In-memory computing heads for the mainstream, but obstacles remain

Defence, banking and telecoms may be among the current main users of in-memory computing but the technology will shift into the mainstream over the next two years, despite a number of obstacles.

Cheaper DRAM prices and a wider availability of software to exploit in-memory computing are driving that shift, leading to at least 35 percent of mid-sized and large companies adopting it by 2015 from less than 10 percent in 2012, according to analyst firm Gartner.

But new adopters of the technology — which allows faster processing by storing data in memory rather than on disks — will face challenges including a lack of standards, unfamiliar architectural complexity and monitoring issues.

"There is a learning curve that the industry has to go through. It's nothing unsolvable but it's a change in paradigm and you have to get used to it and learn," said Massimo Pezzini, vice president and Gartner Fellow.

Fundamentally, in-memory computing means all the data that your application needs is already available in memory somewhere, Pezzini said.

"However, this of course poses challenges in terms of how you manage recovery, for example — or disaster recovery, backup, monitoring, management in a scenario like this. We are not used to it. We are used to dealing with databases and tapes and disks. There is some complexity that needs to be addressed," he said.

"To store large amounts of data in memory, you may need to split it across multiple servers. We all know that monitoring and managing these distributed systems is far from simple. That's undoubtedly another challenge that customers are facing. "

Absence of in-memory standards

Pezzini said the absence of standards represented another issue for more widespread adoption.

"In in-memory technology there are no standards. Every product is developed in its own way. There are no standard APIs, tools or protocols, which of course makes things a little harder when it comes to transferring skills, for example," he said.

"If you're an expert in SAP in-memory technology, it's not so simple to turn you into an expert in Oracle in-memory technology, for example. Lack of standards is an issue in terms of building up the skills pool. It also poses application portability and interoperability issues," Pezzini added.

"These issues will slow down adoption — there's no question about that. If there were commonly agreed standards available of course things would move faster but on the other hand the market is growing very fast."

Pezzini said rapid changes in the technology itself may help offset certain obstacles.

"Today the servers and the blades you see have not really been designed specifically to support in-memory computing type of applications. They simply have a lot of memory but their design point is a traditional one," he said.

"But now we are beginning to see some vendors providing hardware specifically designed to support in-memory computing type of applications."

Embeddedin-memory computing

As well as falling hardware costs and the availability of appropriate software, Pezzini attributed the growth in in-memory computing to an increase in products that come to customers with in-memory computing technology built in.

"The most notable example is SAP. ERP and CRM are now available on the SAP in-memory database, HANA," he said.

Pezzini also cited Software AG's webMethods 9.0 suite of integration and VPN platforms, which has in-memory technology embedded to support scalability and performance.

"So more and more in-memory computing technology comes to customers embedded into traditional products, packaged applications, middleware, VPN tools, and business intelligence tools," he said.

On top of software availability and falling memory costs, the efforts of ERP vendor SAP to promote its HANA in-memory computing is also boosting the profile of the technology beyond it present niches.

"Undoubtedly, all the noise and marketing dollars that SAP is spending have helped a lot in raising awareness of in-memory computing technology," said Pezzini.

"However, it is important also to note that SAP is not the only vendor providing in-memory technology — even if it is very vocal," he added.

IBM, Oracle and Microsoft in-memory technology

Massimo cited the efforts of IBM, Oracle and Microsoft in this area.

"They all have in-memory technologies, either as independent standalone products or as capabilities of their established products. For example, Microsoft in SQL Server 2012 has incorporated an in-memory database, which comes to customers for free," he said.

"But you also have companies like VMware, for example, and Software AG or Tibco. These mid-sized vendors have in-memory computing technology as part of their portfolio."

Pezzini said there are also plenty of little companies or specialist vendors, such as Giga Spaces in Israel and Quartet FS in the UK.

"In our database, we have almost 60 vendors providing various forms of in-memory computing technology. A year ago there were maybe 40," he said.