Discussions

News: Article: Using Terracotta DSO

  1. Article: Using Terracotta DSO (49 messages)

    Terracotta DSO is an open source technology created by Terracotta, meant to provide clustering to Java at the virtual machine level. It does so by weaving code around specific classes, which will communicate with a specific server process to retrieve and update data as needed. This article walks through getting one's feet wet with DSO.
    DSO is deployed as a set of modules that run atop a regular Sun JVM1. A set of scripts to manage the cluster and start the JVM are provided as part of the distribution. The first step DSO users should execute after downloading DSO is running the DISTRIBUTION/dso/bin/make-boot-jar.bat, where $DISTRIBUTION is the home of the Terracotta installation. This will create a set of modules for use with the current JVM. (Any update to your JVM will require re-running this file.) The next step is to have something worth clustering, which will also show off one of the biggest advantages of DSO: codeless clustering.

    Threaded Messages (49)

  2. The 'OSS' claim[ Go to top ]

    For anyone actually trying to use TC, I'd recommend you make yourself VERY familiar with their licensing, there's a fun little clause that says: 'We added an attribution requirement, which says that if a licensee redistributes Covered Code, they must include the phrase "Powered by Terracotta" in legible form at the bottom of any user interface screen' So basically, you can use it without paying if you're willing to carry TC advertising. How is that in the spirit of OSS?
  3. Re: The 'OSS' claim[ Go to top ]

    It is incorrect to say “you can only use it if you are willing to carry TC advertising” Terracotta only asks for attribution if you are redistributing. People who just use the product are unaffected by this. That is the majority.
    For anyone actually trying to use TC, I'd recommend you make yourself VERY familiar with their licensing, there's a fun little clause that says:

    'We added an attribution requirement, which says that if a licensee redistributes Covered Code, they must include the phrase "Powered by Terracotta" in legible form at the bottom of any user interface screen'

    So basically, you can use it without paying if you're willing to carry TC advertising. How is that in the spirit of OSS?
  4. Re: The 'OSS' claim[ Go to top ]

    It is incorrect to say “you can only use it if you are willing to carry TC advertising” Terracotta only asks for attribution if you are redistributing. People who just use the product are unaffected by this. That is the majority.
    Seems to me if this is correct then there is very little to complain about. It is free to use and you don't have to carry advertising when using it. The source is open and can be changed to suit your needs. Is this not open source or am I missing something ?
  5. Re: The 'OSS' claim[ Go to top ]

    Yep...Sort of cheapens what the rest of us are doing - doesn't it? I wouldn't exactly call the faq entry pointing to Mule / Zimbra as doing the same things a ringing endorsement either. Maybe other people here those names and get warm fuzzies. "does Marsellus Wallace look like a bitch?"
    For anyone actually trying to use TC, I'd recommend you make yourself VERY familiar with their licensing, there's a fun little clause that says:

    'We added an attribution requirement, which says that if a licensee redistributes Covered Code, they must include the phrase "Powered by Terracotta" in legible form at the bottom of any user interface screen'

    So basically, you can use it without paying if you're willing to carry TC advertising. How is that in the spirit of OSS?
  6. Hmmm it seems like this thread is carrying on from the Mule discussion. If the Adware/Badgeware people continue to falsely ride on the Open Source bandwagon, this can't be said enough. Firstly Terrecota, MuleSource, Alfresco and others all have great products, no one is denying that, and if they write the code they absolutely have the right to license it as they wish. However they do not have the right to remove the assurances assumed by being Open Source and still claim to be Open Source. Attribution, in a passive form, is nothing new - even Apache have some form of attribution for source code - but quite clearly there is line to be drawn, and this is that line. A big problem is that Open Source implies certain assurances that puts the end-user, not the vendor, in control. It is assurances like this that have accelerated the adoption of Open Source by end users. This success is also why commercial companies want to hop on the Open Source bandwagon, it's great marketting, they get to imply their product has all those same assurances - it doesn't. Quote from Ross Mason
    The Mule team are dedicated to keep up the quality and innovation that Mule is known for, so a little protection was needed.
    You are at the mercy of them doing the right thing, might work for now, might be a PITA in the years down the road, or if they go bust or are bought out. Maybe corporation X buys them in 2 years and closes the source the software, the lock-in due to attribution would force you into a close source world. Quote from Ross Mason
    I think you protest is misguided in this context. Mule is open source software... If they want to fork the code they can.
    If the real Open Source vendor/community refuse to accept patches or implement features or go a particular direction for what ever reason the end-user is free to fork and create a new community - this is an extremely powerful mechanism to ensure the end-user has control, not the vendor. It means that vendors have to compete by offering the best services, writting the best software, building the best communities - or they can have the rug taken out from underneath them. Adware/Badware removes, in all practicalness due to the viral nature of Attribution, the ability to fork thus it puts the vendor back in control - this is antithesis to the way Open Source works, and what people expect from the brand. The vendor can't claim attribution protects their IP and that their stuff is forkable in the same sentence - period - the two are complete diametric opposites. Further to this when a company tweaks the license the first time to gain control and drive a bit more revenue, what's to stop them tweaking it a bit more in the future to drive a bit more revenue - after all the protection of their IP, due to adware attribution, ensures end-user lockin. The MPL+Attribution crowd keep emphasising that almost anyone can use their software free - It's not about "free" stuff, Open Source gives far more assurances than something being free (as in price). Also see http://www.buni.org/mediawiki/index.php/GAP_Against Mark Proctor JBoss Rules http://markproctor.blogspot.com/
  7. Re: The 'OSS' claim[ Go to top ]

    What is the point of putting this on every UI screen anyway? How many non-technical end users are going to say, "Wow! Terracotta rocks! I must rush out and get my own copy. Why didn't I know about this marvellous piece of software before?" Non-techies don't care what it's written in or how it works or what plumbing it's using, they really don't. As long as it works reliably and does what they want it to do, they're happy. "Powered by..." badges on the screen will almost certainly be ignored by 99% of the people using the app and will achieve nothing other than using up valuable screen real estate.
  8. Re: The 'OSS' claim[ Go to top ]

    What is the point of putting this on every UI screen anyway? How many non-technical end users are going to say, "Wow! Terracotta rocks! I must rush out and get my own copy. Why didn't I know about this marvellous piece of software before?"

    Non-techies don't care what it's written in or how it works or what plumbing it's using, they really don't. As long as it works reliably and does what they want it to do, they're happy. "Powered by..." badges on the screen will almost certainly be ignored by 99% of the people using the app and will achieve nothing other than using up valuable screen real estate.
    I believe the advantages are more than just free advertising. You get to protect your IP by inhibiting forking and also drive revenue from those looking to steal your code to OEM in a close product and not pay you a penny. You also inhibit Open Source project, who must provide the assurances of Open Source, from building on your IP, which provides a competitive advantage. All very compelling reasons for an Attribution style license.
  9. Re: The 'OSS' claim[ Go to top ]

    I believe the advantages are more than just free advertising. You get to protect your IP by inhibiting forking and also drive revenue from those looking to steal your code to OEM in a close product and not pay you a penny. You also inhibit Open Source project, who must provide the assurances of Open Source, from building on your IP, which provides a competitive advantage. All very compelling reasons for an Attribution style license.
    All of that can be achieved by a licence that requires attribution but does not dictate that it must be visible on every UI screen. In what way does putting the attribution on every UI screen strengthen the legal position with regards to forking and/or stealing the code? Anyone who is willing to steal the code for commercial gain is hardly likely to pay attention to the part of the licence requiring a "Powered by XYZ" message on the screen.
  10. Re: The 'OSS' claim[ Go to top ]

    Do you get the source? Do you get to use it without paying cash money? Does it get your job done? Quit bitching about a [kcp edited for language]logo and use it if you want. If you don't like the logo and the terms, don't use it. It's not your property, so why do you think you have any right to judge what the owners do with it? You're acting like a damn child who got the wrong color toy for Christmas. Message was edited by: kirk@javaperformancetuning.com
  11. Re: The 'OSS' claim[ Go to top ]

    It's not your property, so why do you think you have any right to judge what the owners do with it?
    No, not your property.. But in most countries people are entitled to something called an "opinion". At the end of the day, its up to the IP owners to decide whether to take that opinion into regard or not, and up to the user to decide whether their opinion is a show stopper for use or not.
  12. Re: The 'OSS' claim[ Go to top ]

    Only the opinions of paying customers count.
  13. Re: The 'OSS' claim[ Go to top ]

    Only the opinions of paying customers count.
    One of the things I like about open source projects is that my opinions generally count far more then they do when I give them to vendors of products that we pay a lot of money to license.
  14. Re: The 'OSS' claim[ Go to top ]

    Only the opinions of paying customers count.


    One of the things I like about open source projects is that my opinions generally count far more then they do when I give them to vendors of products that we pay a lot of money to license.
    You already gave them your "it's good enough to buy" opinion when you gave them money. That's the end of the relationship in most cases. If you think that your opinion is really that important to "open source" projects, you must not be aware of the closed nature of a significant portion of "open source" projects. They usually don't care, or are more interested in egotism (Apache "we're meritocratic, and we don't take patches" or "it's MY project, you're wrong", Hibernate "send five test cases or we don't care, and even if you do, we still don't care, at least not until ten more dups are opened, and then we'll maybe think about it when we understand what the actual problem is"), following specs to the bare minimum (Tomcat), dangling carrots (Resin), or any other "open source" business model. "Open source" has been in quotes because everyone has their own idea of what it should be. As far as I'm concerned, if I get the source code without paying and I can make derivative works without having to ask for permission ahead of time (even if the owner has a right to the derivative), it's open. The source code hippies can suck it because peace and love don't pay the bills and I'll respect anyone else's effort to put food on their plate... and maybe get rich one day.
  15. Open Source has a definition.[ Go to top ]

    "Open source" has been in quotes because everyone has their own idea of what it should be. As far as I'm concerned, if I get the source code without paying and I can make derivative works without having to ask for permission ahead of time (even if the owner has a right to the derivative), it's open. The source code hippies can suck it because peace and love don't pay the bills and I'll respect anyone else's effort to put food on their plate... and maybe get rich one day.
    Open Source actually has a definition that most people seemed to agree with just fine ( http://www.opensource.org/docs/definition.php). In addition, OSI has done a pretty reasonable job of vetting new OS licenses (here's a list of all that are considered Open Source (http://www.opensource.org/licenses/)). The problem with letting a few companies unilaterally change the definition of Open Source is that it leads to confusion. My guess, is that this is what these companies are looking to achieve. A developer realizes that his company has a policy allowing MPL, uses Terrecotta, then realizes later on that this appendix is NOT Ok with management. Many companies that create product are NOT going to be O.K. with shipping with someone else's logo all over their UI. Now they are forced to pay a licensing fee. It is lock-in all over again. If you want to prevent people for doing this, use GPL or LGPL. The problem with "distribution" is that term can be taken advantage of to try to force companies into buying licensing later on. For example, it I had consultants build a web site for me, and they gave me a code drop. Did they just "distribute" the code to me? Do I know need to show this logo for every project that I farm out to 3rd parties? If distribution only triggers code transfer, this point is moot. My consultants are going to give me the code I paid them to write. Whether it is intentional or not, Badgeware is hijacking the word Open Source. Why don't any of these companies submit their license to OSI for approval? My guess is that they realize it will not pass. As for the arguments that you need these types of extremes to make money in Open Source, look at Red Hat or JBoss. Fleury just make tons of cash using the plain ole LGPL. I'm not against proprietary software. I'm just against trying to trick consumers by hijacking a well-known and well-defined term.
  16. Re: The 'OSS' claim[ Go to top ]

    You already gave them your "it's good enough to buy" opinion when you gave them money. That's the end of the relationship in most cases.

    If you think that your opinion is really that important to "open source" projects, you must not be aware of the closed nature of a significant portion of "open source" projects. They usually don't care, or are more interested in egotism (Apache "we're meritocratic, and we don't take patches" or "it's MY project, you're wrong", Hibernate "send five test cases or we don't care, and even if you do, we still don't care, at least not until ten more dups are opened, and then we'll maybe think about it when we understand what the actual problem is"), following specs to the bare minimum (Tomcat), dangling carrots (Resin), or any other "open source" business model.

    "Open source" has been in quotes because everyone has their own idea of what it should be. As far as I'm concerned, if I get the source code without paying and I can make derivative works without having to ask for permission ahead of time (even if the owner has a right to the derivative), it's open. The source code hippies can suck it because peace and love don't pay the bills and I'll respect anyone else's effort to put food on their plate... and maybe get rich one day.
    What company do you work for? I would like to make sure to avoid it.
  17. Re: The 'OSS' claim[ Go to top ]

    What company do you work for? I would like to make sure to avoid it.
    *cry* :)
  18. Re: The 'OSS' claim[ Go to top ]

    Only the opinions of paying customers count.
    Therefore, nobody here should bother to read your messages?
  19. Re: The 'OSS' claim[ Go to top ]

    Quit bitching about a [kcp edited]logo and use it if you want. If you don't like the logo and the terms, don't use it. It's not your property, so why do you think you have any right to judge what the owners do with it? You're acting like a damn child who got the wrong color toy for Christmas.
    Wow, those zen classes are really working out for you. At least you didn't take it personally. I agree with you in one respect: my opinion has alway been that people should vote with their feet. I have no intention of using Terracotta's undoubtedly fine product, not because there's anything wrong with the licence but simply because at this time I don't need it. I'm just interested to understand the thought process that lies behind this licence. I can't understand what their thinking was when they decided to include this particular condition. Message was edited by: kirk@javaperformancetuning.com
  20. Re: The 'OSS' claim[ Go to top ]

    Quit bitching about a [kcp edited for language]logo and use it if you want. If you don't like the logo and the terms, don't use it. It's not your property, so why do you think you have any right to judge what the owners do with it?
    The first post was not about the tool itself but about the licence. Open Source is a set of rules and also a spirit. Since it has gained a lot of success, it is becoming a marketing argument for some companies. Please understand that this appropriation doesn't please everybody. In particular people who believe in the spirit of Open Source software (and that include great developpers who made great contributions to the software world).
  21. Re: The 'OSS' claim[ Go to top ]

    Open Source is a set of rules and also a spirit. Since it has gained a lot of success, it is becoming a marketing argument for some companies.
    This 'marketing argument' you talk about is usually a ploy they take when they have failed to gain significant revenue from selling a license. When I see a company take a proprietary product and "make it open source", I usually see this as desperation and that the company isn't going to be around much longer. Anyways...I consider the gaining popularity of companies using OSS as a marketing argument is a good thing. It means the open source model is winning.
  22. Re: The 'OSS' claim[ Go to top ]

    Open Source is a set of rules and also a spirit.
    Since it has gained a lot of success, it is becoming a marketing argument for some companies.


    This 'marketing argument' you talk about is usually a ploy they take when they have failed to gain significant revenue from selling a license. When I see a company take a proprietary product and "make it open source", I usually see this as desperation and that the company isn't going to be around much longer.
    That's a pretty subjective statement. I guess Bill just wrote the history chapter on Terracotta. How about open sourcing the product to get more folks to actually use the product and realize it's potential and then deciding to pay for support, vs. forcing to work with a 30 day trial period to make that decision? Or how about utilizing a bunch of free programming resources willing to contribute their time for free, by either providing bug fixes or even adding features? Isn't that what JBoss is doing? Oh, no, I forgot, JBoss is doing this in the spirit of OSS.
  23. Its OSS, just stupid OSS software[ Go to top ]

    For anyone actually trying to use TC, I'd recommend you make yourself VERY familiar with their licensing, there's a fun little clause that says:

    'We added an attribution requirement, which says that if a licensee redistributes Covered Code, they must include the phrase "Powered by Terracotta" in legible form at the bottom of any user interface screen'

    So basically, you can use it without paying if you're willing to carry TC advertising. How is that in the spirit of OSS?
    Personally, I'm sick and tired hearing of this "peace and love" bullshit on how project/product "X" isn't in "the spirit of OSS" or "not true open source". Calling Terracotta fake open source is like a Catholic telling a Protestant that he's going to hell because he doesn't recognize the Pope. Its just plain religious and stupid. Terracotta should be allowed to experiment with their own version of an open source business model. Sure, attribution is annoying and you're probably more likely to take a look at something like JBoss Cache that doesn't have such a requirement. But let Terracotta live and learn, IMO. BTW, I took a look at their
  24. triggered by distribution[ Go to top ]

    BTW, I took a look at their license description and it seems that the attribution is only triggered when you distribute their stuff.
  25. Re: triggered by distribution[ Go to top ]

    Suppose, for example, that I am a consultant; and that I write custom software for clients. If I write such software using DSO, does this trigger the attribution portion of the license? I would never ask a client to slap someone else's logo on their site. Does this mean that I should exclude DSO from the delivery toolbox?
  26. Re: triggered by distribution[ Go to top ]

    Suppose, for example, that I am a consultant; and that I write custom software for clients. If I write such software using DSO, does this trigger the attribution portion of the license?
    I would never ask a client to slap someone else's logo on their site. Does this mean that I should exclude DSO from the delivery toolbox?
    No, it would not. Your client would simply need to use DSO.
  27. Re: Its OSS, just stupid OSS software[ Go to top ]

    Personally, I'm sick and tired hearing of this "peace and love" bullshit on how project/product "X" isn't in "the spirit of OSS" or "not true open source". Calling Terracotta fake open source is like a Catholic telling a Protestant that he's going to hell because he doesn't recognize the Pope. Its just plain religious and stupid. Terracotta should be allowed to experiment with their own version of an open source business model.
    I agree that they should do what they want with their code. I also agree that attribution in itself is not a bad thing at all (don't we all want credit for our hard work?) However, there are valid concerns with respect to forking for such open source + attribution licenses. For example, a true fork is difficult to create since its user interface would always link back to the original project. The attribution clause could even potentially be used to assert trademark rights over the fork (lawyers do all kinds of weird things these days). Since forking is one of those basic guarantees of open source -- one that has already been used to great effect on projects like the X server. I, like others, am wary of any "Open Source" license that would inhibit that in any way. Still, on the other end, I see the profit motives (and cannot deny their necessity) that drives these kinds of licenses instead of something like the GPL. I think what everyone is fighting against, Bill, is Open Source becoming yet another completely meaningless buzzword. Projects and people can use whatever license they want, but they also shouldn't be riding the "Open Source Wave" if their license does not really comply with the idea of open source.
  28. Re: Its OSS, just stupid OSS software[ Go to top ]

    The problem gets stickier if you really do want to use it with attribution. Some of us have to live within the walls of governance bodies and a bureaucratic machine that doesn't move from these points. I would actually like for Mule and others to take the attribution clause out if one is just using the software as it is, otherwise, many (including myself) will have to move off of it just to avoid the pain of having to fight in one or more directions. It then becomes a question of how much of a fight one is willing to endure to still use the software, and often times, the threshold of pain vs alternatives is pretty low. I would imagine that if Ross and others modify that license it would be more tractable. A colleague of mine recently remarked about the absurd proposition of having to include the logo on a mobile app with no real estate left for the UI on a cell phone. Sounds silly, but that seems like what you would have to do to follow the letter of the license. This is quite different IMO from not writing air traffic control software in java ;-)
  29. Would having a simple logo pop up while loading the app count for showing the logo? This would at least solve the "real estate" UI problem.
  30. Re: Its OSS, just stupid OSS software[ Go to top ]

    I would actually like for Mule and others to take the attribution clause out if one is just using the software as it is
    I suggest you go read Mule's and Terracotta's license. AFAICT, the attribution is only triggered by distribution. If you are in IT at a government agency developing a portal for instance, AFAICT, you do not have to put the "Powered By" gif because you are not distributing Terracotta/Mule binaries/source outside your company... Although I too hate attribution, it is important to understand license clauses that are triggered only by the distribution of the software. This is very important for instance in understanding other licenses like GPL/LGPL.
  31. Re: Its OSS, just stupid OSS software[ Go to top ]

    just reread it I think you're right ;-). I guess this goes to show how touchy people (myself for example) can be when it comes to open source and licensing.
  32. Re: Its OSS, just stupid OSS software[ Go to top ]

    I would actually like for Mule and others to take the attribution clause out if one is just using the software as it is, otherwise, many (including myself) will have to move off of it just to avoid the pain of having to fight in one or more directions.
    I think that you might be confused by statements others have been making [in perhaps in a self-interested way] about the license. (SEE http://www.javalobby.org/java/forums/t88553.html#92126080 for a technical thread that almost morphed into licensing too.) Do not worry about attribution in any of these companies' licenses unless you actually intend to sell software. And, if you want to make OSS software that you give away or don't profit from you should try approaching the license owner and asking for a waiver of attribution tailored to your scenario. Atlassian, for example gives JIRA and Confluence away to companies that can prove their source and products are freely available. Companies need to protect against (the following are purely made up): "Microsoft announces 'Exchange for Free,' an entry level product for Microsoft users based on Zimbra." "Mulesource technology to embed inside Webmethods...Mulesource simultaneously closes doors." "Tangosol embeds Terracotta to add JVM-clustering to their offerings." As Fleury's Blog [referenced by Kunal above] points out, BSD -licensing would kill the OSS company in favor of the closed-source / commercial one. This is not a discussion about "which is _real_ OSS." It is a discussion about theft (See The NCache discussion here on TSS) becoming legal usage when we make a mistake in our license. We intend to help the community address the need for a license that protects the community and vendor in the best balance possible for all. Drop me a line at [ari AT terracottatech DOT com] if you have ideas / suggestions.
  33. Re: Its OSS, just stupid OSS software[ Go to top ]


    Do not worry about attribution in any of these companies' licenses unless you actually intend to sell software.
    Then why does the license say "redistribution" rather than "selling"? It seems like these are very different acts.
    Companies need to protect against (the following are purely made up):

    "Microsoft announces 'Exchange for Free,' an entry level product for Microsoft users based on Zimbra."

    "Mulesource technology to embed inside Webmethods...Mulesource simultaneously closes doors."

    "Tangosol embeds Terracotta to add JVM-clustering to their offerings."

    Why not just use the GPL? It protects you much better from this, and is actually Open Source (as defined by OSI). Theoretically, if one of these companies wanted to "steal" your software. A possible workaround, would be to have the installer have a "Don't show Attribution" checkbox on it. They would have distributed the software with attribution on each page, and the customer would have removed it. Since, the customer is not redistributing it, everything would be fine. If you think that another company embedding your product in theirs is "Stealing", then you should by all means use GPL, or a GPL dual license. No one would then argue that your license is not Open Source, and it would be safer for you.
  34. Re: Its OSS, just stupid OSS software[ Go to top ]


    Do not worry about attribution in any of these companies' licenses unless you actually intend to sell software.


    Then why does the license say "redistribution" rather than "selling"? It seems like these are very different acts.


    People were being confused by others' statements that using Terracotta requires placement of our logo. I was trying to give a simple rule of thumb to folks who are just developers in-house at a company. Terracotta is not badgeware, period. If you do not distribute, you needn't attribute.
    Why not just use the GPL? It protects you much better from this, and is actually Open Source (as defined by OSI).

    Theoretically, if one of these companies wanted to "steal" your software.
    GPL was an option. I believe I have been quoted in interviews as saying "we started with GPL, but found much of the community afraid of its 'copy-leftness'." Check out the following link. Someone else--not me--discussing the market's fear of GPL: http://www.oreillynet.com/onlamp/blog/2005/07/bsd_best_license_for_it_an_int.html This is all important discussion and the ideas are good. I appreciate everyone's candor and willingness to provide feedback. At this point we should move this discussion to: http://www.terracotta.org/confluence/display/orgsite/Licensing We are working on a simpler version of that page and will, within a day or so hook up a forums topic on licensing from there. I think that will prove a more relevant forum. BTW, I am pretty sure creating a "disable attribution checkbox" would violate the attribution clause.
  35. why not GPL[ Go to top ]

    Will, you suggested GPL. Though GPL is OSI, I have seen several issues with it and several folks especially in the Java field are not happy with it. Here are some of the reasons I can think about - despite I think the world would be quite different if more of Java open source framework were under GPL... - FSF claims GPL is not compatible with Apache 2 license despite ASF claims it is (that is GPL code can use Apache licensed code). People are sometime frighten by that which is sad (the issue arise when the GPL code links to Apache licensed jars f.e.) - Apache licensed / BSD licensed projects will quickly complain they cannot use your GPL bits - unless your GPL bits is a standalone server they can ask their users to dowload separately and then just link to your client jars which has FOSS exceptions (or is licensed in something else than GPL). See MySQL f.e. As for TerraCotta or Mule they could possibly be GPL as beeing mostly standalone things but things might actually get more complex than a MPL+Attribution license as soon as they'd expose API and not just config files (Mule does have API obviously). May be Ari can comment on why TerraCotta did not went GPL. Alex
  36. For a brief moment...[ Go to top ]

    For a brief moment I was so envious of the 49+ comments this poster was getting...then I clicked through to see the commentary. Gasp! I'll be happy to join the benchmark club if a data and compute intensive batch processing benchmark is created for Java (not J2EE, but J2SE). E.g. a bank doing fraud detection on 10 million transactions stored in an ugly multi-record file format off a mainframe... http://www.pervasivedatarush.com NOT OPEN SOURCE ! (but thinking about it) LOL...
  37. Re: why not GPL[ Go to top ]

    Will, you suggested GPL.

    Though GPL is OSI, I have seen several issues with it and several folks especially in the Java field are not happy with it. Here are some of the reasons I can think about - despite I think the world would be quite different if more of Java open source framework were under GPL...

    - FSF claims GPL is not compatible with Apache 2 license despite ASF claims it is (that is GPL code can use Apache licensed code). People are sometime frighten by that which is sad (the issue arise when the GPL code links to Apache licensed jars f.e.)
    - Apache licensed / BSD licensed projects will quickly complain they cannot use your GPL bits - unless your GPL bits is a standalone server they can ask their users to dowload separately and then just link to your client jars which has FOSS exceptions (or is licensed in something else than GPL). See MySQL f.e.

    As for TerraCotta or Mule they could possibly be GPL as beeing mostly standalone things but things might actually get more complex than a MPL+Attribution license as soon as they'd expose API and not just config files (Mule does have API obviously).

    May be Ari can comment on why TerraCotta did not went GPL.

    Alex
    Personally, I'm glad Apache is stupid enough to ban (L)GPL because JBoss would have been outright forked there back in 2003 rather than the "refactoring" that actually happened. Bill
  38. Re: The 'OSS' claim[ Go to top ]

    For anyone actually trying to use TC, I'd recommend you make yourself VERY familiar with their licensing, there's a fun little clause that says:

    'We added an attribution requirement, which says that if a licensee redistributes Covered Code, they must include the phrase "Powered by Terracotta" in legible form at the bottom of any user interface screen'

    So basically, you can use it without paying if you're willing to carry TC advertising. How is that in the spirit of OSS?
    I would encourage the audience of this discussion to read this article - http://blogs.jboss.com/blog/mfleury/2004/08/02/From+GPL+to+BSD+to+LGPL:+On+the+Issue+of+Business+Friendliness.html
  39. Not Open Source[ Go to top ]

    I'm not sure TSS should keep calling badgeware, "open source". As far as I know, none of the badgeware addendums to MPL are OSI compliant. This is important, because there are a lot of assumptions people make about open source, such as the ability to do a reasonable code fork that do not apply with badgeware. The problem with badgeware is this is not the case. If you were to fork the code, it would still be tied to the original product via the logo you would still be forced to display. In addition, when I hear "open source", my initial thought is that If I use it, I may be license constrained, but I would not be design constrained. With badgeware, I am design constrained as well (My UI needs to reflect certain changes, which may have business ramifications). Badgware may be a fine license for some, but calling it open source is rather confusing and could lead to people misunderstanding how they can use the software.
  40. What about performance?[ Go to top ]

    To change the subject of this thread slightly (the validity of the open source claim doesn't really interest me), I think Terracotta DSO is an interesting and possibly innovative technology but the main question mark I have is performance. Some benchmarking against comparable products, e.g. Tangasol Coherence, clustering through the application server etc. would be very helpful in understanding how fit for purpose this technology really is. Anyone out there got any info on this?
  41. Re: What about performance?[ Go to top ]

    To change the subject of this thread slightly (the validity of the open source claim doesn't really interest me), I think Terracotta DSO is an interesting and possibly innovative technology but the main question mark I have is performance. Some benchmarking against comparable products, e.g. Tangasol Coherence, clustering through the application server etc. would be very helpful in understanding how fit for purpose this technology really is. Anyone out there got any info on this?
    Anthony, Here are 2, albeit old, whitepapers comparing TC to Tomcat-internal clustering: http://www.terracotta.org/confluence/download/attachments/151/TerracottaSessions_Tomcat_Comparison.pdf?version=1 and a "popular commercial app server": http://www.terracotta.org/confluence/download/attachments/151/TerracottaSessions_CommercialAppServer_Comparison.pdf?version=1 Problem with benchmarking TC versus Tangosol is that there is no standards-based test of clustering. The architecture of each is quite different so it would be easy to design a test as a one-off where one would easily beat the other. While it is a headache, you could test performance on your own. We have helped customers through their own performance shootouts many times. We have always made a more-than-respectable showing. A couple of the largest websites in the world have found that, because of our fine-grained replication TC can cluster their website session out of the box. I am currently writing an article on Terracotta's scalability that should go live in the next few weeks. Stay tuned for that. --Ari
  42. Re: What about performance?[ Go to top ]

    All of these distributed caching/POJO applications have pretty decent sized followings. It would be really nice if each of them put their money where their mouths are and agreed to develop a performance suite, illustrating some common use cases. One of which may be a heavily overused non-sticky HTTP session, another as a Hibernate L2 cache, generic data grid for a compute cluster, etc. There are plenty of other very common use cases that they all could probably address. So.... GigaSpaces Coherence Terracotta JBoss Cache NCache :P What say you? :)
  43. Re: What about performance?[ Go to top ]

    All of these distributed caching/POJO applications have pretty decent sized followings. It would be really nice if each of them put their money where their mouths are and agreed to develop a performance suite, illustrating some common use cases. One of which may be a heavily overused non-sticky HTTP session, another as a Hibernate L2 cache, generic data grid for a compute cluster, etc. There are plenty of other very common use cases that they all could probably address.

    So....

    GigaSpaces
    Coherence
    Terracotta
    JBoss Cache
    NCache :P

    What say you? :)
    Christopher. I have been asking for exactly this for a while. I think at least 25% of people I speak with ask me for a "good clustering benchmark." I point out that most benchmarks are pseudo business apps that are specifically designed to stress pieces of infrastructure. Your list is exactly that. Examples include a web app with large sessions and small. A Hibernate app with various read / write mixes. And still others exist...like CommonJ (master / worker interface). I would be happy to invest in defining a standard set of apps and associated metrics for testing clustering.
  44. Re: What about performance?[ Go to top ]

    All of these distributed caching/POJO applications have pretty decent sized followings. It would be really nice if each of them put their money where their mouths are and agreed to develop a performance suite, illustrating some common use cases. One of which may be a heavily overused non-sticky HTTP session, another as a Hibernate L2 cache, generic data grid for a compute cluster, etc. There are plenty of other very common use cases that they all could probably address.

    So....

    GigaSpaces
    Coherence
    Terracotta
    JBoss Cache
    NCache :P

    What say you? :)
    This is a pointless exercise, almost any of these could excel against any other depending on the exact requirements. Add this to the fact that you're not comparing like with like, each of these products/companies (you mixed the two) are designed to solve different problems. Most banks (where I work) have benchmarked the first two and each have came back with different results depending on what they were doing. The only two banks I know that tried Terracotta dropped it because it didn't do what it said on the tin and I don't know anyone that actually uses the other two in a production environment. This is about as useful as a benchmark between Java, Perl, Ruby, PHP and C++, perhaps a good laugh to read but a big waste of someone's time. -John-
  45. Re: What about performance?[ Go to top ]

    All of these distributed caching/POJO applications have pretty decent sized followings. It would be really nice if each of them put their money where their mouths are and agreed to develop a performance suite, illustrating some common use cases. One of which may be a heavily overused non-sticky HTTP session, another as a Hibernate L2 cache, generic data grid for a compute cluster, etc. There are plenty of other very common use cases that they all could probably address.

    So....

    GigaSpaces
    Coherence
    Terracotta
    JBoss Cache
    NCache :P

    What say you? :)


    This is a pointless exercise, almost any of these could excel against any other depending on the exact requirements. Add this to the fact that you're not comparing like with like, each of these products/companies (you mixed the two) are designed to solve different problems. Most banks (where I work) have benchmarked the first two and each have came back with different results depending on what they were doing. The only two banks I know that tried Terracotta dropped it because it didn't do what it said on the tin and I don't know anyone that actually uses the other two in a production environment.

    This is about as useful as a benchmark between Java, Perl, Ruby, PHP and C++, perhaps a good laugh to read but a big waste of someone's time.

    -John-
    Couple of points of clarification: 1. I am unaware of any banks that ejected Terracotta. I am aware of several that are in evals and are not yet complete (banks sometimes take upwards of a year on new technologies), but none where we were ejected. However, if it has happened with our open source software, it is possible that we wouldn't have heard of it. No software is perfect for all use cases, and if I were to claim we were, no one here would believe it anyways. Banks are currently using us or testing us for HTTPSession clustering, reference data caching in a cluster, Spring clustering w/o a database, and distributed computing / OSS-based grid. I will make sure to document the use cases in as much detail as customers will allow for TSS users as soon as testing is complete. A few are scheduled to complete in the next couple of months. 2. The original suggestion was for a complete battery of tests. I interpret that as session, Spring, O/R mapper caching, and 2 types of share-nothing grid (embarrasingly parallel workloads and stateful workloads). There is nothing in that request that suggests the vendors get to define a test that only one does well at and that the others subsequently have to live with. Terracotta is indeed interested in helping build a test harness. I am sure the other vendors would agree / not shy away from such a test. --Ari
  46. Always the same story[ Go to top ]

    instead of looking of what is important: Does Terracotta DSO perform well and what can it be used for, what are the experiences with it ... nope ... what do we do ... be a cry baby about licensing. Yes, it is good to know what type of license they have, but is that relevant on TSS? Wasn't there a time in the past where technology still mattered? Nowadays people just complain about licensing... Btw. Terracotta has some nice ideas about clustering pojo's, you should check it out when you're all done wining...
  47. Terracotta - Scale Up Limited Use?[ Go to top ]

    Terracotta looks great! I have several server apps where this would work wonderfully. These all run on less than 20 jvms. This is a great example of adding caching to 'scale up'. The question I have, is what happens when I implement a 250 jvm application I am working on? What if the 250 jvm test is successful and we go to 10000+ jvms? At this point we need to 'scale out'. Am I correct that I would be better off using a single instance of Terracotta for a cluster of say 16 jvms so that I would have dozens of Terracotta instances running vs one overloaded performance blocking Terracotta instance? It seems that caching apps such as this have limited application in large real world environments. If I have distributed caches, what happens when the data changes in one cache? How is that replicated across all caches? PS. We are planning on using grid on demand services for testing. This is a neural net type app with large datasets.
  48. Re: Terracotta - Scale Up Limited Use?[ Go to top ]

    Terracotta looks great!

    I have several server apps where this would work wonderfully. These all run on less than 20 jvms. This is a great example of adding caching to 'scale up'.

    The question I have, is what happens when I implement a 250 jvm application I am working on? What if the 250 jvm test is successful and we go to 10000+ jvms?

    At this point we need to 'scale out'. Am I correct that I would be better off using a single instance of Terracotta for a cluster of say 16 jvms so that I would have dozens of Terracotta instances running vs one overloaded performance blocking Terracotta instance?

    It seems that caching apps such as this have limited application in large real world environments. If I have distributed caches, what happens when the data changes in one cache? How is that replicated across all caches?

    PS. We are planning on using grid on demand services for testing. This is a neural net type app with large datasets.
    Mica, Excellent question. If I understand correctly, you are focused on the article's example. There are other use cases for Terracotta. Jonas just wrote an article on grid computing here: http://jonasboner.com/2007/01/29/how-to-build-a-pojo-based-data-grid-using-open-terracotta. Several folks are using us in this manner. One I can think of off the top of my head is currently testing on 1000 servers. The basic concept in scaling out with Terracotta is to move from a "share everything" or Single System Image model to a "share nothing" model where you use us to coordinate the movement of workload around a cluster. And, as your workload scales, you add JVMs, each of which adds a linear amount of capacity to the mix. If you reach out to me directly [ari AT terracottatech DOT com], I can also walk you through the roadmap of how Terracotta Servers will scale as well. I would talk about it here, but I am concerned about making roadmap commitments to the community on features that are still a few months out. Short answer though is using the ComonJ work / manager pattern, you can scale quite well. Jonas is speaking about the topic at TSSJS Vegas: http://javasymposium.techtarget.com/lasvegas/speakers.html#JBoner And, here is a tutorial on the grid where it is used to create a parallel web-crawling spider: http://www.terracotta.org/confluence/display/orgsite/TutorialTerracottaDsoSpider --Ari
  49. Scaling out[ Go to top ]

    This is a great example of adding caching to 'scale up'.
    The approach scaling by adding VMs is "scaling out". :-) See my blog entry on scaling a data grid. (BTW - Scaling "up" refers to adding CPUs to a big box, for example.)
    The question I have, is what happens when I implement a 250 jvm application I am working on? What if the 250 jvm test is successful and we go to 10000+ jvms?
    The key is to understand the mathematical principles behind the system. Otherwise, predicting scalability (and bottleneck points) is just a guessing game (what I call "PIOOMA", or "pulling it out of .."). Different types of work-loads will [obviously] exhibit different scaling characteristics based on the architecture selected. I can attest to the fact that there is no one "best architecture" for building scale-out distributed systems, and the very things that make one approach perfect for one situation will hobble it for another situation.
    It seems that caching apps such as this have limited application in large real world environments. If I have distributed caches, what happens when the data changes in one cache? How is that replicated across all caches?
    Generally speaking, modern software doesn't have "all caches"; rather, there is "a cache" that is somehow virtualized across many machines. Our Tangosol Coherence product has always provided that "single system image" capability. From reading about the approach that Terracotta uses, it seems to have a similar capability (globally shared object graph, although not necessarily fully replicated).
    PS. We are planning on using grid on demand services for testing. This is a neural net type app with large datasets.
    Perfect :-) Peace, Cameron Purdy Tangosol Coherence: The Java Data Grid
  50. Re: Scaling out[ Go to top ]

    Cameron, Maybe you didn't understand. Adding JVMs/CPUs and using a single cache is 'scaling up'. (has a hump perf) Adding JVM/CPU's with multiple independent caches where each cache cluster functions independently, is 'scaling out'. (has near linear perf) Scaling up has a threshold where adding JVM/CPUs does not add computational function, where as Scaling out, does not or should not have that threshold issue. My point here Cameron, and maybe I didn't ask it clearly, was if this tech was a scale up or out technology. Scale up would perhaps work for a test, but not for implementation, therefore the scale up is out for us. Thank you for pointing out that your company impl is a scale out.