Getting Fed Up With Sun: Can’t Get Systems, Breaking Existing Ones
Posted on April 16, 2007
I’m sooooooooooo tired of making excuses for Sun. I just got two X4200 M2’s, not because I want them but because Sun can’t get Joyent X4100’s. Why Sun can’t produce systems to meet customer demand constantly amazes me! There aren’t many people inside or outside of Sun who evangelize the company as much as I do, but I’m tired of getting people fired up about systems and then hearing they can’t get systems!
Not only can’t Sun produce systems to meet customer demand, they cheap out and break the existing models. My X4200 M2, which I didn’t want in the first place because of its size, has 2 NGE’s and 2 Intel E1000g’s!!! This is absolute bullshit. The X4x00’s should all have 4 E1000g’s. Thats what I want, thats what I run, thats what I need. I refuse to run NGE’s on an $5,000 server!
And while I’m at it, SunSolve is a pile of crap. I’ve waited year to see it improved, but nothing. Its utterly useless. The search page is integrated with the Sun.com site search which makes you wonder why you pay so much, plus most of the information is ancient and or impossible to find. Doc.Sun.com has useful information but the site is a pile of crap, difficult and archaic to navigate, and no one is doing anything about it. For years I’ve heard things will improve, but here in 2007 still nothing has been done and nothing looks like it will.
I want X4100’s, NOT M2 BULLSHIT. I want lots of them and I want them quickly. I want a SunSolve worth paying for. I want a docs.sun.com that has been updated and more easily navigated than what we had 5 years ago. And most of all, I don’t want to keep hearing that Dell doesn’t have these problems!!! Damnit Sun, lets get it together and fix these things and lets do it NOW.
UPDATE: My fellow Joyeur, Mark Mayo, just wrote up a sister-post in the Joyent company blog: How to completely ruin a great piece of server kit (regarding the Sun X4200 M2)
UPDATE2: I’m getting an absolute flood of mails from people expressing the same frustrations that I mention and that we see in the comments here. Sun Executives are aware of this and responded immediately from the very top down. Sun may have its problems, but one thing that I’ve always found to be true is that they are forthright, honest, and responsive. I’m extremely appreciative of Sun’s understanding and response. Remember, they can’t change if we don’t make our demands known! Just switching to Dell doesn’t send a clear message unless you tell them why you’re switching to Dell. And what’s clear is we don’t want to buy Dell, but rather people feel they don’t have a choice. But we do! Sun is our company! Let’s help it be successful and drive it towards excellence. Remember, despite it all we’ve got one hell of a base to build on!
UPDATE3: Due to my crappy spam handler many people commenting are getting “Spam is not appreciated” notes. I’m sorry about that. Please send your comment to firstname.lastname@example.org and I’ll post it for you.
UPDATE4: I’m now talking with a Platform Architect in NVIDIA’s chipset group. Its awesome to see wheels turning. Lets see if something happens though, but still I’m encouraged!
UPDATE5: Wes Adams, Corporate Account Manager, Sun Enterprise Sales Group, has helped us overcome our supply issues helping to push through pending orders that previously were 3 weeks away, now the system will be in our VAR’s warehouse on Monday morning at the latest. We’re very pleased and appreciative.
UPDATE6: I posed the architectural decision to use NGE to Mr John Fowler who confirmed my theory that “Its there, we should use it” (my words, not his), and he went further to pose the question to both Ali Alasti and Andy Bechtolsheim. The reason for NGE is to both reduce chip count (and thus reduce heat, reduce power consumption, and improve reliability) and cost. Ali added “The future of x64 chipset is in direction of further integration of all types of functionality including GE.” There are no plans to use NGE on X4600 or X4600M2. I suggested that the data sheets explicitly declare what onboard solutions are provided, as the current data sheets say only “4 Gigabit Ethernet Ports” and further to consider, if technically possible, to at least swap the NGE’s and E1000g’s so that interfaces 0 and 1 are E1000g and 2 and 3 are NGE, thus allowing the most commonly used ports to be E1000g and to also better leverage the far superior PXE Boot Agent of the Intel NIC’s. Andy Currid, Platform Architect in NVIDIA’s chipset group, was glad to have feedback on the PXE boot ROM (Boot Agent) and is already working with Ali on solutions. Even though the Boot Agent is the thing post causing me pain, I’ve still made it clear that I’ll always consider a platform with differing onboard NIC’s to feel “cheap” and avoid such systems whenever possible. Whether or not this will have any impact on future designs I’m unsure, but while I have strong opinions I’m not about to dictate to Ali or Andy, I do trust their judgment.
I want to thank John Fowler, Andy Currid, Wes Adams, Johnathan Schwartz and everyone at Sun for being so responsive to my inquires and taking all of them very seriously. I’m very greatful.