The Private Cloud

  • Comments posted to this topic are about the item The Private Cloud

  • Steve,

    I couldn't agree more, but have a couple of questions...

    1. When were you working with this DB2 system -- I'm not particularly versed in that flavor, but have worked with a couple of other [acquired] IBM systems in my time. Just curious -- I'm not that familiar with virtualization prior to about the last 5 years.

    2. Regardless of the "Hot-Add" MS, already, in my opinion has this basic feature built into Win2008. We use their servers _almost_ exclusively, and we love them -- have tried others, but stuck with MS virtualization. You can create and shutdown/pause servers easily. For some more $$ you can move them off to another machine, but you can do the basics, as I understand what you're describing, with nothing more than a kick-a$$ machine.

    We have two of those and migrate virtual servers between them over a virtual SAN with [pardon language] virtually no issues. Worst issues we've had are drive failures. Hardware on servers and OS have been so solid we've moved more servers to this model than we had originally planned. LOVE IT!

    D.

  • Private clouds require considerable upfront investment in both hardware and software, so they are not a good choice for small and medium sized companies operating currently under tight budgets in this economy IMHO. Also, that technology is changing so quickly that if you don't have the skilled IT resources onhand, as well as the financial resources to keep up with it you will find that your "cloud", that you have sunk a lot of money into, will become outdated very quickly as well :-D.

    "Technology is a weird thing. It brings you great gifts with one hand, and it stabs you in the back with the other. ...:-D"

  • Interesting certainly. I've been looking into the "cloud" recently and I'm interested that there appear to be some very different architectural models emerging, but nobody seems to have mentioned it yet.

    I've started to come into contact with NoSQL and the concept of "Eventual Consistency", the basic premise being that proper ACID databases just have too much overhead to cope with massively distributed applications like Amazon, Google and Facebook. Looking into it a lot of "Cloud" offerings appear to be using this kind of technology, which to my mind is useless to anyone trying to run any kind of live transactional system, but they keep this information buried.

    The fundamental problem seems to be that when you're serving objects to thousands or millions of people you just can't afford the overhead of keeping track of each individual object and it's state, so the user will pick up whatever the system chooses to throw at them and the updates will be done in the background (and certainly in the case of Facebook the actual consistent view of the data may be thrown up many days later).

    And that's about all I know on the subject, but I think as SQL professionals we should understand the differences between these massively scalable but not particularly reliable models which most of the major "cloud" app vendors seem to be throwing at us and "cloud" as described by Steve, if for no other reason than we can fight the bosses arguments when they try and outsource the whole IT department to the web.

    Anyone know enough to write a good article about this stuff? I for one would find it very useful...

  • Interesting certainly. I've been looking into the "cloud" recently and I'm interested that there appear to be some very different architectural models emerging, but nobody seems to have mentioned it yet.

    I've started to come into contact with NoSQL

    ____________________________________________________________________________

    Richard,

    I agree we need more info in that space, but I think Steve's point is more directed toward the "private cloud" -- the one you have [near total] control over.

    One of the IBM DB's I've got significant familiarity with is a "NoSQL" DB. It was interesting, but as a SQL guy, and a developer that prides himself in being a fairly decent [MS SQL / Oracle] / relational] DBA "on the side", I found even that DB (which is a commercial/high $$ per concurrent connection DB) to be far less reliable than any relational DB I've worked with.

    This was/is, however, in the "private cloud" -- in the "public cloud" space, Google is certainly leaps and bounds ahead of most of us, and obviously Amazon is doing something right. I'm sure Microsoft is as well, and I hope to explore that soon, but my current projects limit me to something more in the "private" realm.

    D.

  • I've looked at cloud computing from both sides now, and still somehow, its the cloud's illusions I recall - no one really knows how to sell the "benefits" of cloud computing at all...

    (My apologies to Judy Collins for hijacking her lyrics)...

    Yes, techies are frothing at the mouth over cloud computing, but having attended two non-IT business seminars on it, few if any outside the techie-sphere are buying into it.

    I have yet to see anyone who can remotely sell the "benefits" against the potential huge cost, major overhaul, and maintenance of this old, rehashed, renamed idea.

    As one "C" level professional stated at one of these conferences (to a Microsoft marketing pro)... "...and what will you guys come up with next? Take a 5 & 1/4 floppy disk, give it a new name like "magic data saucer", and try to sell us that???"

    Whether actually good, bad or just plain ugly, no one seems to be able to sell the cloud computing concept in the real world because it may very well be nothing more than a white elephant running around in repackaged skin.

    There's no such thing as dumb questions, only poorly thought-out answers...
  • Hrrmph. That'd be Joni Mitchell, though Judy Collins release the recording first.

    Cloud is old news. The Russian mafia and zombie botnet owners have been using it for years. And SETI.

  • “Cloud Computing”?

    I guess someone came up with a new BS name for external hosting and/or timesharing.

    Wake me when the next “next big thing” comes along.

  • Hmmm... I seem to have witnessed quite a few transitions of marketing b/s from the mainframe to the thick client and outsourcing/insourcing and back again in my time. It seems to come and go in tandem with times when the market wants to thin out middle management.

    And nobody has ever satisfactorily answered the question "how do I guarantee 100% uptime on my internet connection?"

    Having said that our server virtualisation tech is great, I love it, but I still have a server sitting on a rack, and clients accessing it, not sure it's a cloud, although some of the clients are abroad....

    Maybe I could paint the ceiling of my server room like the Sistine Chapel? That's pretty cloudy...

  • ben.mcintyre (7/13/2010)


    Hrrmph. That'd be Joni Mitchell, though Judy Collins release the recording first.

    Cloud is old news. The Russian mafia and zombie botnet owners have been using it for years. And SETI.

    Oh please, everyone knows that's a Leonard Nimoy song!

    --------------------------------------
    When you encounter a problem, if the solution isn't readily evident go back to the start and check your assumptions.
    --------------------------------------
    It’s unpleasantly like being drunk.
    What’s so unpleasant about being drunk?
    You ask a glass of water. -- Douglas Adams

  • Richard Gardner-291039 (7/13/2010)


    Hmmm... I seem to have witnessed quite a few transitions of marketing b/s from the mainframe to the thick client and outsourcing/insourcing and back again in my time. It seems to come and go in tandem with times when the market wants to thin out middle management.

    And nobody has ever satisfactorily answered the question "how do I guarantee 100% uptime on my internet connection?"

    Having said that our server virtualisation tech is great, I love it, but I still have a server sitting on a rack, and clients accessing it, not sure it's a cloud, although some of the clients are abroad....

    Maybe I could paint the ceiling of my server room like the Sistine Chapel? That's pretty cloudy...

    This is exactly why I think the idea of a "Private Cloud" will happen sooner and work better than the "Public Cloud". If your cloud is over your LAN/WAN or is just how you're talking about getting your data and apps to an off-site, secure location, you have a chance of it working. Letting someone else deal with it over the internet? Security issues, reliability issues, access issues all make it far less than ideal. I think the public concept is just a way to charge for something that is currently "free".

    Personally I'm looking forward to when they tell us green-screen monitors and screen scrapes are the wave of the future.

    --------------------------------------
    When you encounter a problem, if the solution isn't readily evident go back to the start and check your assumptions.
    --------------------------------------
    It’s unpleasantly like being drunk.
    What’s so unpleasant about being drunk?
    You ask a glass of water. -- Douglas Adams

  • dhamilton-905368 (7/13/2010)


    Steve,

    I couldn't agree more, but have a couple of questions...

    1. When were you working with this DB2 system -- I'm not particularly versed in that flavor, but have worked with a couple of other [acquired] IBM systems in my time. Just curious -- I'm not that familiar with virtualization prior to about the last 5 years.

    D.

    I believe it was a P690 that we had (http://www-03.ibm.com/systems/power/hardware/pseries/highend/p690/index.html). It was expensive, close to, if not over, $1,000,000. We had 2-3 of them, and they arrived populated with 32 CPUs and 48GB or more of RAM. This was 2000/2001 timeframe. However we had only "purchased" 24 CPUs and 32GB of RAM, or something like that.

    The hypervisor was essentially built into the firmware as the BIOS. You could allocate CPU/RAM/HBA slots dynamically to virtual machines on this box. Since it was built so low to the hardware, very low overhead. We could also "unlock" or "rent" CPUs/RAM from IBM for short periods of time. We could call, get a code, and unlock for days/hours. I think that if we rented things for less than 90 days a year it made more sense. We did rent some CPUs and RAM for end of year processing to get things done quicker, but not EoQ stuff.

    If the x64 architecture and Windows boxes allowed you to do the same, I think it could be cool. I think the problem has been the host hypervisor hasn't been slim enough in the past and the size of boxes under x86 really allowing for larger scale boxes like 32way boxes at a reasonable price.

  • TravisDBA (7/13/2010)


    Private clouds require considerable upfront investment in both hardware and software, so they are not a good choice for small and medium sized companies operating currently under tight budgets in this economy IMHO. Also, that technology is changing so quickly that if you don't have the skilled IT resources onhand, as well as the financial resources to keep up with it you will find that your "cloud", that you have sunk a lot of money into, will become outdated very quickly as well :-D.

    I think this depends on how it's built. People used to say this about clustering, but I'd argue that's not the case anymore. If it was a relatively simple software install that would allow me to put 3 SQL boxes into a "cloud" and let people just deploy dbs to this cloud, or connect to this cloud, I think it easily makes sense for small companies.

  • Richard Gardner-291039 (7/13/2010)


    The fundamental problem seems to be that when you're serving objects to thousands or millions of people you just can't afford the overhead of keeping track of each individual object and it's state, so the user will pick up whatever the system chooses to throw at them and the updates will be done in the background (and certainly in the case of Facebook the actual consistent view of the data may be thrown up many days later).

    Yep, and stuff gets dropped. There are cases/places where FB/Google/Amazon fail to get consistent and the user gets an error. For them, no big deal. For many web sites, it's probably fine. Heck, if forum posts here might be delayed slightly from another server that a person was reading from it wouldn't be an issue most of the time.

    However in some cases it is a big deal. In many companies people would freak out. Perhaps they would learn to live with it, but if one of those servers that was pushing consistency out failed, it can be a big, big deal in some applications. Not sure if it's most, but the paradigm for many corporate applications will be slow to change.

    NoSQL makes good sense for places that don't need ACID and need lots of scalability. There are plenty of situations where ACID is not required. More places where scalability isn't needed. However for the majority of small servers, small apps, I think SQL makes more sense than NOSQL.

  • blandry (7/13/2010)


    ...

    Yes, techies are frothing at the mouth over cloud computing, but having attended two non-IT business seminars on it, few if any outside the techie-sphere are buying into it.

    I have yet to see anyone who can remotely sell the "benefits" against the potential huge cost, major overhaul, and maintenance of this old, rehashed, renamed idea.

    ...

    Not sure I see techies frothing at the mouth. Some like it, most seem scared (to me). Heck, I have run into a number of developers this year that didn't understand virtualization. I think it's easy to see the press and assume that lots of people buy into it.

    I think the costs are like early cell phone plans. They're nebulous. Once you start to pay per use, it can go either way. Lots of friends starting sites, companies, love buying a slice of a server and database online. Simplifies things, cheaper, and they have a little scalability. None of them have hit it Farmville-big, so it's not a big deal. Most people aren't likely to do so.

    However for corporations, would you want to pay/GB or transaction? Do you know what that would cost you? In my mind, it's unplannable. Like early cell phone plans. We don't know how many minutes we're using, and it's hard, in real time, to sometimes determine if it's worth making/taking another call. If the plans are large enough, enough pad included, then it can save you money.

    Or perhaps it's cheaper on unlimited use plans (or high use plans), because the majority of people don't use that much capacity.

    The thing that scares me more than the reliability, security, is the patching. I have had patching issues, and if I have a cloud vendor forcing updates on me that break an application, I am stuck. A private cloud gives me more control. I would like the idea of being able to drop hardware into a database pool and just send data to it.

Viewing 15 posts - 1 through 15 (of 24 total)

You must be logged in to reply to this topic. Login to reply