I think we need to look at support in terms of people instead of systems or devices. End User Support starts with the person using the tools, and works toward resolution by the most appropriate means. In some cases, that means simply advocating on their behalf to other groups who provide services used. It also includes fixing things, monitoring security and updates, advice and assistance for procurement and so on. The paradigm shift comes from defining the service in terms of the person and their job responsibilities.
This means viewing IT as a set of tools, a means to an end, instead of an end in itself. That's going to be scary for many folks, but if we truly want to be more efficient, open and agile we're going to have to redefine ourselves in terms of how our clients use the tools we provide and develop. That has been the guiding principle in our service development in SDI.
I used the drivers here: Precompiled JDBC SQLite drivers
The source is here, though it may not be the same: Source
I dropped the Linux JDBC and .so files into [cfmx7root]/lib/ and restarted ColdFusion. Then I created a database using SQLiteQuery on my Mac and added a table, column and single value. Then I copied the db file to a test folder on my server and just ran a select * from table.
My datasource setting is as follows:
JDBC URL: jdbc:sqlite:\var\www\html\sqlite_test\cfmltest Driver Class: org.sqlite.JDBC Driver Name: sqlitejdbc.jarAt that point, running the code threw an error that the table didn't exist. Permissions were all okay, so I executed a CREATE TABLE via the .cfm file with a different name, column and value. That table showed up. Turns out that the table I created on the Mac only shows up on the Mac via the GUI tool, and the table I created via CF only shows up for CF, even if I query the master table to get a table listing.
I'm thinking that has to be something to do with driver versions. If I get a chance to test further, I'll run some speed tests versus reading and writing files, XML, etc. to see what the speed is like. This is also obviously insecure, you don't want your db file to be readable by the web server.
UPDATE: now that CF8 includes the Derby java database, this is somewhat moot. Though it does give a way to interact with existing sqlite databases. I still don't know why some tables don't show up, again probably a version/driver issue.
- Creating virtual labs that can be recreated anywhere, even on a single server
- Deploying custom toolsets to remote locations as images
- Sharing whole development environments among developers
- Disaster Recovery and backup
Our service is in beta now and we're planning a launch in early March. More on that to come.
The last major component of the VM Service infrastructure should be in place December 12. EMC is coming to configure and power up our SAN, and work with us on our fiber network and tuning for VMWare.
We have been discussing pricing and comparing our service to others. There has been quite a bit of discussion in and around ITS lately about Google and Amazon and which services we should build and run versus outsourcing. I believe that there are services that we can and should offer within PSU, either centrally or as cost-recovery, and that there will continue to be. The key is that we have to evaluate the service based on aspects that can only be offered or implemented by ITS within our networks and systems. We must also consider cost, and always be mindful of overall savings for the University.
When designing the VM Service my target has been cost-effectiveness relative to running a physical box and staff time to maintain and secure it. Can we save people time and money and at the same time make it easier to implement custom operating environments for administration, teaching and research? Can we offer features and services that can't be offered by any external vendor or service? We'll be running a limited beta in the next few weeks to determine what we can and should offer, and how to configure the VMs by default.
The storage array has arrived and we are scheduling the setup. The holdup right now is electrical installation in the new server space. We're also in the process of acquiring the tape backup system for DR.
I'd like to have everything up and running by Thanksgiving, but that will depend entirely on the status of the space.
www.work.psu.edu has a new interface, and I really like it. As I was checking it out today, I thought about the fact that we don't seem to encourage each other much within ITS. I don't know who exactly worked on the new site, and it's probably a case in point that I don't, but here's a big thumbs up to you.
There are many large, complex systems behind that pretty new face, and I keep thinking about how much of what we do in ITS gets lost because we don't seem to have a good handle on the face of data. Google and Apple don't discover fire, then invent Zippos. Somebody has to discover fire, and keep it burning, but when you cook something with it you'll capture your audience.
That was a really belabored metaphor for a simple idea that I'm trying to express. We need to get better at exposing the really cool stuff that we do to the "average" users in our extended community. Architects and engineers really need to start to understand HCI and interface design, even if they never actually build an interface. I'm sure that's not a popular sentiment, but there it is.
Jimmy V. has a couple of posts on his blog that I really want to comment on. First, I like Jim. We were in ITLP together and I respect his experience in software development. He does good work.
Second, I agree with many of his points. I've worked on more than a few web applications that brought interface and design experts in too late or not at all and ended up being unused and/or harshly criticized even though the application logic and data model were very well done. I've also seen web apps that had good designers who were asked to put lipstick on a pig.
Having said that, Jim and I are very different in our philosophies. I build web applications, though I really don't like HTML. I build databases, business models and application logic, though I do dabble in interface design on occasion. I really believe that web applications, and the types of applications that they're evolving into, are going to dominate ITS development for the foreseeable future. Furthermore, those applications are being built in dynamic languages that are interpreted rather than compiled. That's why the model presented in this article is much, much closer to the way I work. Most of the coders, designers and architects that I talk to and work with work this way too. In fact, I doubt that I could code without knowing what problems I'm solving and seeing the bigger picture.
The web has issues, but the paradigm of rapidly developed (and rapidly replaced) data-driven, networked applications is here to stay. It seems that so much of the discussion in ITS over the last couple of years has been about meeting student and faculty needs and anticipating demand. I think I can sum that up for you - any information I want, any time I want it, on any device I choose to use. I'm loving that challenge.
The new Penn State Undergraduate Bulletin Online is going live, the previous site is slowly being phased out as links are changed to the new URL.
The site is built on ColdFusion 8, Red Hat Linux AS 4 and MySQL 5. I built the database and wrote the application logic, while the design and standards compliance was handled by Web and Media Design Solutions. The course data comes from the University Course Master, while the majors and other degree programs are HTML files with metadata in the db. I built a very simple CMS to allow the text stewards to maintain the programs and general information sections.
We've already had requests to integrate with other projects in ITS and share data with groups like the Registrar and Commonwealth Campuses. I think this is an excellent opportunity to model new ways of collaborating and especially exposing information in the University's enterprise infrastructure to the web and other channels. I think that we will start to see quite a few new projects in a similar vein in the coming months, as more departments and groups look for ways to share information and meet student needs (and demands).
I worked with quite a few talented and dedicated people on this project, including the sysadmins in my own group, Web and Media Design, AIS and Undergraduate Education. People like these make Penn State a great place to work.
After months of planning and discussion, our virtual machine hosting service is moving to the procurement and setup stage. We're still dependent on the completion of our renovated server space in Willard, but we'll be ready to move in and set up by mid-October.
There seemed to be a <collegewords> serendipitous confluence </collegewords> of circumstances that made this project not only possible but very attractive right now. It seems like virtualization is on everyone's radar, when discussing power and cooling for a server space the engineers asked me several times if we've "looked into virtualization". I explained that we are in fact building a virtualization infrastructure, but even VMs need power and cooling.
Our model will be a "virtual colocation center". We will charge a setup fee and monthly maintenance per VM, with optional contract and hourly consulting. The setup fee will include OS install (Red Hat AS 5 will be our default, with first year updates included), real IP, DNS registration (psu.edu), 1 year SSL certificate, WebAccess registration, and install/configuration of any other software the client may want to run. Access will be via SSH, and we're testing integration of the Virtual Console with Kerberos to allow console access.
I'm thinking our customer base will primarily be small groups who need custom web applications or CMS, researchers with grants who need short-term application or database servers, or IT shops who don't have the time or resources to handle every request for a web site or mail server that comes along. This service is intended to complement the volume license agreement that the Computer Store is preparing with VMWare. i can't really discuss pricing yet, but at this point the service is looking decidedly less expensive than purchasing a physical box, housing it, securing it and hiring someone to administrate it for an individual. This might help to alleviate the "boxes under desks" problem that many departments are having, and Commonwealth Campuses may be interested in quickly deploying DR or alternate servers at UP.
We did consider apportioning processor, memory and disk resources per client instead of actual VM instances, but given the nature of the service and the clients we're targeting, the value-add of setup, registration, etc. seemed the more appropriate model. We're not just offering VMs, we're offering full-service deployment and maintenance as well. Of course we'll evaluate this as the service matures.
I'll post some updates as the service develops. Perhaps an overview of the technology we're using and performance tests from some of our beta customers.
Recent Comments