By pelegri on Jul 02, 2009
This note provides a summary of GlassFish adoption statistics updated through May 2009. The numbers for May 2009 are down a bit from the record-breaking March 2009 but the overall pattern continues to be of growth. The drop seems to be mostly due to seasonal variations (i.e. Spring break) although possibly the Oracle announcement may have also caused some disruption.
Data for GlassFish Server downloads from Jul'05 to May'09 is presented through two data sets, shown aggregated vertically: pure run-time bundles (Sun Java System Application Server, Sun GlassFish Enterprise Server, GlassFish Server, the Java EE SDK and the Java Application Platform SDK, etc) and tools bundles that include GlassFish (NetBeans, the Eclipse Bundle, etc). As a reference point, I'm also including the d/l stats for all versions of JBoss AppServer as reported by SourceForge.
Downloads during May'09 totaled 172,190 for run-time bundles and 508,171 for tools bundles for a grand total of 680,361. Downloads of all versions of JBoss AS totaled 71,986.
The best month for run-time bundles was Mar'09 with 250,756; that for tools bundles was Apr'08 with 695,550, and the one for the combined number was Apr '08 with 882,489. The best month for JBoss 5.0 in this period was Dec '08 with 126,989 downloads.
We started aggregating the GlassFish Admin Console pings in January 2007 to create our GeoMap. The data is mostly useful to track size and geographic trends due to several limitations - we will have improved data after GFv3 releases.
Like with the download data, May'09 was a good month but not as good as Mar'09. Monthly Hits were 529,449 while monthly IPs were 48,666. Cumulative Hits were 8,659,387 and cumulative IPs reached 687,881.
Downloads from Sun have optional registration. I don't currently have monthly trend but the cumulative data as of this writing is:
• GlassFish v2 - 434,687
• GlassFish v3 - 129,707
Update Center Data
Update Center Pings - N/A at the moment; will update when I get the data.
Caveats and Comments
Although the trends shown here and elsewhere have methodological limitations, I believe they are accurate indicators as they all point in the same direction.
We mostly count completed downloads, but, starting in 2009, some of the tools numbers
are attempted d/ls.
• We count downloads from Sun.Com, Java.Net and NetBeans.ORG, but not from our Maven Repositories.
• GlassFish is freely redistributable and we don't track other distributions like Ubuntu, Solaris or OpenSolaris
• IP addresses overcount due to dynamic IP allocation.
• IP addresses undercount due to firewalls and offline.
• We can only count activations through the the Admin consoles, in particular IDEs don't count, and more than 50% of the downloads come from tools bundles (see top section).
Overall, I believe these are solid numbers, but if we discover a problem I will correct it and let you know.