The latest BriefingsDirect SOA Insights Edition, Vol. 24, provides a roundtable discussion and dissection of Services Oriented Architecture (SOA)-related news and events, with a panel of IT analysts. In this episode, our experts examine virtualization trends through the acquisition this summer of XenSource by Citrix. We also discuss the $1.6-billion purchase of Opsware by HP as a way of analyzing IT management, automation and operations, as well as the impact on SOA.
Join noted IT industry analysts and experts Jim Kobielus, principal analyst at Current Analysis; Neil Macehiter, research director at Macehiter Ward-Dutton; Dan Kusnetzky, principal analyst and president of the Kusnetzky Group; Brad Shimmin, principal analyst at Current Analysis; Todd Biske, practicing enterprise architect for a Fortune 500 firm (formerly with MomentumSI); JP Morgenthal, CEO of Avorcor; and Tony Baer, principal of onStrategies. Our discussion is hosted and moderated by me, Dana Gardner.
Here are some excerpts:
On Virtualization and Citrix-XenSourceListen to the podcast. Or read a full transcript.
Virtualization is definitely something that organizations are looking at right now. For the clients I've worked with, it's been a mix. Some are really trying to embrace it on the server side and make use of it right now. Others are looking at possibly using it on the desktop for developers, when they need to get a specific development environment, but it’s definitely in people’s minds today. So, I would definitely classify it as in the list of strategic initiatives that companies are looking at and determining how to use appropriately.
The interesting thing about XenSource is that it’s been considered to be the leading, emerging alternative to VMware. It essentially virtualizes the machine to a slightly more native approach than VMware. It's a very interesting acquisition because Xen has had a relationship with Microsoft, where it gets access to Microsoft's virtualization technology, and it also fills a key gap for Citrix. ... Microsoft will still need some way of interoperating its hypervisor with the Linux environment. So, even though the relationships may change somewhat in the long run, there will still be some sort of technology sharing here.
The XenSource acquisition by Citrix is good for Microsoft, because it allows Microsoft to buy some time until Viridian is ready. A year from now, Microsoft can say, “Oh yeah, we don’t have Viridian ready just yet, but look at this. Two of our primary virtualization partners have gotten together to field an ever more comprehensive virtualization product portfolio, which is integrated or will be integrated fairly tightly with Viridian when that comes out." So, we'll hear Microsoft saying, “We don’t have it all together today, but we have partners who can give you a fuller virtualization portfolio to compete with EMC/VMware."
If we look at Citrix's portfolio, every single piece, service, or product offering is matched by something Microsoft is pushing now. That, in essence, means that Microsoft is trying to acquire the business that Citrix has and slowly remove Citrix from the limelight and off to the sidelines. ... [Citrix] needed a broader strategy, one that wasn’t focused solely on access mechanisms. The acquisition of XenSource gives them a broader story.
If we look at just the idea of what XenSource was doing with their processing virtualization management security, particularly their recent announcement of partnership with Symantec for the Veritas Storage Foundation software to be included in XenEnterprise, you could see that Citrix starts to have a more top-to-bottom virtualization story than they every had before. So, from a product portfolio view, this acquisition appears to make some sense.
We've been dealing with issues of the Microsoft platform for a long time around resource management, where we're fighting with SQL Server and other applications or resources, and each one has different memory requirements. This virtualized environment allows us to focus on giving our application 100 percent of the resources, and thereby never running out of things like TCP/IP sockets or having memory thrashing errors slowing down wireless communications, which is critical to Web services doing their job. So, it’s having a profound effect on the production environment.
This is a really interesting acquisition that will help XenSource at least get more mind share in the enterprise. Companies obviously have lots of Microsoft investments on all their desktops. There's a good chance that major enterprises have significant investments in Citrix as well, if they've got any need for remote access for their systems, terminal services, etc. It will open it up to a few more environments to add in this virtualization capability for organizations that were still unsure about what to do with open source. It’s a good thing from that perspective.
On the datacenter side, the promise there is better utilization of resources. As I said, if you really want to get into it, you could find a way to tune Microsoft. But I have a sys admin working in one of my clients who is fighting with this 3GB initialization parameter. When he puts it in one way, one app gets hit when he puts it in there. When he doesn’t put it in, other apps get hit. But mine [virtualized] works fine.
This is a clear case of where you go out and get an additional operating system license and put this into each application and in its own virtual machine running on a four-way or eight-way Intel Duo Core 2 machine -- they are running a storage area network (SAN) -- and you have one of the cleanest, most high performing environments I've ever seen.
SOA is, in effect, heightening this issue, because of the need for discrete services running with their own horses and their own power. ... What virtualization does is let you set up that verified stack on your server and not have to worry about breaking it down the road, because it's sitting in it's own virtual environment.
On HP-Opsware and SOA Operational Efficiency
If you start getting into the automation space, the HP-Opsware deal is obviously the more interesting one. There's a natural connection between the virtualization space and some of the movements in the management space.
When you really embrace SOA, you're going to wind up with more moving parts for a given solution. And in doing so, you could create this management challenge of how to allocate resources for each of these individual services that have their own life cycle. There's a natural potential to move towards server virtualization to do that, so you can get your arms around that whole management concept. Where I've been disappointed in the management space, however, was that we really haven’t seen anything from the large systems management vendors to start tackling this problem.
So, if we are creating lots and lots of services -- you may now have 500 or 1,000 services -- you have to look at that and say, "I have a bigger management problem." There's no reason we can’t take the concepts of SOA and apply them to the management environment.
So, whether it's automated provisioning of solutions, automated policy management, a need to change SOA’s or enable more resources for a particular consumer, there is no reason that I shouldn’t be able to have my management systems call a service to do that. I may want to set up custom orchestrations for how to manage my infrastructure. I may want it all automated out of the box and just push a switch and have it happen.
In order to get there, we've got to have management services on all of our infrastructure, and that’s where there's a huge gap. Everything is still intended to be used by a person. Maybe with some creative scripting, people are able to do it, but you can compare it to the days of Web-enabling mainframes, where the technique was to screen scrape off the green screens. You almost have to do the same thing from the management side now. Look at these user-facing consoles, figure out what glue you can put in front of it to script it, and automate it.
I want to see my SOA installations able to speak to and hear from my datacenter systems management solutions. And right now, for that closed loop you were talking about, everything we see is the basic SNMP traps that may get read by Tivoli’s program. That lets you say, "Okay, there is an alert that one of my servers is overrun on memory. I’ve got these applications running on it, I am going to need to do something, and I see an alert that I can drill down on, and do some basic root-cause analysis."
That’s not enough for a true Business Technology Optimization (BTO) and being able to utilize the resources you are trying to marshal for a SOA environment. I want my Tivoli Application Manager to fully automate that process, look at the variables and the event stream coming from my systems, correlate that, and put it into some sort of context with my applications that are running on it.
I had a conversation with Paul Preiss, who heads up the International Association of Software Architects (IASA) about this very thing. He has raised a point and is trying to drive attention toward the exponential growth of SOA, as people start to add services and services become dependent upon services and organizations. ... From a product perspective, when you deliver a product as a SOA, you're delivering the architecture, and then you are delivering the implementation of that architecture. Therefore, you have a very controlled instance in which you can manage very easily without needing large governance controls, because you're providing all the infrastructure for management of that SOA as part of your application.
[However] when you have a lot of legacy infrastructure and legacy investment, and people without the sophisticated SOA architecture experience on staff start developing services, you open yourself up for potential disaster. In those instances, the organization has to ask itself, "Are we ready to invest here?" Every organization obviously wants to take advantage of the latest technologies. This is one of them that can really end up biting them if they are not careful. So they need to step back and think about what it takes to invest in SOA and start to wrap their legacy systems and make them available.
Opsware is very good at automating provisioning in the lifecycle, but it’s around the infrastructure that’s running the services, not the services themselves. That’s where the linkage needs to come. ... The vendors in the space -- the BMCs, IBMs, HPs, all of them -- have really missed this, and they’ve been lacking in explaining how they're really going to manage the services, because they're so fine-grained. Historically, managing an instance of an SAP application server is very coarse-grained, and that’s comparatively straightforward. But, when you're talking about disaggregating that and having application components everywhere, you have to disaggregate the way you manage as well.
You have it completly correct. Services management must be automated. Cannot be done by people. Factory automation faced this decades ago for the same reason. The precision is too much for people to manage. They touch it, they break it.
ReplyDelete