Billy Marshall was founder of rPath a few years ago, pioneering the delivery of software appliances. Prior to that he spent several years at Red Hat as creator of the Red Hat Network and then as VP, North America Sales. Marshall is currently spending the summer as a consultant at EMC in its Cloud Infrastructure Group.
In Part 1, we focused on rPath's view of software appliances (SWAs), hurdles for Linux vendors that want to get into the SWA business, which vendors have the best chance of succeeding, and two important topics: pricing and support. In Part 2, our focus is on which applications are likely to make their way into SWAs, how some large ISVs are creating SWAs around their applications, and the relationship between SWAs and clouds. (Marshall wanted us to note that the comments made in this interview are his own and do not in anyway reflect rPath's views.)
Linux.com: Let's talk about ISVs and which applications are more likely to make their way into SWAs. Which ISVs did you go after and why?
Billy Marshall: Most folks who have traditionally shipped appliances generally already have some approach for lifecycle management of the software in place, developed by some engineer inside the company. It is a difficult sell to these folks and it is rare that you can switch one of these application owners. At rPath we switched a couple of BSD-based folks over to Linux. But it was rare. They would look at our system and say, you have optimized for lifecycle management and it is better than what we have, but its different than what I‚Äôm doing. That means that I have to invest in change and we‚Äôd rather not change. So we went after those ISVs that did not have a strategy already in place because they are confronted with the challenge of delivering their applications atop a virtualized stack. And they can either depend on their customer to virtualize the stack and do it wrong (the ISV would then have to support the customer and his errors), or the ISV could do it and get few, if any, support calls.
Linux.com: Are there any particular types of applications that are more likely to be applianced than others? IT infrastructure, Web infrastructure, etc., appear to be the most popular.
Marshall: I agree with that. Things that have historically been appliances are good candidates to be virtual appliances or SWAs. The only really interesting case of a SWA is the virtual appliance. This is because people just don‚Äôt want to trust anybody to lay something down on hardware. They want it to hit a hypervisor. We (rPath) allowed appliances to be installed on hardware, but that was the least favorite option. If ISVs were going to install SWAs on hardware they did it before the SWA left the factory, and it was a hardware appliance. They were not going to let the customer do it. They might give customers some choice of hardware configurations, but they were not gong to let the customer do the installation.
The best workloads are those that have classically been appliances such as firewalls, business intelligence, analytics, search, etc., not I/O intensive but typically more compute intensive. Workloads that have batch oriented workflow so that they can come and go on the network as they need to come and go. For example, SAP R3 will always run by itself on the hardware, but it might have BI stuff, CRM stuff, and analytics around it. It makes sense that these surrounding applications are delivered as SWAs.
Linux.com: Are ISVs with large, expensive applications such as DB2, Oracle DB, and etc. creating SWAs around their applications?
Marshall: Much of IBM‚Äôs middleware is showing up on Amazon as IBM AMIs (Amazon Machine Images) on SUSE Linux Enterprise Server. If it‚Äôs showing up there it is a virtual appliance. The pricing for IBM AMIs is by hour usage and server instance size. If you go to http://aws.amazon.com/ibm, you will find that DB2, Informix, WebSphere sMash, Lotus Web Content Management Server, and WebSphere Portal Server are all listed as AMIs with pricing by the hour and by the type of server instance on which it runs. For example, IBM DB2 Workgroup Edition is priced around $1.30 per hour for a standard large server instance; whereas, IBM WebSphere Portal Server is priced at over $6 per hour for a standard large server instance. You can actually charge the same or more for SWAs as you do applications that run in the traditional manner on a server because you will save on services and support.
Linux.com: SWAs and clouds seem to be tied closely together. Give us your view.
Marshall: Platform as a service is where the provider specifies the policies and language and everything else. Infrastructure as a service is where the hardware is abstracted away from you, but you are able to bring your software stack to bear on the infrastructure. This is what I call a cloud. This is what Amazon does. Some people call platform as a service a cloud.
The fact that SWAs and virtual appliances are engineered to be loosely coupled to the infrastructure by definition fit closely to what Amazon is calling a cloud infrastructure--the applications come and go as demand dictates at a variable price. I think that cloud represents the infrastructure side and virtual appliances represent the application side of clouds.