Accelerate Application Scalability with BigMemory Hybrid

This is a guest post by Foram Gandhi, Senior Product Manager, Terracotta.

According to a research on ZDNet, SSDs are set to grab over one third of PC storage solutions market by 2017[1]. Enterprise SSD demand has been growing exponentially and is expected to grow through 2017 [2]

Why is SSD so popular?

SSDs are known to be energy efficient & maintenance efficient over long run as compared to spin disk. SSDs are known to be lightweight compared to spin disks [3]. SSDs perform significantly better in terms of IOPS as compared to spin disk [3].


Why would application architects choose SSD over HDD for their architecture?

Applications have an ever-increasing need for speed. When developers want performance improvement in their high-scale application without re-architecting it, a combined solution of in-memory and SSD can phenomenally change the performance characteristics of application. The use of SSD is especially relevant in cases of high scale in the order of hundreds of terabytes.

Of course if an application is at cutting edge of performance then a pure in-memory solution is the best choice. Example of such application is fraud detection. Online transaction approvals and denials need to happen in sub-milliseconds and the payment industry is getting ever more competitive. Fast execution of fraud detection algorithm is key to sustain business. A pure in-memory solution is best fit for these use cases where micro second latency expectation is a norm.

However, if developers want a mix of cost efficiency with improved performance for a high scale system, use of SSD can fit the bill.


What is BigMemory Hybrid?

At Terracotta we focus on high-performance, high-scale applications. Terracotta BigMemory Hybrid is an extension to In-memory data management platform to also support data management on SSDs. The same quality of service in terms of fault tolerance, high availability, consistency, persistence can be obtained by BigMemory Hybrid. The only difference is that data will be stored in SSD as well as in-memory. So In addition to managing memory, the system will manage data on SSD and take care of file handling and data compaction.

When a read operation is triggered, data is fetched from SSD. However, when a write operation is triggered, data is written to write ahead log in memory to get an initial acknowledgement. User does not have to manage movement of data between heap memory, off-heap memory and SSD. BigMemory Hybrid will implicitly manage this.

In terms of resource utilization, when BigMemory Hybrid is turned on, the usage of RAM reduces drastically thus resulting in cost savings. The results are shown in chart below.

With BigMemory Hybrid is turned on, DRAM requirements are reduced for the same live dataset size at the cost of longer read latencies and slightly increased heap usage. The reason we are showing high Flash usage in this diagram for off-heap deployment is because we are assuming that persistence is turned on for the deployment.

The customer just needs to add a tag in tc-config.xml to be able to use hybrid capabilities which will reduce their requirement for additional RAM. Configuration example below:

To learn more, please contact me at



Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>