
While it may not have the same evocative imagery as “cloud computing” or the conceptual heft of “big data”, there’s a third computing trend gaining momentum in the enterprise space. It’s called “in memory computing” and it has the potential to drive just as much change as any IT trend in the enterprise today. Let me explain.
IN MEMORY = SOLID-STATE MEMORY
Like cloud computing and big data, the concept of in-memory computing (IMC) has been around a long time, waiting for the right market conditions to trigger widespread adoption. IMC relies on the same solid-state flash memory modules found in everything from USB thumb drives to iPads and smart phones in order to operate.
The prices for this flash memory, known as DRAM and NAND in the industry, have dropped precipitously over the years as you can see in this graph developed by EE Times, making it feasible to purchase terabytes of it at a time.

While flash memory will likely never compete with rotating media (conventional hard disc drives, or HDDs) for the lowest price-per-gigabyte, it has one advantage hard drives can’t touch: lower latency.
Latency measures the time it takes to retrieve and deliver a specific block of stored data. Flash memory performs this task in about 83 nanoseconds. Rotating media performs the same task in 13 milliseconds. Nanoseconds vs. milliseconds. What’s the difference?
Christian Paredes, in a blog post on Scoutapp.com, explains the difference masterfully below:

So while the difference may be less than the blink of an eye, scaled up over millions of transactions the difference in performance is massive.
OVERHAULING THE DATA CENTER ARCHITECTURE
In-memory computing taps into this performance difference by flipping the usual data center architecture on its head. In all of today’s data centers, including Avnet’s, data is stored on predominantly on hard drives due to their low cost and good reliability and performance. That data is often backed up on tape drives, due to their extreme low cost.
But hard drives are slow, and tape is even slower. That’s an issue as the importance of speed is becoming just as important in the data center as reliability and price.
In IMC solutions, company data is stored in solid-state memory modules where it can be accessed almost instantly due to low latencies. Backup and disaster recovery then takes place on hard drives, and tape is phased out over time.
Though the initial investment may be slightly higher than rotating media, the lack of moving parts in the solid-state flash memory promises to reduce maintenance costs over time. It’s a fundamental shift and an evolution in the way data centers are architected.
IMC & BIG DATA
As I discussed in a recent blog post, Big Data promises to help businesses manage and make sense of the exponential growth in information coming at them in real time. But the bigger the data set being managed and the faster it changes, the more IMC architectures are needed to meet the minimum performance thresholds that Big Data applications demand.
As a transaction-driven company ourselves looking to roll out our own Big Data initiative this year, Avnet is increasingly embracing in-memory architectures to ensure that our infrastructure isn’t the bottleneck in an environment where every millisecond counts.
You can expect to hear more about Avnet’s own experiences with in-memory computing and architectures in the coming months.
Like cloud computing and big data, it’s an old idea whose time has finally come.
- - Steve
Share or print this post: