Performance and scalability: Coherence vs. GigaSpaces in the simplest java.util.Map use case

  1. Greetings, I am considering using one of these products for the simplest use case of a very large distributed Map object. I am primarily interested in get(), put() and locking operations. Filtering, querying, notifications must be supported but will be used rarely so their performance is not important. The important performance criteria are * scalability - let's say several thousand nodes * low latency especially on put and lock * cpu, ram and network traffic footprint * redundancy and recoverability. If half of the datacenters go down, how long until the grid heals? Licenses for both products are available so this will not lead to any purchase. Best Regards, Zlatin
  2. Zaltin You many find a good starting point in Gojko review Oracle Coherence vs Gigaspaces XAP. He ended up selecting the later. Nati S. GigaSpaces
  3. Zlatin, We have some interesting benchmark results and scalability reports you might be interested with. Please let me know if you want us to share these with you. These essentially show the major difference between the different solutions. My contact info below. Shay Hassidim GigaSpaces Deputy CTO GigaSpaces US (NY) office shay @ gigaspaces.com
  4. Zlatin, At the risk of upsetting the principals mentioned in the subject line, have you considered GemStone's GemFire? There are some who might argue that since your use-case is very "simple"--with a humble Map interface being the main requirement--that a complex enterprise-level product like GemFire might be overkill. However, when it comes to deployments on "1000's" of nodes, you will eventually find out that there is no such thing as "simple." Careful architecture/design, testing, tuning, and production deployment planning are simply core to any large-scale software deployment, and only products that have have cut their teeth being torture tested in VLS mission-critical applications will stand-up to the challenge. This is ESPECIALLY true for the key criteria you are looking at . . . low-latency, high-scalability, reliability, and fast/smart/deterministic recovery in complex failure scenarios. Today's GemFire is a culmination of nearly 30 years of continuous engineering work highly focused on building the most advanced Distributed System software technology. I encourage you to take a look at GemFire's community site (community.gemstone.com) where you'll find specific information on your requirements, and to reconsider limiting your evaluation to Oracle and Giga. You may be pleasantly surprised. We can also, of course, provide some very impressive benchmarks and testimonials from our VLS customers. Cheers, Gideon Principal Architect GemStone Systems gideon.low AT gemstone.com http://community.gemstone.com
  5. Still looking for scalability?[ Go to top ]

    Hi Zlatin,

    Don't know if you're still looking for scalability for your aps, but we have a newly released distributed transaction manager that helps with scalability at the data level. We integrate with in-memory data grids to persist data to databases.  We currently work with Gigaspaces, and you can try our stuff by going to our website at CloudTran dot com.