Discussions

News: JXInsight 5.7 Released - Metering the Cloud

  1. JXInsight 5.7 Released - Metering the Cloud (12 messages)

    JXInsight 5.7 released today is the first feature rich and scalable activity-based costing (ABC) solution for enterprise applications that dynamically assigns costs to applications, services and components based on mulit-resource consumption, delivering a unified approach to performance management and cost management for cloud computing. Major Enhancements: + New JVMTI Agent and Java 6 Support + Dynamic Group (Cost Center) Tagging + Custom Resource Meters + Custom Resource Counters + Cluster-wide Resource Counter Monitoring + Unit Cost Metering + Cost Breakdown Analysis (Chargeback) + Activity Cost Tracking + New Metering Strategies + Reduced Memory Footprint + Cross Language Metering: JRuby-to-Ruby + JMX Integration + Oracle Coherence (Grid-Wide) JMX Integration + Java Logging Integration + Azul Systems Resource Meters Extension A What's New in JXInsight 5.7 has been published here: http://www.jinspired.com/products/jxinsight/new-in-5.7.html Some background information on our metering the clouds initiative can be read here: http://www.jinspired.com/products/jxinsight/meteringthecloud.html William
  2. + Azul Systems Resource Meters Extension
    What's the Azul support about? Peace, Cameron Purdy Oracle Coherence: Data Grid for Java, .NET and C++
  3. Hi Cameron, For standard JVMTI agents a JNI call to obtain various metrics or event callbacks results in a round-trip from the compute appliance back to the client/proxy JVM. This makes it practically impossible to use any standard native profiling solution as the JNI/network call would perturb the figures significantly. But since our probes technology can be extended to providing metering for any type of resource meter we can easily plug-in Azul specific replacement meters with the native call executed in the compute appliance. We provide 5 Azul thread based meters (they call them counters) currently but we hope that this will be expanded in the future to include other meters related to the internal performance of the VM itself such as IO/Network wait times between proxy-and-compute appliance. It is important that counters be thread scoped, unlike JMX based metrics, otherwise any reporting is ** last ** generation as you are effectively using a legacy & traditional costing approach to performance management inaccurately apportioning cost and causality (unless of course one JVM equates to one activity or one service). JXInsight Probes uses an activity based costing approach which means counters/meters can be metered in terms of the request/workflow/process/activity/task/tag/thread. We provide 30+ built-in base resource meters along with the capability to define custom meters based on metering for a particular a group/cost center (com.coherence.* or com.tangosol.*). We also offer the ability to dynamically create resource counters which can be mapped to meters. These can be built very easily be extending an single abstract aspect class in our SDK and defining a pointcut in an aop.xml. William
  4. It looks quite interesting, I think metering is one of the big technical challenges when it comes to providing "cloud on a tap" based services.. ..but the pricing looks a bit confusing/not thought through (or maybe I'm misunderstanding it) - at USD5700 per 2CPU's it seems you could rack up quite a cost if providing some sort of SAAS or hosting solution, as you're hardly likely to be able to charge much more than USD100 per CPU/month unless you have some very unique, high value niche product that has nothing to do with cost efficiency (which would seem to go against the case for cloud computing in the first place). Maybe I'm misunderstanding the pricing, as I thought?
  5. Hi Wille, I am confused why you would compare a monthly rate with a fixed once off payment and why you would use a bare metal monthly CPU rate when surely the software managed/metered within the cloud is adding much more value on-top of this base cost otherwise I think you have a flawed business model or you are the cloud itself. Some items you might have overlooked. 1. JXInsight includes 6 technologies (probes, traces, transact, diagnostics, insight, metrics) one of which is a resource metering solution that can uniquely scale in terms of usage (dev, test, prod) and performance. With each technology we have out-innovated the competition by leaps. The technologies are listed at the top of the page: http://www.jinspired.com/products/jxinsight/ We have already provided a competitive offer for our probes technology to a major US bank for a deployment across up to 1,000 servers - not even an internal cloud. 2. The server edition which includes the above mentioned 6 technologies also includes a development-only edition which can be distributed to all development teams within the organization free of charge. This is a major cost saver and even more so if the product is used by experienced performance engineers. One user has said previously "JXInsight can do just about anything". This is evident in the fact that all our extension libraries (current count is 630) are built on the same open API offered to customers. IMHO you are not buying a product you are buying "the solution". 3. The unit cost is 1,700 USD for all technologies. The price you have mentioned is a base installation cost for the first 2 CPU's. If you are really in the cloud and not just being hosted then eventually there will be no concept of an actual physical rented server (or base installation). Amazon and others are offering this as it until the market matures and is ready to accept and understand the real next step in computing. Hint: If you can actually see a one or more physical/virtualized server then obviously you are not in the clouds. 4. The cloud computing infrastructure is currently priced at near utility levels but the software and services running on-top of the cloud is certainly not as the scale is not there yet outside of Google and Microsoft. For companies using the cloud the value of the software and service delivered must be sufficiently high (value or volume) to warrant such a step and here service quality and reliability will be paramount. 5. JXInsight Probes technology aims to cut costs across the board by providing real production metering data which can be filtered back through testing and into development. This rarely happens today. In fact I have lost count of the number of times I have walked into a war council and questioned the level of activity per use case/workflow/transaction only to get a blank look back. No one could run a manufacturing facility like this but yet IT does. William
  6. Thanks for the clarification, USD1700 makes a lot more sense, although still a bit pricey.. I guess it depends a lot on the use case - I was thinking from the perspective of "being the cloud" at which point it doesn't make much sense because of the low margins, for clients "using the cloud" USD1700 for additional processors is probably not bad at all.
  7. Hi Wille, If you become the cloud, a very big one at that, then you move from a standard software purchase to a technology license which will reduce the cost substantially. It would like to point out that we have designed the API, runtime and tooling with multiple extension points in mind from the point of cloud computing platform vendor, cloud service, and cloud service user. CC platform vendors can (1) create and updates probe counters (pushed base) which map to meters or (2) integrate out-of-box meters (pull based) into the runtime, (3) map the metering data for a particular group to a meter. These meters and metering reflect the variable cost of the cloud computing platform or cloud service such as the number of bytes written to storage, the number of objects written to storage, the amount of cpu consumer during a request processing. CC service providers running their platform on a CC platform (though this is not required) can meter the firing of probes (named group) that map to the main service entry points into their cloud service (i.e. the number of twitter messages published) in terms of the underlying cloud platform meters. In turn the CC service provider can use the same activity metering data to define a a custom service meter which can then be used metered by probes that are named according to the cost center hierarchy of the user (org, unit, dept, role, user). Metering reflects the usage of a resource meter in the context of an named activity (group, cost center, cost object) which can be hierarchical. A meter can itself be based on the metering for a particular meter by a particular named activity. This recursive nature in the definition and aggregation allows us to cost and charge from multiple perspectives (vendor, provider, user). Do you need to have all CC parties involved to use our probes technology? Certainty not, in fact you can use our costing approach to improve performance of any Java application by using our metering data to better inform developers, testers, operations, managers and users of the usage and activity of the software in the wild. William
  8. http://www.manyniches.com/cloudcomputing/business-model-influencing-software-architecture/
    In a world where SaaS vendors haven’t yet figured out how to do consumptive based pricing for their offerings, the very real possibility exists that a small handful of customers can abuse your application and destroy not only the profitability of the account, but of your entire P&L.
    Consumptive costing models, automatic scaling of applications and difficult to define atomic quota units have the potential to create serious financial challenges for cloud based application vendors. The new challenges will manifest themselves by enforcing rigid software efficiency design goals on development teams, and forcing the operations team to entertain the notion of firing bad customers. Designing good software is certainly not a new topic, but the possibility of bankrupting a company is not something about which the architects have ever really had to think.
  9. I think you have a flawed business model or you are the cloud itself.
    I always look forward to posts about William's product.. He never fails to insult a potential customer..
  10. Threads Gary P has been participating in Forum Thread Post Started By Messages Last Active News JXInsight 5.7 Released - Metering the Cloud William Louth 9 November 11, 2008 News JXInsight 5.6 Released William Louth 14 June 11, 2008
  11. Threads Gary P has been participating in
    Forum Thread Post Started By Messages Last Active
    News JXInsight 5.7 Released - Metering the Cloud William Louth 9 November 11, 2008
    News JXInsight 5.6 Released William Louth 14 June 11, 2008
    Thanks I forgot about that one from June.. What a hoot!
  12. Wille, I think it would be much better to compare our product/technology pricing with the pricing of the cloud service offered and not the ** cost ** of the cloud provision/allocation. Would you price a cloud service you developed and maintained at 10 cents if one 1 CPU/month can handle 1000 users. This would seem to undermine the value of the software stack. Software licensing or software subscriptions will still have to be paid for irrespective of the location of the computation (to some degree). Eventually elements of the software stack will become an integral part of any cloud computing platform itself but there will still be many companies with hybrid architectures and platforms and they will want a solution (API, Runtime, and Tooling) that will offer a consistent metering and performance management solution across platforms and more importantly across application life cycle stages. JXInsight certainly meets this requirement more than any other solution on the market. I strongly believe that our Probes technology (API and runtime) is close to the optimal design solution which we plan on porting to other languages and runtimes in the new year. William
  13. high value niche product that has nothing to do with cost efficiency (which would seem to go against the case for cloud computing in the first place).
    Our resource metering technology will work within and outside of a cloud computing environment. Activity based costing (http://en.wikipedia.org/wiki/Activity-based_costing) can and should be used to manage the design, development, and delivery of IT services internal or external. Cloud computing is a nice way to get across this approach and its concepts (cost objects, cost drivers, resources, activity) as the correlation between performance (resource usage) and cost is much more evident in the cloud computing context (monthly billing) though I fail to see why internal IT management teams cannot operate so. William