UNIX Consulting and Expertise
Golden Apple Enterprises Ltd. » Archive of 'Apr, 2009'

Larry Ellison on Cloud Computing Comments Off on Larry Ellison on Cloud Computing

I’m very much indebted to @stewarttownsend for linking to this on Twitter. During an analyst call Larry was asked for his views on Cloud Computing. This is his response:

There seems to be a tragic loss of high profile CEOs who feel able to speak their mind fully. Scott McNealy was great back in the day, and Larry is still not shy about holding forth – but apart from that, the IT industry lacks some really good leadership these days.

Oracle have bought Sun 1 comment

Well, this was a bit unexpected – but they were one of the only other companies who could realistically make a move.

Oracle have purchased Sun, at $9.50 a share, in a deal valued at approximately $7.4 billion, or $5.6 billion net of Sun’s cash and debt.

Larry Ellison has shown for years he has the determination and drive to push a company forward – either by crushing the opposition (a favourite term of a DBA friend) or by buying them out.

Owning the whole stack – from storage, through tin, up to software – plus services will give Oracle a huge amount of leverage and visibility. It’ll also mean they can take on IBM directly at pretty much every level.

Interesting times ahead.

You can read the Sun press release at http://www.sun.com/third-party/global/oracle/index.jsp

Update: Oracle have put up a nice page of FAQs and information at http://www.oracle.com/sun/index.html

A review of the Sun Ultra 24 workstation Comments Off on A review of the Sun Ultra 24 workstation

I’m in a position to start ramping up my involvement in OpenSolaris, which has put me in the market for a new workstation. I’ve just had a shiny new Ultra 24 delivered as part of Sun’s extensive Try Before Buy program – so here’s a quick review.

Some Background

Back in the 90s Sun was wildly successful with their UltraSPARC II processor. They had a good crossbar architecture with UPA to plug it into, and with the recent absorption of Cray’s Superserver team, they’d got a scalable platform, in the form of the E10k.

Someone at Sun had the idea of nobbling the multi-processor bits of the chip, increasing yields, boosting speeds and reducing price. The UltraSPARC IIi was born. It got shoved into the fantastic Netra T1 1u server, and then Sun started looking at the IIi chip for their workstation lines.

Marrying the cheap IIi processor with the UPA crossbar and cheap PC components was a stroke of genius, and the resulting Ultra 5 pizza box and Ultra 10 mini tower workstations sold like hotcakes. They were just what people wanted – cheap, binary compatible UNIX workstations that just worked.

Some Opinion on Workstations

It’s that last point that’s important; UNIX workstations have a very different shelf life from a PC. You need balanced performance – a fast CPU is useless if your graphics and disk subsystem can’t keep up. The machine has to last 4-5 years in a typical environment – I worked with an international communications company who are still using Sun Ultra 10 workstations – over a decade after they were first introduced.

This is important, because if you take the PC economics too far, you end up with a machine that costs more than a PC, but still has to be replaced as often. Silicon Graphics got it wrong with their Visual Workstations, and Sun got it right with the Ultra 5 and 10. (You could argue the choice of OS there was more important than the components)

The Sun Ultra 24

It’s with that in mind that you should read on. My main desktop is a Silicon Graphics Fuel – packed with RAM, fast disks, and a dual head graphics card, it’s more than capable of running the 30+ apps I keep going.

As I posted on Siliconbunny a while ago (see Breathing new life into those old Silicon Graphics machines) the rise of web-centric apps means that older UNIX workstations will have a much longer working life than they’ve had before. IRIX and MIPS (and indeed, Silicon Graphics themselves) may no longer be with us, but the Fuel is just as usable as when it was first brought out.

So I won’t be repeating other people’s work on endlessly measuring the machine’s performance with meaningless benchmarks, and quoted random numbers as if they somehow provide a good measure of how good a bit of kit this is.

I’ve got the Ultra 24 for OpenSolaris development work, and given that it has an Intel x86 processor, it also makes it an ideal test bed for messing around with virtualisation. The bar is set pretty high – it needs to ‘just work’ and be as fast and slick as the Fuel, yet still provide the Solaris hacking environment my trusty Ultra 10 did.

First Impressions

On first unboxing the machine, my first impression was that this is a classic Sun workstation. Dull. Beige.

Sun Ultra 24 vs Sun Ultra 10

Seriously, guys – come on. Silicon Graphics gave you a beating for years with their great lineup of super sexy workstations and servers. If you’re justifying a more expensive machine, then make it at least look like it’s worth the extra money – because otherwise you’ll be losing out to cheap crud from Dell or HP. If you’re offering is essentially a PC, this is more important than ever.

The beige tower with the black DVD-ROM drive is instantly identifiable as the replacement for the Ultra 10. A tacky “Intel Inside” sticker on the front isn’t really impressing either. Apple don’t have to plaster their kit with Intel adverts – why are Sun doing it?

Rear views of the two Ultras

Taking it apart was pretty straightforward – a couple of thumb screws and the side slides off. Inside the case is a comprehensive map showing what goes where, and how to remove all the major components. There is a good range of PCI-E slots, which is going to prove very useful over the life of the machine – 10Gb ethernet and extra graphics cards will be added over time.

The major plus here is the SATA drive slots – they take SPUD brackets! I whipped a drive out of a Netra D130 storage tray, removed the SPUD, fitted it to a SATA drive, and it slid right in. Every Sun shop I know has piles of old 4gb and 9gb SCSI drives with SPUDs – now you’ve got a use for them.

The major minus here is the memory slots. There are only 4 (DIMMs installed in matched pairs). With current 2GB DIMMs that gives you a max of 8GB of RAM – which is really weak. The Ultra 10 could handle up to 1GB, at a time when workstations were usually shipping with a max of 256mb. Given laptops take 8GB of RAM now, and that 8GB is the max in my Fuel, I’m really unimpressed with this. I’m hoping that there isn’t a chipset limitation that would stop the machine using 4GB DIMMs later on, because being limited to only 8GB (on a machine that’s ripe for virtualisation work) is – frankly – a crippling limitation.

With the cover off I whipped out the 1GB RAM the Ultra 24 was shipped with, replacing it with 4GB in 2 2GB DIMMs. I also added a couple of GFX1600 cards, which can convert the two DVI outputs from the Nvidia Quadro card to OpenLDI, for use on my SGI 1600SW screens.

Silicon Graphics monitors on a Sun workstation?

Despite their age, the 1600SW remains the greatest display known to man. If you’re using a machine for long periods of time, nothing else will even come close. Any box that can’t be plugged into a 1600SW is dead to me.

Booting up and the OS

Time out.

Dear Sun,
You have OBP, which is – and remains – awesome. I understand you can’t use this with x86. Silicon Graphics solved this with ARCS, and a nice intuitive and powerful graphical UI for booting and configuring their Visual Workstations.
Apple have bypassed the whole legacy BIOS mess with EFI, and it is truly sweet.
So why I am seeing the same crappy AMI BIOS that’s on my £200 laptop when I turn on the Ultra 24?
Fix this, and fix it now.

Really, this is pretty crappy. It drives home the fact that, for all Sun’s engineering, you’re driving a standard (and expensive) PC. That is not the way to re-inforce the brand or display added value. Please, fix it.

You’ll just have time to gag at the rubbish low-res Sun logo (come back, OBP boot banner, I miss you!) before the machine starts a handful of irritating Intel initialisation messages. I don’t care about Intel’s management interface, and I really don’t care about PXE boot – just hide this stuff and make it go away.

Then you’re dropped into the shiny Grub, offering you a boot route into the pre-installed Solaris 10 OS. Pre-installed Solaris sucks, always – fix that default partitioning guys! Besides, I’m here to slap on OpenSolaris – so in goes the LiveCD and away the machine boots.

OpenSolaris starts up surprising quickly. Most impressive, though, is that it recognises the freakish 1600×1024 resolution of the 1600SW straight away, booting cleanly into the GUI. Colour me impressed – this is the sort of slick “it just works” experience I expect.

OpenSolaris has come along in leaps and bounds, and the installation is both fast and painless. Another reboot, and I can login to the desktop. A few quick clicks and I’ve got an extended desktop across both 1600SW screens.

Yes, I know that this is more a feature of OpenSolaris than the Ultra 24 – but the OpenSolaris community need to be congratulated on this level of ease of use. It’s almost as slick as the IRIX GUI, and that’s really saying something.

Loading up loads of apps and starting to push the machine, and it’s clear that there’s a lot of X server optimisation still to be done. Hesitation, window tearing, artifacts – under load, OpenSolaris just doesn’t seem to perform as well as IRIX on the desktop. Is this Nvidia drivers, or X server optimisation, or both? It’s an issue, but it’s not the Ultra 24’s problem, and I can’t fault it here.

Verdict

The Sun Ultra 24 is a cracking workstation. It’s fast, it’s quiet, and with OpenSolaris it just works. It’s got good room for expansion, although the limited RAM slots could prove to be an issue – especially if you’re playing with virtualisation.

As an entry level UNIX workstation, however, it excels, and is a worthy successor to the line of machines that the Ultra 10 first kicked off.

Latest on the IBM/Sun deal Comments Off on Latest on the IBM/Sun deal

An updated story in the Wall Street Journal sheds some more light on the progress to the deal.

Apparently Sun’s board have rejected IBM’s offer – which I think is a sensible move. IBM asset-stripping Sun seems to be the most likely outcome of any deal.

More interesting is the news that Sun’s board is allegedly split, with Schwartz supporting a buy out, and McNealy opposing it. With the news that NcNealy is allegedly heading up a group that wants to take control of Sun and take it private, this is an interesting development.

Is Scott McNealy the right man to take the helm at Sun? Personally, although I think Sun’s management team are good, they lack experience with anything other than a growing company – and that shows with continued losses.

Arguably McNealy has a similar background – but he’s brought Sun through some tough times before, and potentially has the experience to steer Sun through the current problems. As I’ve mentioned before, he also has the mindset and investor confidence to support that.

The question is – what has McNealy been doing since he stepped down as CEO? He’s still Chairman of the Board, and he’s watched Sun continue to make big losses. Is he taking a long game plan to regain control, or is he as culpable as Sun’s current management team?

What future for Schwartz though? He’s been CEO for a few years now, and hasn’t managed to boost Sun’s fortunes. On top of that he’s overseen open-sourcing Sun’s crown jewels, and not following through with selling the support services around that. That’s a pretty major failing in execution.

With Schwartz pushing a deal that would see Sun swallowed whole by their main competitor, and with nothing else to offer investors and Sun’s customer base, I can’t see how he can continue as CEO. Likeable though he is, his position is now untenable, and with him backing a deal with IBM he’s showing Sun’s customers that he’s out of ideas and sees no future for the company.

I don’t believe that is the case at all, and I think he’s sending a very negative message. With IBM backing away I’m not sure he will retain his position for much longer.

Will Scott McNealy and Carl Icahn take Sun private? 1 comment

I’ve speculated before about Sun taking itself private – it certainly has the capability to do so, even withstanding the dents to it’s balance sheet due to the questionable purchases of StorageTek and MySQL.

News comes from iNewswire that McNealy has resigned from his position at Sun to head up an investment team which will be launching a counter-bid for Sun.

With Paul Ottelini of Intel’s comments that Sun had been hawking itself around Silicon Valley, it’s clear that Jonathan Schwartz either:

  • sees no future for Sun as a stand-alone company
  • is trying to sell the company while he can recoup some value from it
  • has been pushed down this path by Sun investors who don’t understand IT or R&D (yes Southwestern Asset Management, I’m looking at you)

The last point seems the most likely – with the current economic conditions, many investment funds are desperate to recoup whatever cash they can in the short term, as investors panic and want to pull their money out or re-balance their portfolios.

If Scott and his team can follow through, this can only mean good news for Sun and it’s customer base. Like Steve Jobs at Apple, Scott has the narrow focus to push through his vision of how things should be done. Whether he can pull off the whole Jobsian Messiah thing remains to be seen, but I’d place far more credence in that happening than IBM doing anything other than asset stripping the company.

Have a read through the iNewswire release yourself and see what you think.

Top of page / Subscribe to new Entries (RSS)