The Bankwatch

Tracking the consumer evolution of financial services

Photo Matt » Sun Isn’t Relevant to Startups

This is just a fantastic post from Matt (WordPress) that is a clarion call to the Sun, Microsoft, Oracle, group of old guys who really need to get with the programme, or they will be gone.

I’ve given up, I’ve lost hope, and our business has moved on. We’re growing fast and adding infrastructure faster, all on a tight budget, and I’m less inclined to depart from the LAMP-based architecture that Digg, Flickr, Wikipedia, and dozens of other sites I respect use.

Personally I’m far more excited about what Amazon is doing these days.

Source: Photo Matt » Sun Isn’t Relevant to Startups

With regard to Matt’s comment on Amazon, I am so impressed with S3, and Simple Storage in particular. This is the wave of the future, bringing real computing on demand to the masses.

Technorati tags: , ,

Written by Colin Henderson

January 20, 2007 at 06:36

Posted in Uncategorized

9 Responses

Subscribe to comments with RSS.

  1. As a self-appointed “Developer of All Things Fantastic and Efficient”, I would have to agree. I actually feel kind of bad for them. It is not that large companies are no longer relevant. I think that both Amazon and IBM and Google have shown this to be too simple of an analysis. Its that Sun is stuck in the same mind set it had 10 years ago.

    I learned my first Unix on a Sun Solaris system only about 5 years ago. But, since then, the University where I learned it has thrown them out for cheapter, lighter, and more maintainable Linux boxes. The reason that Sun has made so much money over the past 10 years is that they have lots of people in lots of offices wayyyy up high who believe that you must go with a company like Sun in order to write “serious” applications.

    Meanwhile, people tinkered in their basements with a passion. A hidden passion from the people at the top. Nerds listened though. They told each other. Word spread.

    You can run a fast, cheap, and solid web-business (that is server-based business) on relatively fast, cheap, and solid commodity boxes. Sun is skating on the remaining goodwill left.

    Personally, I’m rooting for Sun in the long-run. Their legacy is great and I would hate for them to not figure it out.

    Faster versions of what you sold 10 years ago are no longer useful to us. Innovate! Make my job easier! And no, Solaris isn’t making my job any easier.

    -hampton.

    Hampton Catlin

    January 22, 2007 at 23:14

  2. Hey Hampton … thanks for the comment.

    Matt goes on to say he likes the LAMP approach.

    I remember seeing McNeely at conferences, and he was more obsessed with being un-Microsoft, than being better, I always thought.

    Colin

    January 23, 2007 at 20:21

  3. Thanks for the interesting blog and the subsequent comments. Although Sun has over the years misread what was happening in many places of the market and in particular, the open-source world, I think if you look a bit more closely that we’ve also been making a lot of changes. For one, OpenSolaris has been open-sourced (http://opensolaris.org) for almost two years now with five separate distributions available and a growing community forming. Sun also realize that innovation is taking place everywhere and things like LAMP (or should we say, SAMP) are available for OpenSolaris distributions. In addition, Solaris is available for free (one only pays for support.) Finally, Sun definitely doesn’t look at itself as purely a high-end platform. OpenSolaris is available on both SPARC and x86/x64 systems and Sun supports Solaris on more of the latter such platforms than even Red Had does for RHEL.

    Sorry if I sound defensive but given all of the changes in the past couple of years (including the introduction of innovative Opteron systems that run Solaris, Linux and Windows and Intel-based ones in the future) it sounds like you’re talking about a Sun from about five or so years ago! 🙂

    David Comay

    January 23, 2007 at 22:37

  4. Hampton, I don’t necessarily agree with you when you say Sun is still thinking like they did 10 years ago. They have changed rapidly with the market, in ways that are invisible to the average person. Do you have a Blackberry or Motorola RAZR? If you do, Sun is in your pocket or on your belt right now.

    The only thing that Sun is doing that is 10 years old (and older) is maintaining backwards compatibility. A company that needs to replace a backend Solaris 2.7 server that was installed in 1998 can copy their binaries to a Solaris 10 host and expect it to just work. Think back to 1998, do you think any other OS vendor or distribution could boast the same compatibility in their most recent release? While this isn’t necessarily relevant to most web activity today, for the average corporation that has a finance server under Mary’s desk that has just been running since god knows when, this is a serious advantage for Sun.

    As for innovation, in hardware, Sun has completely eclipsed their competition with their Niagara line of processors. If you play the GHz game, they don’t appear very attractive, but when you can deploy a single host and realize can process web service requests at 40K/sec, you understand that their engineering is superior. I’d rather spend a little more money on a few well engineered machines than spend all my money on a fleet of sysadmins to manage the convoy of linux boxes required to do the same workload.

    Innovating in software, Sun continues to improve upon Java and Solaris at a pace that would embarass most other firms. The improvements are too many to list here.

    What I will admit is that we’ve been somewhat disappointed in that Sun’s new innovation model has had them pushing features into the production stream before they are fully complemented with related technologies.

    Example: Solaris Containers without Network Virtualization. We’ve been waiting for Crossbow since Containers was released. Without network virtualization, running zones on the same host but on different networks, while maintaining isolation is non-trivial.

    As for making your job easier, take a look at just three technologies that are in the production Solaris 10 release today: ZFS, Zones/Containers, DTrace.

    ZFS – 128-bit copy-on-write filesystem. Endless storage flexibility, easy administration (point/click or simple CLI), virtually unlimited point-in-time snapshots, cloning, and bulletproof data integrity (even an accidentally power-cycled machine comes up in good state). All in all a good foundation for the services you build/deploy on top of it. Easier than LVM, and easier/cheaper than Veritas.

    Containers/Zones – resource management, isolation, security, portability. You can create and run a zone on host A, and when host A is overloaded, you can move zone in it’s entirety to host B, using a single command line. Combine Zones with ZFS (cloning) and you’ve got complete environment provisioning in a matter of seconds. Now that’s easy.

    DTrace – one word, visibility. Do you want to know what that ISV’s code is doing when it’s consuming 99% of the CPU, and they tell you to just kill it and restart? Want to help your developers write better code by showing them the overhead of their abstractions? Makes me feel easy knowing that I can use DTrace branded X-Ray specs to see what’s going on.

    I think that the complete rework of the OS while maintaining backwards compatibility is a feat for which they do not get enough credit. That makes life easier too.

    All in all, I’d say Sun is innovating the pants off the competition, and making your life easier. They’re giving it all away, free for download, in the hopes that you’ll buy their hardware, support, and services, and once you’re big enough they want to provide you with the development & system administration teams to keep you going.

    Even if you don’t use Sun’s hardware, support, or services, you should use Sun’s software. It is qualified on Dell, HP, & IBM. It runs well on white-label AMD boxes, and after recent announcements, we can only assume that Intel will provide Sun with the chipset drivers so that Solaris will run well on white-label Intel boxes too.

    Gosh, it feels like I’ve written a brochure. I don’t work for Sun, but I’m a happy customer. We’ve taken much of this innovation and used the technologies to drive costs and complexity out of our business. Life is easier. We’re not sitting back and counting our money just yet, there is still work to do, but Sun continues to be there helping us along.

    MK

    Michael

    January 24, 2007 at 03:36

  5. Michael-

    Obviously you have extensive experience with Sun equipment. So, beyond my experience learning Unix on a Sun box, I have had very little experience with them.

    So, I must in general defer to you on Sun’s product line.

    However, I do disagree with the idea that you would need “a fleet of sysadmins to manage the convoy of Linux boxes required to do the same workload”. While I was doing some research for NASA, we built a 16 box Linux Beauwolf cluster to process streaming satellite data from weather satellites. I have to say, the process was relatively simply and straight forward. We were only 2 guys doing it.

    Even Google and AmazonWS are mostly built with commodity hardware. And, I don’t think they are a bunch of people with regrets on their speed.

    You’d be surprised what a generic server can handle. And, I still believe, that unless you are launching cnn.com or something, and by using offsite fileserving (S3), and a share-nothing architecture. You can have a very, very, very fast server setup on the cheap that does not require a team of support technicians. If you *do* need that, then there is something wrong with your hiring practices or your software (see the first one too).

    I may be very flawed in my view of deployments, but so far, I have yet to hear *anyone* say… “My god, my webserver is getting hammered on this Linux cluster! What am I going to do????” At this rate, there should be thousands of web companies screaming that.

    If I am wrong, I’d *love* to be corrected. I’m a pragmatist from the core. Which is why I don’t believe a big company with carte blanche confidence “pay me, and i will make it easy”.

    I need to be proven that free stuff is *harder* and paid is *easier*. In my limited experience, the opposite has been true.

    -hampton.

    Hampton

    January 24, 2007 at 12:25

  6. Hampton,

    Again, we’re not far from each other on this one.

    One divergence. I meant ‘web services’, not ‘web server’. I was referring specifically to J2EE web services out of a BEA Weblogic container, not referring to static content delivery using something like Apache or thttpd.

    And I headed in the wrong direction with the hardware argument, as I do think it is our specific problem set that makes many Linux boxes arduous over fewer Niagara hosts. We use BEA Weblogic as a standard, and it costs us a fortune in licensing and maintenance costs. And for the administrator ratio, that is specifically because BEA seems to make it exceedingly difficult to automate container installation. Though, to be fair, once the containers are there and configured, the applications slide right in. Again, another argument that is off topic.

    The key to what I was saying is that Sun has made available, for *free*, all the software technology innovation that they are developing. You can get it all, and install the LAMP (under Solaris, it would be SAMP) stack, and benefit from the technologies they’ve added to the equation that aren’t yet available in the Linux streams.

    This big company is being forced to provide price/license parity for the most basic component, the operating system, due to force of OSS competition. And in doing so, we all benefit from the advances that they have added and continue to add to try to differentiate themselves from Linux in the marketplace.

    And each of the advances they’ve added are there to make your life easier, so you can spend less time worrying about how to configure storage or limit privileges, and spend more time engineering your software for your business specific requirements.

    But I don’t need to talk to you about software engineering, you know what’s going on. Keep it up, I look forward to more posts on REST.

    MK

    Michael

    January 24, 2007 at 23:33

  7. First off, my apologies to David from Sun, because you were stuck in my spam thing for a few hours.

    Fantastic discussion going on here. I think we veered off a bit on the point of java on Blackberries, and while I love my gmail java thing on my blackberry, the central issue raised by Hampton, and countered by David seems to centre on ease of use and setup.

    Thoughts?

    Colin

    January 25, 2007 at 00:27

  8. I think that which technology your application is developed in makes a massive difference. Honestly, if I had a gun to my head and had to go back to my old friend Java, I’d probably go with a Sun server to deploy. Mostly, because if you don’t go with the vendor on that one, its a literal *nightmare* to deploy.

    Thanks to the technology set I’ve chosen. Setting up a server takes about 2~ hours. And, with properly setup Capistrano scripts (another 2 hours), then I am able to deploy to the servers a new version of an application with one command ‘cap deploy’ to as many servers as I have RAM and patience for. Capistrano makes deploying on Linux (or FreeBSD) fantastically easy for a developer. So literally, after that initial setup cost, there is only *one* command to make it all happen. About 100 steps have to occur in the deployment, but it is all automated so its no sweat off my back.

    So, for a total of 4~ hours of work (Ubuntu + Rails + Capistrano) I am forever ready for near instant deployment. And I’m not even a deployment expert!

    And, general ease of use is hard to compare. Unless we know someone who has extensively deployed production servers in both technologies, then its hard to compare. Maybe Java has a quick setup, but slow deployments. Maybe, Sun is better with this or that, but Linux is better at that or this.

    It’s just one of those things.

    But, I can say that in my mind, BEA Websphere deployment would be much easier on Sun. But, don’t get me started on development effort and time.

    -hampton.

    PS: Collin, you are like a comment pusher. Its fantastic!

    Hampton

    January 25, 2007 at 08:02

  9. Hampton…. just trying to encourage good conversation, but thanks!

    Colin

    January 25, 2007 at 23:17


Comments are closed.

%d bloggers like this: