By: Robert Duffner
Srinivasan Sundara Rajan works at Hewlett Packard as a Senior Solution Architect. His primary focus is on enabling service-oriented architecture (SOA) through legacy modernization for automobile 
Robert Duffner: Could you please take a moment to introduce yourself?
Srinivasan Sundara Rajan: I currently work as a senior solutions architect for HP Enterprise Services, catering to the larger customer accounts. I have more than 19 years of experience in the industry. I write regularly in Cloud Computing Journal on enterprise cloud adoption.
Robert: On the Supply Chain and Technology blog, the author says that, "SMBs are frantically moving to the cloud. Enterprises are not ready for it." What are your thoughts on that?
Srinivasan: I think that is correct to an extent, in the sense that enterprises are still contemplating. For an enterprise, infrastructure as a service is useful , but that is not a business transformation enabler . The real challenge for them is moving to the cloud so that their time to market improves, and software as a service is an ultimate goal for them, although they can never really arrive at the perfect software-as-a-service (SaaS) platform in the short run
The issue for them is that the big ERP packages like SAP or Oracle Financials require a lot of customization. With that in mind, no SaaS offering can straightaway satisfy most of the larger enterprise needs, so the best bet for the enterprises is to get to a platform-as-a-service (PaaS) platform on top of your SaaS building blocks. In other words, you satisfy your basic needs on the SaaS, but you extend on top of it with PaaS to satisfy your enterprise needs.
That type of platform is still evolving, and enterprises are not finding a ready-made answer to get into the cloud. Even though a lot of the standard concerns like security and availability have been addressed, I feel that the business capability enablement has yet to mature so that enterprises can fully adapt to the cloud.
Robert: How do you recommend that enterprises with the full spectrum of both very old legacy systems and new applications approach the cloud?
Srinivasan: I always choose the Microsoft platform because of its PaaS and strong integration tools with the on premises applications, and my approach is to recommend a hybrid environment. For the foreseeable future, a 100% cloud-only solution is not going to be viable for large enterprises.
The enterprises yet continue to have their own data centers to support certain legacy applications and calculations that are not available as a SaaS s, for various reasons, enterprises are not going to to move these critical applications to the cloud.
Still, there is very good potential for some 30 to 40 percent of the workload to move into the cloud.
Let us take an example, in the very specific case of a data warehousing scenario, from a business intelligence perspective, there is an operational data store which is a consolidation of all your transactions systems in near real time.
Near-real-time consolidation of all your operational systems may not move to the cloud, because they have to be near to your data centers to accumulate data as quickly as possible. On the other hand, there are operations such as enterprise reporting and data marts focus on historical, staggered accumulation of data.
You could always wait a week to accumulate your operational data store into the data warehouse, so the latency of cloud may be very compatible with that scenario. On the other hand, enterprise reporting raises issues in terms of licensing and standardization of the reports. For big enterprises facing those kinds of problems, SaaS or PaaS would be ideal for reporting. That's why the overall solution tends toward a hybrid delivery
Robert: I am glad you mentioned the hybrid cloud environment, which is something we are finding a great deal of interest in among customers interested in Windows Azure. You just recently wrote an article about how data warehousing fits in a hybrid cloud environment.
Srinivasan: A data warehouse is not just one database. You have a transactional system, which may be like any kind of a legacy system. It could involve mainframes, ERPs, and other stuff. And traditionally you should have an operational data store, which is at the transaction level. You want to capture a real-time, or near-real-time, consolidated view of all this data.
You may have 15 different technologies like AS/400, mainframes, Oracle, and ERP in your transactional system and want to consolidate them in near real time into an ODS. Because of that near-real-time requirement and having so many adapters in place and so forth, you may want to keep your operational data store inside your data center, because of the latency and availability factors.
Data warehousing typically works on a variable load pattern. For example, a telecom company may load all its billing data to a data warehouse, so naturally during the billing cycles, they would have a very high level of activity, and then it might be cool for the next 10 days. And when the new customer file comes, they may want to do lot of address validation, address cleansing, and so on.
That variable load and the fact that data accumulates over a period, rather than requiring real-time response, as well as the availability of a common SaaS platform for reporting and a PaaS platform for extending that reporting, make the rest of the components a very good candidate for moving to the cloud.
In this scenario, certain pieces like the operational data store and transactional system continue to be on premises. The data warehouse itself and the compute resources that handle the monthly or weekly differential load can move to the cloud. That combination can benefit the enterprise in terms of operational and other expenses while also providing the benefits of a platform tool.
Robert: Are there any other scenarios where you see hybrid cloud as being particularly important, beyond those that you've already mentioned?
Srinivasan: Product lifecycle management is becoming more and more collaborative. Big manufacturers make certain components of a large product such as an automobile, airplane, or industrial equipment, and you want a lot of suppliers to collaborate with the design. You also want to enable connected parts purchasing, quality assurance, and several other areas.
For example Azure has got the Windows Active Directory Federation and other stuff. These tools give you a very good platform for multiple people to collaborate over the cloud while still keeping certain systems isolated. Please also refer to my article on, Federated Security in Windows Azure
Hybrid cloud can also be very useful in terms of reducing dependence on legacy systems. For example, you may want to reduce the amount of processing you depend on mainframes for over time. Mainframes are very good for transactions, especially real-time transactions.
You won't be able to get an equivalent level of performance for the foreseeable future in the cloud, but reporting doesn't have to coexist with the transaction system. That can be moved to the cloud, and there are quite a few scenarios like that where I could see a good amount of cross usage between on-premises and cloud systems.
Robert: Can you talk a little bit about how you see technologies like SQL Azure and DataSync playing a role?
Srinivasan: In a data warehousing scenario, say you have your ODS and your current data warehouse on premises, and you have decided to move your data warehouses to the cloud. Naturally, you cannot use your REST-based service invocations to move those enormous amounts of data, so first of all, you need appropriate tools. SQL Azure is exciting here, because it provides you lot of native tools for the movement of large data. Of course, Informatica and other third-party vendors also support it.
Whenever you are talking about this kind of a hybrid delivery, you also need to address continuous integration, because it is not just one time movement of data. There you've got multiple choices. You can use queues for a publish-and-subscribe approach, which will give you asynchronous access. You can also use SQL Azure for more of an ETL type of data transformation, so you set some kind of a source and a target. This is especially so if your source database is in SQL Server, because ETL plays a major part in moving the data from one place to another.
You are never going to have the data 'as is' getting transformed from a source to a target. You'll have to use multiple transformations, so you need an automated tool to support movement from an on-premises system to the cloud. That means you need several things to be taken care of.
First of all, you need the adapters. You need to have a common tool that will talk to your on-premises data sources and to the cloud. The second consideration is security, including how your credentials, tokens, access accounts, and pass words are integrated. The third concern is about ensuring a rich set of functionalities. Since you have built it on TSQL, you have a rich set of functionality for transformation. Those tools will definitely have a larger role on a hybrid kind of a platform.
Robert: In the article, "Improving the Business Value of SaaS Applications," you talk about the need to link cloud application authentication back to enterprise directories. How do you see SaaS vendors and enterprises handling this requirement today?
Srinivasan: I may be wrong, but what I have seen in Amazon is that you need to go and register yourself as a user. Consider how untenable that is in an enterprise of 150,000 or 200,000 users. It's not like all the users are going to access that same application, but you will need consistency.
Suppose you are using a workflow kind of CRM application, where a service request needs to go to your manager, and it may need to be escalated further, depending upon the requirements of the individual transaction. You can't set up all that authentication data or your organizational directory data into the cloud, because that would require a huge amount of synching and potential chaos from data mismatches.
Therefore, I need certain things to be tied back to my enterprise directory. I still use a CRM application over the SaaS, but when my customer sends me a complaint that I don't handle in a timely fashion, it should automatically be forwarded on to my manager.
Such a synergy is possible only if the authentication and authorization is tied back to the enterprise directories. Of course, it will put a lot on the security mechanisms, but I don't really feel a need to duplicate the security inside the cloud. It may be good for individual users but not for entire enterprises.
Robert: You spend a lot of time on legacy modernization. Is there a fit for cloud in this regard?
Srinivasan: Yes, in a few different ways. First of all, most legacy modernization is targeted at server consolidation or platform migration. I might move a system from multiple mainframes into a single DB2 system on commodity servers.
Suppose the mainframe or other legacy system is implementing content management or document management. You could certainly move that to Office 365, SharePoint, or a similar content management system on the cloud.
This opportunity is not just about moving one to one. You could establish certain dynamic scaling properties on top of it, for example. Moreover, if you really think about your business processes with a holistic approach, you might investigate what other business processes are redundant on the legacy systems and consider whether those can also be moved to the Cloud. It may also be that you would be able to extend SaaS offerings using PaaS.
Robert: You also wrote an article comparing cloud sourcing versus outsourcing. Can you talk us through your thinking on that?
Srinivasan: I have worked both in the US and in India, so I think I know both business environments to some extent. I am not pointing at any single company, but I believe that, under the current outsourcing model, companies spend most of their time on maintenance, which means that they get relatively less value for their business capability.
Say I, as an enterprise, wanted to create a new report to satisfy a federal compliance requirement. What we current see in an outsourcing world, we go to the business unit that requested it and tell them that it would have cost them about $10k, but because their system is on an unsupported platform or an older version of the database, it is going to cost more like $100k.
That has been a relatively common scenario, and while it is not really a problem of outsourcing, but outsourcing is not clearly able to solve that situation. You can't really plan in advance for things like that, and with outsourcing, you generally have a multi-year contract which provides less room for dynamically adding and removing resources.
If it's executed well, cloud gives you a lot of choice that increases your capability to concentrate on business capability versus maintenance.
Security and trust also come into play. When you move your data to your outsourcing provider or let that provider access your databases, there is a very similar set of considerations as when you move the same data to a trusted cloud provider. Therefore, that security aspect remains much the same between outsourcing versus cloud sourcing.
Because there is the potential for a huge benefit in terms of business capability if executed correctly, outsourcing may evolve into a combination of cloud sourcing plus outsourcing plus data center maintenance. The outsourcing companies may also help companies move towards cloud sourcing.
Robert: The cloud is clearly a no brainer for startups because you can pretty much trade your capital expenses for more operating expenses. Do you see any corresponding best-kept secrets in terms of cloud advantages for big enterprises?
Srinivasan: One very interesting concept there is the community cloud perspective. Think about an automobile manufacturer that needs to comply with a large number of separate regulatory bodies from all over the world. It could carry substantial cost to create separate sets of reports from scratch for each of those worldwide governments.
Of course, those reports are largely standard in nature, such as emissions reports that need to be completed by a large number of international car makers. Therefore, each individual car company stands to benefit in terms of the cost and time requirements to create those reports if they band together to create a common SaaS model, which is not available today.
Another aspect of the community cloud concept is an industry exchange for information. For example, certain in-car controls created by big automobiles, collect and track information about collisions on the highway. If I get into an accident, information about it is automatically passed on to a help desk that handles my automated request for help. Aggregating the support for those information stores among multiple car companies represents substantial cost savings, as well as the ability to get better geographical coverage.
So community cloud can create some valuable synergies among different companies, even though they may be competitors. In that sense, this idea is similar to relationships that are commonplace in other areas. For example, Microsoft, IBM, HP and Oracle collaborate on projects like SOA standards or security standards, even though they are competitors in many market segments.
Robert: Large enterprises traditionally have to spend a significant amount of time thinking about things like business continuity and disaster recovery. How do you think those considerations are impacted by the cloud?
Srinivasan: Cloud provides you things like standby server support, replication, and log shipping out of the box, although there is of course the issue of how trusted they are. Large providers are providing very good SLA commitments, and I have written on the topic of infrastructure-as-a-service acting as a disaster recovery platform.
Disaster recovery and that sort of thing follow naturally from some of the basic tenets of cloud, since things are automatically copied and replicated, for example. And the tools are also available as part of the platform to get back your old data from your replicated versions.
Of course, enterprises have to go through certain studies and proofs of concept there, to figure out how cloud suits their particular scenarios, but for applications that are already hosted on the cloud, there is definitely a simple alternative to conventional backup and disaster recovery.
Robert: People have speculated that server shipments might decline as a result of virtualization innovation and now the cloud, but that hasn't happened. How do you see these technologies impacting the need for servers?
Srinivasan: Workloads are continuing to increase, and in fact, the presence of the cloud helps to accelerate that growth. Data volumes are growing, and what we want to do with them is also. What cloud is really giving you is that you don't have to spend lot of your time thinking about how much compute power to have in reserve to avoid going over capacity and losing business.
Robert: That concludes my planned questions for today. Is there anything else that you would like to talk about?
Srinivasan: To touch back on the community cloud concept, it is very difficult to achieve SaaS using a common package that suits every enterprise. You need customization, which is going to be available in the form of PaaS, and how well you provide integration from your on-premises systems will be a key differentiator as businesses try to focus on generating business value instead of using their resources on maintenance.
A lot of the new offerings are pretty exciting, but we still need a lot of case studies and proofs of concept to illuminate the path for enterprises to follow in adopting cloud. There are still a lot of issues to work through, including vendor lock-in and many others, but ultimately the effort will be worth it, and businesses on the whole will benefit.
Robert: Thanks for taking the time to talk today, Srini. I really appreciate your insights.
Srinivasan: Thank you.
source: Technet Blog
