of different aspects of IT. We’ve got desktop virtualization now, and what not. Tell us how this is percolating up and out from its core around just servers.

Strain: Desktop virtualization has been growing, and we expect it to grow further. Part of it is just a comfort within IT organizations that they do know how to virtualize. They feel comfortable with the technology, and now, putting a desktop workload instead of server workload, is sort of a natural way to extend that and to use as resources more wisely.

Probably the biggest difference in the drivers for desktop virtualization is the need for meeting compliance regulations, particularly in financial, healthcare, and in a lot of other industries, where customer or employee privacy is very important. It makes sure that the datas no longer sits on someone’s desk. It stays solely within the data center.

Gardner: So there are a lot of good reasons for virtualizing, and, as you point out, the economy is accelerating that from a pure dollars-and-cents perspective. But this is not just cut and dried. In some respects, you can find yourself getting in too deep and have difficulty navigating what you’ve fallen into.

Easy to virtualize

Strain: That’s definitely true, and because of the fact that all the major vendors now have free hypervisor capabilities, it becomes so easy to virtualize, number one, and so easy to add additional virtual machines, that it can be difficult to manage if technology organizations don’t do that in a planned way.

Gardner: As I pointed out, it’s difficult to go back to just one of the hypervisor vendors and get that full panoply of services across what you’ve got in place at your particular enterprise, which of course might be different from any other enterprise. What’s the approach now to dealing with this issue about not having a single throat to choke?

Strain: There are a couple of dimensions to that. As you said, most of the virtualization vendors do have management tools, but those tools are really optimized for their particular virtualization ecosystem. In some cases, there is some ability to reach out to heterogeneous virtualization, but it’s clear that that’s not a focus for most of the virtualization players. They want to really focus on their environment.

The other piece is that the hardware management is critical here. An example would be, if you’ve got a server that is having a problem, that could very well introduce downtime. You've got to have a way of navigating the virtual machine, so that those are moved off of the server.

That’s really an area where HP has really tried to invest in trying to pull all that together, being

. . . Having tools that work consistently both in physical and in virtual environments, and allow you to easily transition between them is really important to customers.

able to do the physical management with our Insight Control tools, and then tying that into the virtualization management with multiple vendors, using Insight Dynamics – VSE.

Gardner: We’ve discussed heterogeneity when it comes to multiple hypervisors, but we’re also managing heterogeneity, when it comes to mixtures of physical and virtual environments. The hypervisor provider necessarily isn’t going to be interested in the physical side.

Strain: That’s exactly right. And, if they were interested, they don’t necessarily have the in-depth hardware knowledge that we can provide from a server-vendor standpoint. So yeah, clearly there are a few organizations that are 100 percent virtualized, but that’s still a very small minority. So, we think that having tools that work consistently both in physical and in virtual environments, and allow you to easily transition between them is really important to customers.

Gardner: All right. How do we approach this, and is this something that is like other areas of IT we’ve seen, where you start at a tactical level and then over time it gets too complex, too unwieldy, you start taking more strategic overview and come up with methodologies to set some standards up. Is this business as usual in terms of a maturation process?

Strain: I think that’s what we’ve seen in the past. I certainly wouldn't recommend that to somebody today that’s trying to get into virtualization. There are a lot of ways that you can plan ahead on this, and be able to do this in a way that you don't have to pay a penalty later on.

Capacity assessment

It could be something as simple as doing a capacity assessment, a set of services that goes in and looks at what you’ve got today, how you can best use those resources, and how those can be transitioned. In most cases you’re going to want to have a set of tools like some of the ones I’ve talked about with Insight Control and Insight Dynamics VSE, so that you do have more control of the sprawl and, as you add new virtual machines, you do that in a more intelligent way.

Gardner: Tell us a little bit about how that works? I've heard you guys refer to this as "integrated by design." What does that mean?

Strain: We’ve really tried to take all the pieces and make sure that those work together out of the box. One of the things we’ve done recently to up the ante on that is this thing called BladeSystem Matrix. This is really converged infrastructure that allows customer to purchase a blade infrastructure complete with management tools with the services, and a choice of virtualization platforms. They all come together, all work together, are all tested together, and really make that integration seamless.

Gardner: And HP is pretty much neutral on hypervisors. You give the consumer, the customer, the enterprise the choice on their preferred vendor.

Strain: We do. We give them a choice of vendors. The other thing we try to do is give them a choice of platforms. We invest very heavily in certifying across those vendors, across the

What we’re finding is that we can’t say that one particular server or one particular storage is right for everybody. We’ve got to meet the broadest needs for the customers.

broadest range of server and storage platforms. What we’re finding is that we can’t say that one particular server or one particular storage is right for everybody. We’ve got to meet the broadest needs for the customers.

Gardner: Let's take a look at how this works in practice. Do you have any examples, customers that have moved in this direction at a significant level already, and perhaps had some story about what this has done for them?

Strain: I’ve just pulled up a recent case study that we did on a transportation company, called TTX Company. I thought this was a good example, because they’d really tried a couple of different paths. They’d originally done mainframes, and realized that the economics of going to x86 servers made lot more sense.

But, what they found was they had so many servers, they weren’t getting good utilization, and they were seeing the expenses go up, and, at the same time, seeing that they were starting to run out of space in their data center. So, from a pure economic standpoint, they looked at this and said, “Look, we can lower our hardware cost.”

TCO 50 percent lower

In fact, they saw a 10 percent reduced-hardware cost, plus they’re seeing substantial operating expense reductions, 44 percent lower power cost, and also 69 percent reduction in their rack footprint. So, they can now say they are removing it from the datacenter and, compared to their mainframes, they think they have about a 50 percent lower total cost of ownership (TCO).

Gardner: So, if you do this right, they're not just rounding-error improvements. These are pretty substantial.

Strain: These are substantial, and, particularly today, that’s a great way to justify virtualization. What they also found was that, from an IT standpoint, they were much more effective. They project that they can recover much more quickly -- in fact, 96 percent reduction in recovery time. That's going from 24 hours down to 1 hour recovery.

Likewise, they could deploy new servers much more quickly -- 20 minutes versus 4 hours is what they estimate. They’ve reduced the times they have to actually touch a server by a factor of five.

Gardner: So, we’ve seen quite a few new studies that have come out, and virtualization remains in the top tier of concerns and initiatives from enterprises, based on the market research. We’re also seeing interesting things like managing information explosion and reducing redundancy in terms of storaging data. These all come together at a fairly transformative level.

How big a part in what we might consider IT transformation does virtualization play?

Strain: It plays a very substantial role. It’s certainly not the only answer or not the only

The investment in the industry is around management, making it simpler to deploy, to move, to allow redundancy, all those kinds of things, as well as automation.

component of data center transformation, but it is a substantial one. And, it's one that companies of almost any size can take advantage of, particularly now, where some of the requirements for extensive shared storage have decreased. It's really something that almost anybody who's got even one or two servers can take advantage of, all the way to the largest enterprises.

Gardner: So, at a time when the incentives, the paybacks from virtualization activities are growing, we’re seeing sprawl and we’re seeing complexity. This needs to be balanced out. What do you think is the road map? If we had a crystal ball, from your perspective in knowledge of the market, how do we get both? How do we get the benefits without the pain?

Strain: Clearly, this is an area where the entire industry is investing heavily in not just the enabling of the virtualization. That’s been done. There’s still some evolution there, but the steps are getting increasingly smaller. The investment in the industry is around management, making it simpler to deploy, to move, to allow redundancy, all those kinds of things, as well as automation.

There are a lot of tasks that, particularly when you think about a virtual machine, can be run on a range of different hardware, even in different datacenters. The ability to automate based on a set of corporate rules really can make IT much more effective.

Gardner: Great. We’ve been talking about better managing server virtualization expansion across enterprises, and we’ve been joined in our discussion by Doug Strain. He is the manager of Partner Virtualization Marketing at HP. We appreciate it, Doug.

Strain: My pleasure.

Gardner: This is Dana Gardner, principal analyst at Interarbor Solutions. You've been listening to a sponsored BriefingsDirect podcast. Thanks for listening, and come back next time.

Free Offer: Get a complimentary copy of the new book Cloud Computing For Dummies courtesy of Hewlett-Packard at www.hp.com/go/cloudpodcastoffer.

Listen to the podcast. Find it on iTunes/iPod and Podcast.com. Download the transcript. Learn more. Sponsor: Hewlett Packard.

Transcript of a sponsored BriefingsDirect Podcast on how companies need to deal with the complexity that comes from the increasing use of virtualization. Copyright Interarbor Solutions, LLC, 2005-2009. All rights reserved.