GridGain confirms that Apache Ignite has twice the performance of Hazelcast
A provocative blog written by Mr. Greg Luck, CEO of Hazelcast, accused the Apache Ignite community of "Forging" test results. This blog has caused some confusion, I think it is necessary for me to clarify.
Honestly, we are very surprised to see this blog from Hazelcast. Should Mr. Luck be at least active with the Ignite community or GridGain team before such gross criticism? The general premise is that some people conduct an open test for an open-source project and "deliberately" fake it all over the world. Does this seem silly?
I should also mention that through the Hazelcast user group email list, the Ignite community actively contacted them to publish test results and code, and then hoped that the Hazelcast team could review and provide feedback. However, Mr. Luck chose to delete the email from the Hazelcast public mail list and made an annotation:
We think it is inappropriate to submit your beautiful numbers in our email list. We intend to delete your emails. Please put them on your own email list and website.
Delete an email from a public email list simply because you do not like its content? This crude behavior clearly shows the substantial difference between a business company and an Apache project.
Needless to say, after reading the Hazelcast blog, we re-tested all the tests and re-confirmed all the results on Amazon AWS, in most tests, Ignite must be at least 50% to 100% faster. Below are some results:
No Deadlock transactions (optimistic, serialization)
Hazelcast does not support non-deadlock transactions, so we compare the non-deadlock transactions of Ignite with the pessimistic transactions of Hazelcast.
Atomic operation
These results, as well as many others, are posted on the Ignite website at: https://ignite.apache.org/benchmarks/ignite-vs-hazelcast.html
My comment on Hazelcast is:
Map. put ()-> Map. set ()
We have updated the test, but this change does not produce significant differences and does not affect the previous results.
Synchronous and asynchronous backup
The configuration we tested between Apache Ignite and Hazelcast is 100% the same, so the Hazelcast declaration is completely incorrect. In addition, according to Greg's blog, they are testing on their own hardware and have changed the test scenario for some reasons, making it impossible for others to verify. Our tests, on the other hand, run on Amazon AWS and put the code on Github. Anyone can run and reproduce the results.
When we compare and test the performance of Ignite, GridGain, and other products, we are very serious and formal. In our test, we tested more than 20 different scenarios and use cases. Today we re-run all the tests and the results are the same as before.
GridGain Confirms: Apache Ignite Performance 2x Faster than Hazelcast
This article permanently updates the link address: