The top Sun Ray myths

One of the lessons I believe we can learn from productivity change in the telecom industry is that centrally provisioned services can be extremely attractive to consumers provided that the service provider does not impose artificial controls on what the user can do with the device.

My favorite tool for implementing this idea in IT combines Solaris on SPARC with Sun Ray desktops and the devolution of systems control (as opposed to systems operation) to user management.

I'll talk about the management aspects of this a bit next week, but for now thought I'd recognize the reality that very few IT people have ever actually seen, never mind used, a Sun Ray by summarizing the top reasons people put forward to argue that it's not as good as a PC:

1. if the network goes down, you go down.

   This is true. You should note, however, three things about this claim:

   • the problem affects all network dependent systems, including PCs whose users need central authentication, web access, or server stored documents to work;

   • with the typical Unix distributed data center strategy a local failure taking place anywhere on the network inward of the first switch or hub the Sun Ray connects to, will have little or no discernible effect on operations; and,

   • outside of consumer telco or cable company operations, network failures not caused by PC style switching are extremely rare.

     I haven't been able to find out how many user-hours are lost to network issues (other than planned construction or moving) per year at Sun where more than 33,000 people use Sun Rays every day -because the number is apparently so small it's either not considered reportable or not specifically tracked. Certainly no user I talked to could remember it ever happening.

2. laptops are a requirement.

   There is a Sun Ray laptop, but I don't normally recommend it. Instead I usually suggest that laptop requirements be carefully reviewed and appropriate Apple, Linux, or Wintel gear issued where needed.

   In most cases such reviews produce non obvious results: today's iPhone (and tomorrow's clones) can do a lot of what laptops used to be needed for; most people who claim to work effectively on the road, don't; most people who want laptops as desktop replacements are really trying to escape centralized IT controls on software change and/or shared licensing; and, most client-site use of laptops poses significant but unquantified risks to both proprietary and client data.

3. screen space and resolution are limited.

   A few people who send me email have experience with the earliest Sun Rays - and 13 inch early laptop style screens. Those days are long gone - now Sun's "thinguy" puts pictures in his blog showing dual and quad head Sun Rays displaying Windows server software running at 3840 x 1200

4. there's no software.

   The Sun Ray runs no user software - it's a just a user interface device displaying whatever it's sent and capturing user input for transmission to the application. As a result it is currently usable with essentially all Windows, Linux, and Solaris software - and can handle Mac software written for use with the X11 interface too.

   In other words, the device is usable with more software than anything else out there.

5. the fact of central control renders systems using this as unresponsive as 1970s data centers.

   It is true that IT tends to centralize control - and it is also true that in the Windows world larger organizations have no other reasonable choice because decentralizing control is simply too expensive both in terms of support and in terms of the "security" risks inherent in any PC use.

   However, how a Unix/Sun Ray data center treats its users depends on its managers, not the hardware. You can centralize both control and processing (and most people do, because that's how they've been trained), but you don't have to..

   What you can do, instead, is centralize processing while decentralizing control.

   Remember: the business costs and usage constraints imposed by the 1970s IBM data center; costs and constraints that are now being re-invented and re-applied in big organization Windows environments, are artifacts of cultures, costs, and software limits which simply don't apply in the Unix/Sun Ray world.

   In other words, you can get this effect: but it indicates managerial incompetence with respect to the technology and is not a necessary consequence of the technology.

6. it can't handle video conferencing, movies, sound

   Yes, it can.

7. everything costs more - support people, hardware, software

   Quite the reverse: if you've got a thousand desktops, anything you need for any significant chunk of them will cost less to implement on Unix/Sun Ray than on any other environment.

   The people side of this argument is particularly compelling. It's true that for every good Unix sysadmin you can hire you'll get ten or more MCSE style applicants hireable in a Windows world - and that the sysadmin will cost you at least a third more than the Windows people. On the other hand, appropriate staffing ratios for Windows are much higher than for Unix: so for a thousand desktops you'll probably need about 20 IT FTEs for Windows versus 4 or 5 for Unix - and one of those would really be an understudy and holiday fill-in kept on in case someone else gets hit by a bus - or, worse: a spousal demand requiring off-net travel.

8. It's no good for graphics intensive processing or any other application requiring lots of desktop processing.

   This is true: Sun Ray is best suited to a shared processor environment and performance limited by bandwidth to the server.

   There are, however, two things to note about this:

   • first this is a consequence of the implementation, not the computing model. Sun's long gone NeWS sent PostScript to the terminal and was ideally suited to collaborative engineering because bandwidth requirements were only marginally affected by imaging complexity.

   • second it's only true for a fairly narrow range of requirements in which the display graphics computations significantly outweigh the back-end calculation and/or data access requirements.

     If, for example, your application relies on a grid for the back end calculations then there's a scaling point on the computation above which giving every user a Sun Ray and depending on a central imaging server will provide cheaper and faster access to data than doing display on dedicated workstations.

     This is outside most people's experience - but consider the application of this idea to something as pedestrian as e-mail processing. Give 50 people access to Microsoft Exchange on a Xeon via desktop computers and they'll get one or two pieces of mail well before colleagues using Sun Rays on a 5220 T2 server will. Make that 5000 pieces of email, however, and the Sun Ray users will get much better performance than the PC users.

So what's the bottom line? most of the more widely cited negatives for Sun Rays are pure nonsense. Sun Rays are not the right answer for a hobbyist who wants to play games, surf the web, and pretend to program - but for large organizations the negatives simply don't hold up - and we haven't looked at the positives at all yet.

(Note: I'm "off-net" until Sunday (a victim of spousal abuse!)) and won't be able to respond until then. Meanwhile Sun's Craig Bender - Mr. Thinguy - will be responding in defense of the Sun Ray. )