View All Videos
Operations Above the Level of a Single Device

Operations Above the Level of a Single Device

Damon Edwards / 

Tim O’Reilly has another excellent post about the concept of “software above the level of a single device”. While most of his post is dedicated to famous consumer-oriented client/services offerings like itunes, if you glance around our industry it’s quickly apparent that the concept of “software above the level of a single device” is the predominant vision amongst leading business strategists and software architects. But once that thinking is translated into actual code, the care and feeding of those innovative services tends to look a like like decades old box-by-box thinking. Lots of hand-editing of configs, lots of manually maintained procedural scripts, lots of jumping from log file to log file looking to divine the root cause of a problem or outage, lots of pain and inefficiency everywhere.

As the elegance and distributed nature of business ideas and their resulting software has improved, the the prevalent thinking about how you support it surprisingly has not. Sure there are projects like ControlTier, SmartFrog, Puppet, CFEngine, etc. that deliver on the promise of a better way, but the collective thinking of the industry has a long ways to go before these kinds of solutions become standard. Maybe we need a catchy tagline to close that gap…. “Operations Above the Level of a Single Device” anyone?

Sun and IBM virtualization announcements are more than just a move on VMware

Sun and IBM virtualization announcements are more than just a move on VMware

1

Damon Edwards / 

Both Sun (launched with Oracle’s PR help) and IBM have been making big waves in the virtualization space. On the surface these are clear moves on VMware/EMC, the current collector of the the vast majority of revenue around virtualization. But under the surface there are some other interesting things going.

First, does anyone else agree that this is a way to give the “on-premisis” ecosystem (including hardware sales, services groups, and partners) a way to fight back against the scalability and manageability promises of SaaS alternatives? It’s a safe bet that some form of a “do-it-youself SaaS is good” message will be hoisted upon the industry in the upcoming year.

Second, both initiatives talk a lot about “data center management”, which is generally code for deployment automation and monitoring. I’m glad to see increasing recognition of this problem space. But I am surprised that very little has come down the pipe in terms of new development on true deployment or configuration management tooling. The default answer still seems to be to let Global Services (or one of Sun’s services partners) come in and build a custom solution from a grab bag of tools.

Of course their are open source tools like ControlTier (service provisioning targeted at developers and deployment engineers ) and Puppet (system provisioning targeted at systems administrators), but you hear little from the industry titans about the need for these types of tools. Follow the money for the answer why.

Package-Centric Application Release Methodology: What is it?

Package-Centric Application Release Methodology: What is it?

2

Alex Honor / 


Spend some time in the trenches at a software as a service (SaaS) or e-commerce company and you’ll find a prevailing opinion that traditional release management processes don’t keep up with the rapid pace of application changes and rising system complexity. SaaS and e-commerce environments are the ultimate example of IT moving from being a supporting cost center to the actual source of revenue production. For SaaS and e-commerce companies, the application release process is what governs their “factory floor” and this process needs to be run with as much predictability, measurability, and automated efficiency as any modern manufacturing process. To make this governing process run smoothly you need to handoff well defined units of change from one part of the process to the next, and this is where a package-centric application release methodology comes into play. The package-centric application release methodology augments traditional change management to help implement and execute change.

Traditional vs. Package-Centric Methodologies
Traditional release management methodologies are dominated by human process definition and organizational workflows that govern how change tasks should be controlled and who should carry it out. Toolsets to support these traditional methodologies are centered around task approval, auditing, and planning. In other words, a traditional release management methodology is all about managing people and their activities. These human-oriented workflows are important, but on their own they do little to solve the inefficiencies and complexities that plague technicians in SaaS and e-commerce companies when the time comes to execute procedure to carry out those changes. Plainly put, you can get a good handle on what all your people are doing, but you don’t get a good handle on how their changes are technically performed.

In contrast, a package-centric release management methodology focuses on how to facilitate collaboration of decentralized groups to coordinate the implementation of change. Under this methodology, the product of each team is released as a standard unit, one that contains instructions on how to apply the change as well as any relevant content needed for the change. These standardized units of change enable an automated change management approach that rationalizes and coordinates changes originating from disparate teams but each potentially effecting the greater application service operation. In other words, package-centric release management methodology focuses on format, modularity, mechanisms, and technical workflow all in the context of the larger integrated application.

The basic element of the package-centric methodology is, not surprisingly, the package. The package concept prescribes a standard unit of distribution as well as standardized methods for installation and removal. The essential benefit of formal packaging is the provision for changes to be predictably migrated, done, and undone. A package also carries with it essential information like what version it is, who made it, and what does it depend on. Ultimately, packages become the common currency of change within the IT organization.

What belongs in a package?
One often imagines a release to be a distribution of an entire application. In reality, for online systems, change happens at a much more granular level. One typically observes several kinds of changes released to live environments:

  • Code: Application files executed by the runtime system. This could be compiled objects or interpreted script.
  • Platform: Files that comprise the runtime layer. This is generally server software like Apache, JBoss, Oracle, etc.
  • Content: Non-executed files containing information. These could be media files or static text.
  • Configuration: Files defining the structure and settings of an online service. These could be files for configuring the runtime system or the application.
  • Data: Files containing data or procedures for defining data. These could be database schema dumps, SQL scripts, .csv files, etc.
  • Control: Configuration and procedures consumed by management frameworks

Within the package-centric paradigm, each type of change noted above is bundled, distributed and executed via a package.

 

Standard vehicle of change
The package serves as a vehicle of change, both as a unit of transmission between teams, but also as a standard means to affect the change within the context of the running application service. The package concept helps cope with rising complexity by encapsulating change-specific procedures and underlying tools behind the package’s standard interfaces. In an object-oriented fashion, the package provides the standard interfaces and encapsulation needed to run efficient operations. Teams no longer have to understand the internal structure and procedures in order to apply another team’s change.

Scaling up
Because packages establish a standard interface to distribute and execute change, an IT organization can more readily scale up operations. Knowledge transfer no longer needs to be done between individuals but can instead be explicitly specified within the package. By declaring the essential knowledge and process within the package, teams can more quickly execute change by avoiding repeated inter-personal communication. Because changes are packaged, they can be stored in a repository to facilitate auditing and reporting. Package-centric change also makes it possible to leverage a generalized automation layer to rationally execute changes en masse. Automation is key to gaining the efficiency needed to run any SaaS or e-ecommerce application.

The human workflow emphasized by traditional release management methodology is an important component to governing the business service’s “factory floor” . The package-centric methodology picks up where the activity planning and coordination leave off by assisting the technicians to distribute and apply change.

In future posts I’ll discuss the various aspects, benefits, and examples of the Package Centric Release Methodology.

Hey internal IT, you’re a SaaS provider too

Hey internal IT, you’re a SaaS provider too

2

Damon Edwards / 

If you don’t provide service to external users are you still a “Software as a Service” (SaaS) provider?

As the term SaaS becomes increasingly popular, I hear this question more and more. Here is my $0.02:

Yes. Saas is about the relationship between the service provider and the service consumer. The physical characteristics of service delivery are almost inconsequential.

In any SaaS relationship there is a service provider and a service consumer. In my opinion, the defining feature of SaaS is that the business contract between the two parties is the sole recourse the service consumer has to manage the service (aside from in-application config options). The service consumer has no hand in running or maintaining the service other than what they request via the contract.

Over the past few years there has been a major push inside enterprises towards being “service-oriented” (and ensuring clear lines of demarcation between business units and IT units). The end goal has always been for business users to be able to establish a business contract on which IT delivers a service. From the both the internal IT and end user point of view, SaaS is just another acronym on this service-oriented path on which they have already been traveling.

So internal IT, congratulations you’re a SaaS provider too. In defining a SaaS world, focus on the relationships… physical characteristics of service delivery and who signs who’s paycheck are just distractions.

Knowing Where You Are Going: Identifying ITSM Requirements and Maturity

Knowing Where You Are Going: Identifying ITSM Requirements and Maturity

1

Damon Edwards / 

Charles Betz’s ERP4IT blog has an interesting post on what Betz calls “The Four Pillars of ITSM”. It is an interesting read because not only is he spot on, but his “pillars” also appear to reflect implicit stages of maturity

The “pillars” (see his post for the details):
1. A service-focused view of IT
2. Process orientation
3. Master data management (business and internal IT data)
4. Well-architected internal IT

The maturity levels I see in this would progress something like this:

First you wake up and realize that you operate software as a service. This often results in a bit of an upper management hand-wave, but none the less, it’s a critical step because it means you recognize that you have to change how your organization runs. (Pillar #1)

Then you realize that the service lifecycle actually cuts across many different disciplines and you need coherent processes that span them all. Once you have those processes defined you also need to know how to measure their effectiveness. (Pillar #2 and #3)

Finally you need to bake those processes and methods of measurement into your IT systems (Pillars #3 and #4)

Note: Pillar #3 is the linch pin between #2 and #4.

Page 24 of 26First2223242526