Intel Announces Optane DC Persistent Memory For The Enterprise, Expected To Go On Sale This Year

Intel has today announced the introduction of its Optane DC persistent memory, which as the name suggests, is targeted squarely at the datacenter, behaves similarly to DRAM, and offers persistent storage. Persistence is just part of the equation, though; Optane DC persistent memory is designed to improve performance, and in effect improve the capabilities of a given server.

At a press briefing this morning, Intel’s VP of Datacenter and GM of Xeon Lisa Spelman said that the company’s new persistent memory helps break through constricting methods that have governed computing for “more than 50 years”. That’s a lofty statement, but some enterprise customers might begin to agree after seeing some Intel’s statements.

During this briefing, Intel couldn’t wait to jump into benchmarks, so VP and GM of Datacenter Memory & Storage Alper Ilkbahar benchmarked a simple database reboot. It’s not entirely clear how large this database was, but considering the fact that it took 2,100 seconds to load after being rebooted on a typical DRAM-only server, we can tell it’s lofty. With the persistent memory version of the server, the database start time was reduced to a meager 17 seconds.

A speed-up of 125x is no joke, and it’s the kind of difference that is sure to raise some skepticism. I can’t validate any of the findings, but the technology’s design is easy enough to understand. If you cache data in fast memory that won’t disappear after a reboot, nothing has to be reloaded after a crash or reboot – once the data is reengaged, it’ll become accessible again.

To be truly useful for what it’s set out to do. Optane DC persistent memory modules need to offer densities far better than what we expect from DRAM. To that end, Intel has 128GB, 256GB, and 512GB persistent modules on tap, with support for up to 3TB per CPU socket.

Database reboots are a significant use case here, because it’s inexcusable to many to have to wait so long for a database to work on its startup for a half an hour or more every time it needs to happen. With the super-fast Optane DC persistent memory under-the-hood of a system, security patches won’t seem like such a massive inconvenience given the dramatic speed-up of reboots.

To help get the technology into the hands of developers quicker – because yes, supporting Optane in this way is not automatic – Intel is launching a developer subsite that will offer plenty of information to accelerate development. Qualified users will be able to essentially borrow server time, including a certain amount of Optane DC persistent memory, giving them real-world testing before committing to a test or full rollout. If you are immediately interested in learning more, you’ll want to hit up the link and sign up for a webinar which takes place tomorrow.

In addition to the Optane DC persistent memory, Intel also talked about its forthcoming QLC (quad-bits-per-layer) SSDs that it expects will become a natural progression in datacenters moving from mechanical storage, especially as cost continues to be brought down (something QLC itself helps with). One of Intel’s SSD rulers can hold 16TB, and its overall form-factor is very modest in comparison to even a single hard drive.

Intel says that its persistent memory is sampling to customers now, and it fully expects to release SKUs to market at some point this year. Since we’ve been hearing about this type of solution from Intel for a while, it’s nice to know that we won’t have to wait much longer to see large-scale deployments take place. Unfortunately, it will require a next-gen Cascade Lake Xeon processor to take advantage of, but we’re told that current-gen motherboards could support the modules just fine – though you’d of course want to double-check with the vendor when the time comes.