OracleWorld began today and features an impressive lineup of enterprise java related sessions as well as live online coverage of the keynotes and other events. Also, Oracle's new JDeveloper 10g preview is available, and the latest dual and multinode JAppServer2002 results position Oracle's appserver 10g ahead of IBM and BEA in price/performance.
- Posted by: Floyd Marinescu
- Posted on: September 08 2003 18:38 EDT
Live coverage of the keynotes at OracleWorld will be available all week, here:
Oracle JDeveloper 10g Preview is also available and features: Visual page flow design for Struts. One click deploy to Oracle Application Server, BEA Weblogic, JBoss, Tomcat. A Visual layout editor for JSP, HTML, and Swing. Drag and Drop data-binding between UI (JSP and Swing) and business services such as EJB, TopLink, Web services and ADF business components. Integrated Top-Link mapping. Automatic design pattern implementation for EJB. Visual Database design and reverse engineering, etc. The preview is available here:
The latest JAppServer2002 results, in which Oracle's new application server 10g won the dual node category and won in price/performance in the multinode, are available here:
Reflective of Oracle's continueing emphasis on the J2EE industry, this years OracleWorld is not a DBA show anymore - it features an impressive lineup of conference sessions about Enterprise Java, summarized here:
Conference website link:
- OracleWorld begins, JDeveloper 10G & new JAppServer2002 Results by Cary Bloom on September 08 2003 19:06 EDT
- OracleWorld begins, JDeveloper 10G & new JAppServer2002 Results by Cameron Purdy on September 08 2003 22:16 EDT
- OracleWorld begins, JDeveloper 10G & new JAppServer2002 Results by Ivan Ooi on September 09 2003 02:18 EDT
- Nice to know by Stephane Valseme on September 09 2003 03:59 EDT
- OracleWorld begins, JDeveloper 10G & new JAppServer2002 Results by Mileta Cekovic on September 09 2003 06:03 EDT
- OracleWorld begins, JDeveloper 10G & new JAppServer2002 Results by Ignatius Reilly on September 09 2003 16:11 EDT
what else is new? :-)
I didn't look at the details yet (e.g. what pricing tricks were used, if any) but I will say that the numbers were fairly symmetrical across the different categories, and they did achieve the best price/performance (as well as the best performance in one of the categories, IIRC).
It's not going to cause a stampede. (I mean, Oracle already has like 99% of the app server market, right? 16000+ customers already in production with it, or something like that according to the Oracle product manager on the other thread. BEA and IBM just pick up the scraps, I guess. JBoss who?)
OTOH, it's a good line in the sand, and (assuming they didn't pull anything sneaky in the submission) they deserve to be congratulated.
Coherence: Easily share live data across a cluster!
Is there any features like auto indent which available in JBuilder ?
That what I hope from JDeveloper~
BTW got JFormatField included ?
nice to see that they used Bea's JVM :)
This is the first time that two EJB benchmarks are published using almost the same hardware. BEA/HP previously published benchmark on WebLogic 8.1 using also:
AppServer tier: 3x2x3Mhz Xeons (although BEA used DP Xeons with 12KB+8KB (instruction + data) L1 cache and Oracle used MP Xeons with 16KB+16KB L1 cache, and BEA has 2GB RAM, but Oracle 4GB RAM)
Database tier: 4x2.8MHz Xeons MP. BEA again used 2GB RAM and Oracle 4GB RAM.
Both systems used the same JVM (JRockit 8.1) but BEA used SQL Server 2000 and Oracle used Oracle 9i DB.
I suppose that Oracle's better result (1165 vs. BEA's 1037) is more due to Oracle database and more L1 cache and RAM and that Oracle and BEA AppServers perform very similarily.
I've missed that Oracle used Red Hat Enterprise Linux AS 2.1 and BEA used Windows 2000 Advanced Server on both, AppServer and DB tiers.
Oracle also used 15K RPM SCSI HDDs and BEA used 10K RPM SCSI HDDs.
Thus, software stacks are:
Oracle: Oracle AppServer/Oracle DB/RedHat Linux
and hardware stacks are 'almost' identical, except:
BEA: 12KB+8KB L1 cache on AppServer tier, 2GB RAM and 10K RPM HDDs on both tiers
Oracle: 16KB+16KB L1 cache on AppServer tier, 4GB RAM and 15K RPM HDDs on both tiers
1) Oracle is not using Generally Available products
The biggest weakness in Oracle's announcement is that is that they are using a "technology preview" or beta release of their Oracle 10G application server and database drivers. BEA's benchmarks all use generally available products. The results have very little relevance if you cannot deploy the products.
2) Oracle is playing pricing games again
The price/performance metrics are not directly comparable to BEAs or IBM's. Oracle is using a 3-year term license compared to BEA's perpetual license, which distorts the relative cost comparison of the two products.
3) Some of the BEA benchmark numbers use older and more expensive hardware and software
There is a very large discrepancy in the third BEA submission listed above. This is due to two issues. First, that benchmark was done on an older Intel Itantiaium2 platform, which increased the cost and did not show currently competitive performance. Second, it was on an older version of WebLogic Server. The entire product stack has improved considerably since then, so make sure to look at the more recent BEA benchmarks. Don't put it past Oracle to compare their new numbers with BEA's old ones.
4) Oracle is using BEA's products to achieve their SPEC and TPC benchmark results.
Oracle is using BEA WebLogic JRockit as their JVM for SPECjAppServer. Furthermore, all of Oracles top ten TPC-C benchmarks are front ended by BEA Tuxedo (http://www.tpc.org/tpcc/results/tpcc_perf_results.asp). When Oracle demonstrates performance, they quietly turn to BEA.
5) BEA still holds the throughput record for SPECjAppServer2002
BEA still has the highest absolute throughput record in the current benchmark.
6) BEA is the only truly hardware and database independent application server
WebLogic Server has achieved world record numbers on the widest variety of hardware, operating systems and databases.
Hardware: 32 and 64 bit Intel, PA-RISC and SPARC
Operating Systems: HP-UX, Solaris, Linux, and Windows
Databases: Oracle and Microsoft SQL Server
Oracle users are somewhat tied to one database. For customers and ISVs that need to deploy to heterogeneous environments, BEA is a proven choice.
Lets stay to the facts here and cut the marketing:
> 1) Oracle is not using Generally Available products
> The biggest weakness in Oracle's announcement is that is that they are using
> a "technology preview" or beta release of their Oracle 10G application
> server and database drivers. BEA's benchmarks all use generally available
> products. The results have very little relevance if you cannot deploy the
That's funny, one of BEA's benchmarks, which is mentioned as the highest throughput was also published BEFORE the product was Generally Available:
"Test date: Mar-2003, EJB Container Avail: May-2003, EJB Container JVM Avail: Mar-2003"
I'd like to watch the keynotes, but they require the RealPlayer, and we all know what a rat's nest that is.