On-server memory solutions are emerging as a key element in warehousing, data lakes and other initiatives targeting Big Data and the Internet of Things, and are even making a run at high-speed transactional processing and other more traditional enterprise applications.
But is it always the right answer for high-speed workflows? And do the multiple varieties of memory have any bearing on successful outcomes?
Like nearly all infrastructure, more advanced technologies produce better results but come at a higher cost premium. DDR4 silicon, for example, offers 50 percent more bandwidth and is 35 percent more energy efficient than its DDR3 predecessor. To date, however, DDR4 has been seen in top-end servers and desktops, although vendors like Dell and HPE are starting to trickle it down into lower-end PowerEdge and Proliant machines, says IDG’s Agam Shah. These models can be had for less than $1,000 and boost internal storage by anywhere from half to more than three times that of previous servers.
Currently, however, the prevailing form of in-memory in the server farm is Flash, not DRAM, which makes sense considering it is both cheaper and more resilient, according to Diablo Technologies. Integrating Flash as system memory can deliver four times the capacity of DRAM, which translates into a 90 percent server consolidation rate, depending on the architecture, and a reduction in power consumption of about 70 percent. And with solutions like Diablo’s Memory1 platform, which uses JEDEC-compliant LRDIMM/RDIMM hardware to fit in a standard DDR4 memory slot, Flash can be deployed on standard servers with no changes to CPU, the motherboard, operating systems or applications.
The rise of in-memory solutions is one of the key reasons Intel is heading back into the memory business after a nearly three-decade hiatus, says Endpoint Technologies Associates President Roger L. Kay. The company originally bailed on memory in the mid-80s under pressure from Japanese competitors, but has now earmarked upwards of $5.5 billion to purchase memory fabs in Asia. Its first venture will be in the Chinese city of Dalian where it will convert an existing facility to begin making 3D NAND Flash chips. Intel is in a much stronger position these days to dictate the economics of the memory business, Kay says, and the market for memory subsystems now extends way beyond mere PCs and servers.
But in time-honored fashion, just as everyone is getting comfortable with one set of technologies, someone comes along to upend it all. In the memory business, that could very well be Micron Technology, which is looking beyond merely moving memory close to compute and actually turning memory into compute, says Next Platform’s Nicole Hemsoth. The company’s Automata processor utilizes a high degree of parallelism for bringing large numbers of disparate instructions to bear on a single problem. This enables them to conduct extremely fast search and pattern matching across even the largest of data sets, offering support for applications ranging from analytics and security to advanced genomics and neural networking. Once the company works out the programming models, production should ramp up fairly quickly as the device is built on largely the same process used for current DRAM solutions.
The faster things get in the digital economy, the more the enterprise will need to deploy memory-based infrastructure to meet user and application demands. This shouldn’t be seen as the death knell for the traditional storage array, but more of a shifting of priorities as the enterprise seeks to capitalize on emerging trends like mobile computing and Big Data analytics.
And the best news is that memory solutions still appear to be very early on their development path, meaning the real upgrades to enterprise-class performance and reliability are still to come.
Arthur Cole writes about infrastructure for IT Business Edge. Cole has been covering the high-tech media and computing industries for more than 20 years, having served as editor of TV Technology, Video Technology News, Internet News and Multimedia Weekly. His contributions have appeared in Communications Today and Enterprise Networking Planet and as web content for numerous high-tech clients like TwinStrata and Carpathia. Follow Art on Twitter @acole602.