11-30-2011 06:51 AM
In my experience, HP-UX is the best Unix available; better than AIX and way better than Solaris. It’s not without its flaws, but in terms of resiliency, ease of use, and general capability it’s head and shoulders above the competition.
With that in mind, it is really disappointing to see HP be content with HP-UX being an also-ran in regards to market share. Recent communication hasn’t done much to make me believe that HP believes in HP-UX. Sure, HP says it’s committed to supporting HP-UX for many years to come, but at the same time has decided to help competitors gain market share by donating its mission-critical chops to Linux and Microsoft.
So, for fun (and because I’m up at 3:00 AM and can’t go back to sleep) I’m going to lay out some ideas I’ve had about how to bolster HP-UX in the market and start winning market share.
Give away developer tools.
HP should be making it easier for developers to write awesome, high-performance applications for HP-UX, not make them pay for the privilege. Make tools like aC++ and Caliper freely available for everyone. Yes, AllianceOne does look interesting (I don’t really know too much about it), but it is hoops a developer need not jump through on a RHEL box where a simple ‘yum install gcc’ will do the trick. By making HP-UX more developer-friendly a richer ecosystem of software for it will arise, which will make it a more enticing platform for users.
Include OnlineJFS in BOE.
While there might be some licensing issues to work out with Symantec, this would be a very good thing. How many other OSs currently charge to be able to resize file systems online? I know of no other.
Include Glance in all OEs.
Glance is awesome and is an excellent way to showcase the exceptional management tools available for HP-UX. Let people watch performance stats on individual machines with Glance for free, but still charge for centralized performance monitoring with OVPA/OVPM.
Port HP-UX to x86.
Blasphemy, I know, but hear me out. Create another OE called x86OE or something and place it below BOE in the OE hierarchy by stripping out DRD, VxVM, SRP, PRM, maybe Ignite, and stuff that requires Integrity hardware. Leave in OnlineJFS and Glance. Give the whole thing away for free.
The idea is to create a version of HP-UX that isn’t geared towards the mission-critical market, but towards workloads that are naturally more tolerant of individual server downtime; things like DNS, HTTP, SMTP. Sell support for it to folks who want it, sure, but make the bits free for anyone to download, play with, use, and update.
One of the reasons Linux and Windows enjoy a large amount of market share is because they are available for people to learn and learn to love; Windows because it’s on nearly every desktop and Linux because it’s freely available. Making an HP-UX Jr. that is free and runs on x86 is a win in two ways: it makes it easier for people to become familiar with HP-UX thus fosters popular support and it provides a natural migration path to full-blown HP-UX on Integrity for applications that do require mission-critical features.
Lean on Intel to make Itanium a priority.
Itanium has had a rough childhood, no doubt, and because of it the platform suffers from a bad reputation. Improvements made over the past few years have gotten Itanium mostly past its rocky start, but there is much room for improvement. Four cores per die is a nice development, and eight will be even better, but neither is that impressive. And while I know clock speed isn’t the only factor in processor performance, it’s hard to defend Itanium’s mid-1 GHz clock speed against Power7’s 3-4 GHz or even x86 and SPARC’s 2-3 GHz clock speeds.
There are more things that could be done, I'm sure, but these are the ones that have been floating around my head the past few months.