Sun's Desktop Selection Guide and Pusillanimous Counsel
Dec 18, 2003 8:32 AM PT
The word "pusillanimous" is normally used to label actions judged to reflect a lack of moral courage on the part of the person taking the action. Thus, people who can be rather easily pressured to undertake actions contrary to their own beliefs are often said to be pusillanimous, cowardly or weak-spirited, while those who advise "getting along by going along" are said to offer pusillanimous counsels.
It might not be a word you want to drop into casual conversations about middle management. But if you do, be sure to have someone send a video of the reaction you get to the people at America's Funniest Videos.
On the other hand, it's not just the perfect word to describe a recently released Sun document titled the Desktop Architecture Selection Guide, it's the only applicable adjective the editor will let me use here.
Tilted the Wrong Way
What's most disheartening about this "blueprint" is that it looks as if someone took an older source document that probably did a pretty good job of stating the case for the Sun Ray sometime in late 2000 or early 2001, then mealy-mouthed it with some minor updating for corporate release in November 2003.
Thus, references to StarOffice are to version 6 and 6.1 (June 2002) with embedded links to the version 6.0 FAQ, while user authentication is linked to the federated services facilities from the mid-nineties. In fact, by word count the document's emphasis isn't on the Sun Ray at all. It's on running Windows applications like "advanced directory" on traditional Wintel clients like Citrix or Tarantella licenses.
It was reasonable, in 2000, to express some ambivalence about leaving the Microsoft client-server architecture behind mainly because critical office applications weren't yet up to the job. Today, however, that's not true. Sun's Java enterprise desktop meets most requirements quite nicely. As a result, an updated version of something that provided a reasonable balance between Wintel and Unix applications in 2000 should now lean far more heavily in favor of the Sun Ray.
Instead, the updater appears to have focused on the Windows side while assuming stability on the Sun Ray side. As a result, a document that probably started with a reasonable balance is now tilted heavily the wrong way.
Case Is Compelling
An architecture document intended to help people configure and deploy the Sun Ray should assume that the reader is committed to the architecture and needs help getting it in place. Thus, just summarizing the cost and productivity argument for it instead of building an extended case makes perfect sense -- and the document's authors do provide a few paragraphs summarizing the case against using Wintel, but they don't proactively make the case for the Sun Ray.
In reality, the case for the Sun Ray is utterly compelling and needs to be told with the kind of blunt good humor characteristic of Sun's highly unofficial BigBlueSmoke.com.
For a small business with less than 100 employees, the case for the Sun Ray comes down to this: All the software needed for office automation, Web services, secure document management, remote access and IT management is free. The hardware itself is marginally cheaper to buy than the Wintel equivalent but lasts far longer, and operating costs are a small fraction of their Wintel equivalents.
Gains in Flexibility
A minimalist install with 100 Sun Ray 1g displays using Dell 19" LCDs could be split evenly between two Sun 440 servers for a total of just under US$2,000 per user, including network gear and a couple of printers. That would provide redundancy along with room for accounting or other applications while requiring only one part-time sysadmin. The whole setup would probably run for something like five years with only software updates.
For a busy office with a complex online application -- like a time-based billing system and integrated VoIP communications -- those 440s would be marginal. Replacing them with six-way V880s improves performance, allows more room for growth and reduces the operational impact of having one server shutdown but raises capital cost to just over $3,000 per user -- just about what a Wintel environment would come to before business application licensing and support.
For larger businesses, the case is typically even better, but the importance of a small capital cost savings and even the substantial savings in support pale in comparison to gains in flexibility and user productivity. The reason is that getting rid of the Microsoft client-server architecture also means getting rid of all the organizational structures -- like the help desk and the consequent separation of application expertise from operational support -- put in place to support it.
Deploying and Supporting
Of course, an architecture guide should fundamentally be about what it takes to deploy and support the Sun Ray desktop. When I first downloaded the document, this is what I skipped to first, only to find stuff like the following:
There is no simple guideline for the ratio between Sun Ray ultrathin client devices and server resources required to support them. Guidelines provided by the Sun Ray ultrathin client-engineering group suggest that each Ultra Sparc III CPU can support 25 concurrent Sun Ray ultrathin clients. Such a guideline, however, ignores other software that is running on the server and user interaction with the server. In a real-world situation, the typical user might be running a Tarantella or Citrix native client to present Windows applications and might also be accessing an Internet browser and an office productivity tool (for example, the StarOffice Office Suite), running on the same or a different server. Experience in the field suggests that some of these tools might require frequent screen refreshes caused by user interaction or, in the case of the Internet browser, caused by Shockwave animations or streaming video.
This was followed by truly brilliant and Solaris-specific advice like the following:
Always use a server with at least two CPUs whenever possible. This avoids the situation in which a single thread spins on a CPU and locks out other Sun Ray ultrathin clients. A thread is a section of executable code that is recognized by the operating system as an executable entity in its own right.
Not to be pusillanimous or even pugilistic about it, but this sucks. Sun should withdraw and revise this document instantly.
Paul Murphy, a LinuxInsider columnist, wrote and published The Unix Guide to Defenestration. Murphy is a 20-year veteran of the IT consulting industry, specializing in Unix and Unix-related management issues.