Welcome to the latest BriefingsDirect SOA Insights Edition, a weekly discussion and dissection of Service Oriented Architecture (SOA) and related news and events with a panel of independent IT industry analysts and guests. Yours truly, Dana Gardner, principal analyst at Interarbor Solutions, is your host and moderator.
BriefingsDirect SOA Insights Edition brings a new focus to the SOA thought leadership field by assembling noted independent IT industry analysts and guests to investigate the trends and news of the week. Take it via this blog and excerpts, by listening to the podcast, and/or by perusing the full transcript of the discussion.
This week's panel consists of Steve Garone, a former program vice president at IDC and founder of the AlignIT Group, and Jon Collins, principal analyst at Macehiter Ward-Dutton. The topics are Sun's virtualization announcement and the relationship between SOA and virtualization, as well examining the notion that SOA will kill ERP ... or not.
Among the more intriguing issues of the day is the intersection of SOA and virtualization. Sun Microsystems made a recent announcement, and IBM and HP, to name a few, have been very active in this field. The vendors are expanding the definition of virtualization: They're running services in a closed individual environment that's secure and can exploit the power of the hardware, such as blades and multicore, and they are increasingly virtualizing data, storage, and middleware resources.
There’s been another hot topic in the discussion around SOA recently, and that's about the impact of SOA on ERP, and CRM, as well as other packaged business applications. On one side of the fence, we have people saying that SOA is going to be the death-knell, if you will, or certainly diminish the role and impact of packaged applications. We’ve had some large analyst firms come out with reports and some sensational headlines to that effect. And, yet, when I speak to people in the field they say, "No way"; that SAP and Oracle and other packaged application vendors are in for the long haul, and that they’re entrenched.
And, they say that SOA might become a layer above those packaged applications in order to create a data services layer and perhaps a transactional and logic services layer above those. But that doesn’t mean you’ve ripped them out. And, like the mainframe, they'll be there in 20 years. Let’s talk a little bit about the impact of SOA on the packaged business applications field.
Here are some excerpts from Collins, Gardner, and Garone this lively show:
All of the big players obviously -- meaning IBM, Microsoft, and HP -- are talking about their virtualization offerings. Interestingly, if you take IBM as an example, virtualization can mean very different things, depending on which part of the organization you're talking to. If you are talking about, for example, virtualization of information access, that’s almost a middleware question of how to access multiple types of structured data.
But, if you're talking about virtualization as infrastructure, then it’s very much about the resource management question. So, any one of the organization may have different virtualizations. It's a question of how you bring those together. One company that’s probably coming up alongside -- it’s trying to get as far up as it can before people stop it -- is EMC.
And, yes, obviously EMC has VMware, so they have done a great deal in virtualization on the server side, but equally with their nLayer acquisition, and other management-type acquisitions, they're putting in place a lot of the pieces they need to be able to manage the virtualized infrastructure they end up with. I think they're doing that quietly and hoping that no one sees they're trying solve that problem.
I’m reminded of when Microsoft stood up and said: "Of course we’ve always been a virtualization company." They tend to say that they've always been a "this" company, they’ve always been a "that" company. The reason they said they’ve always been a virtualization company is, of course, virtual file systems -- caching and paging and swapping all these mechanisms inside the operating system.
I think extreme virtualization ... will succeed ... (and) ... we’ll stop talking about it, and it will become parts of the distributed platform of the future. And it will just be an expectation -- SOA will run and run because it’s a way of doing things. Virtualization should just become part of the standard, if you like.
When I’ve spoken to actual customers of these things, no one is looking to go and replace them. And if they are, that it's not because they are thinking about SOA. What I’ve written here is that at the bottom line, you’ve got a database, you’ve got a rules engine, you’ve got a workflow engine, and it's been crafted and customized for that organization. Why on earth would they want to rip it out and put something else in place? Probably other analysts think about it in terms of market segments and products, and so on. But that’s not the way that we need to think about this.
There is a little bit of that deja vu all over again. We talked earlier about SAP and components. I remember discussions like this back when software components started to appear in the mid 1990s, and even before then, when people were talking objects and CORBA and how we're going to get all this stuff to work. If you go by those kinds of predictions and supposed trends, SAP and PeopleSoft -- well, PeopleSoft isn’t independent now -- but an Oracle and the other enterprise app vendors would have been out of business a long time ago.
The fact is that people will want to stay with their current solutions, especially in the case of the ERP, where that high level of integration that brings a lot of the benefits of ERP, has been established and is working for you. And then, tack on to that the technologies that will allow you to say, “Manage the supply chain better,” "Bring your supply chain in better contact and better integration with the systems that you currently have."
Perhaps the way to look at SOA is not what we’ve been hearing a little bit of lately, which is SOA kills ERP. It's not rip and replace. It's not either-or. It's both. From my perspective, SOA’s main value proposition in the short- and medium-term is about risk reduction.
If you can put in some buffer around your legacy systems -- if you can use SOA for more greenfield development, too -- if you can use it for extended enterprise opportunities where you can share services in a close-knit trading or supply chain environments, or a business ecology environment -- then you’re reducing your risk for being backward compatible because your services can continue to be modernized. That is to say, your legacy can be modernized into services and the investments can be exploited. You can reduce your risk in terms of flexibility by being fast to business process development in the current day, and you can reduce your risk in terms of being ready to accept what’s going to be new and interesting, innovative, and productive in the future. What’s wrong with this logic about SOA, as a risk reducer and not a displacer of packaged applications?
I’ve got the phrase written down here, "augmentation not replacement." I think that you're absolutely right to see everything in terms of reduced risk. I would say that companies are always looking for reduced risks, but they are not necessarily saying that they’ve got high levels of risk, and need to reduce them. But, maybe I would re-phrase it to, “Increase flexibility without increasing risk.” With the idea of replacing something and starting from scratch, there’s an inordinate amount of risk involved, because potentially you go through the same problems you went through last time. Whereas if you create an interface to the tried and tested, you can develop new ways of using that tried-and-tested business logic and tried-and-tested the data management logic without being held back.
Listen to the podcast, or read the full transcript for more on this week's SOA news and analysis.