Blog ›Hazelcast Enterprise: A Direct Challenge to Larry Ellison’s Vision for In-Memory Computing
mikojava

By mikojava

View all blogs by the author

Jun 11, 2014

Back to Blog

Hazelcast Enterprise: A Direct Challenge to Larry Ellison’s Vision for In-Memory Computing

Larry Ellison gave his vision for In-Memory Computing today.

His description began with some very sensible description of Row VS Columnar database formatting–for OLTP (transaction processing) or for OLAP (analytic processing).

So he characterizes the innovation of 12c Oracle “in-memory option” as being stored BOTH in row format and in a columnar format.

The core of this is gaining some of the pure performance benefits of columnar databases for OLAP at the same time as OLTP.

This kind of “converged” OLAP and OLTP is definitely something that is happening big time. The requirement to be able to do real-time analytics on top of streaming transactional data is undoubtedly a “holy grail” of data processing.

The mere existence of OLAP vs OLTP is an artificial distinction forced by the prohibitive cost of disks vs main memory.

So this much is very agreeable. Customers are increasingly thinking about things like “Big Data” and are pushing into systems like Hadoop which are principally disk based, and are running batch analytics.

The crux is to have column and row formats coexisting in-memory.

New kind of failure

He is singing the song of “scale out” which is now fault tolerant, available, reliable and he said “there are no other In-Memory Databases that are fault tolerant”

And he says “No other In-Memory Databases are scale out”.

He talks about how any node can fail and the system keeps running.

This again is quite sensible, however, it is also a property of Hazelcast In-Memory Data Grid, the system supports not only no single point of failure, fault tolerance, high availability and full resilience, but it is based on commodity hardware, not expensive proprietary Exadata boxes.

SO what is the fly in the ointment?

The vision is absolutely correct, but if you play the Oracle Twitter “drinking game” you need to take a drink every time Larry says “Exadata”.

Larry is selling the Sun Hardware platform now.

How many Sun boxes will he well with this strategy? This is a big difference with the strategy for HANA from SAP or from most previous Oracle launches. This time Larry is playing for all of the marbles and wants to be the Apple of Enterprise computing.

He’s interested in lots of the correct requirements:

  • Scale Out
  • Highly available
  • In-Memory
  • No single point of failure
  • Converged OLAP/OLTP
  • Resilient to failure
  • Scale out
  • Also Scale up
  • All your SQL runs unchanged
  • Cloud ready

The thing is, every single one of these bullet points are available in Hazelcast open source In-Memory Data Grid.

The fly in the ointment is that Larry is trying to sell extremely expensive boxes from Sun Microsystems now. Hazelcast can deliver this kind of high-performance In-Memory computing on top of ANY database including Oracle, and can do so with commodity hardware on open source and open standards.

Why 2014 wont be like 1984?

Like Apple, Larry Ellison wants to deliver both software and hardware vertically integrated.

This produces extraordinary vendor lock in and closed systems.

On June 10th Hazelcast launched Hazelcast Enterprise, an In-Memory Computing solution built on top of Open Source and Open Standards but made for the Enterprise.

This is why 2014 won’t be like 1984.