HP has knocked back the commercial launch of its memristor storage technology, which is shaping up to be a cheaper, more efficient successor to flash. If its release is delayed, where does that leave the technology industry?
Co-developers HP and Hynix initially planned to bring memristor technology to market in 2013. But this release date has now been moved to summer 2014, because Hynix wants time to shift its business, The Register reported on Monday. HP Labs fellow Stan Williams said that the storage specialist needs to be able to cope with the cannibalisation of its mainstay flash products by the ultra-efficient non-volatile storage technology provided by memristors.
If this occurs, it will be a dismal thing for the tech industry. Memristors are a way to unlock the true potential of the public and private cloud, mobile devices and high-performance computing. Any commercial delay will put a drag on development of such products taking advantage of the storage technology.
When given the opportunity to confirm or contradict The Register report, HP side-stepped the issue.
"As with many other ground-breaking technologies being developed at HP Labs, HP has not yet committed to a specific product roadmap for memristor-based products," it said in a statement. "HP does have internal milestones that are subject to change, depending on shifting market, technology and business conditions."
Why memristors matter
A memristor is an electrical component that has resistance to electrical current, but the resistance changes as you put the current through. When you take the current away, the memristor preserves the last state it was in. Essentially, it's an analogue memory circuit.
HP has already proved that memristors can perform both logic and storage operations. This means computation could be performed much closer to the storage layer than in existing technologies. In turn, this means companies can make devices that perform faster and use less power at the same time.
In fact, the technology is eventually expected to replace flash, DRAM and even hard drives.
This is because it roughly doubles the storage capacity of those storage mediums, while keeping the price the same.
HP has developed the technology through HP Labs. It got into the effort because it buys vast quantities of storage technology for its public and private clouds, servers, computers and mobile devices, and it wants to lower prices and up performance. Hynix is involved because it knows it will be able to sell the technology to everyone that makes sophisticated electrical equipment.
So why the delay? According to The Register, Hynix needs time to alter its business to cope with the expected cannibalisation of its flash by the technology, and as Hynix will be the one fabbing the stuff, there's not much HP can do here.
But is this right for the industry? I don't think so. A delay would be damaging, because the memristor has the potential to solve one of the two major problems facing the technology industry: storage I/O.
Let me put it this way: it used to be that an application's performance was limited by the clockspeed of the processor it ran on. But as applications have become distributed due to the rise of cloud/distributed computing, the block on many applications has moved to their input/output layer. This is true in both private and public clouds.
This block exists in two areas: latency to read and write data, and latency to shuffle the data between bits of kit. Memristors solve the first block, by making application access times significantly faster, while major IT companies such as Intel, IBM and Fujitsu are already working on photonic interconnects to solve the network bandwidth snafu.
If they delay the launch of the technology, HP and Hynix are withholding something that could benefit not just the technology industry, but every consumer of digital products on the planet.
Cost benefits all round
The digital items we buy — whether it be MP3 songs or rentable compute time from Amazon — are costed partially according to how much money it takes to perform a computation. Introduce memristors, and the cost of computation gets dramatically smaller.
When it comes to the enterprise, in cut-throat industries like cloud computing, we can expect these cost changes to be passed on to buyers (Amazon has cut its cloud pricing 19 times in six years; just imagine what they could do if their storage costs halved). In more established, less disrupted industries, we can expect a nice uptick in performance.
Private clouds will get cheaper as well, as the total cost of ownership on a per-server basis will fall. This means businesses will be able to get more storage bang for their buck on technologies based on memristors.
The technology will also have beneficial effects on storage-hungry supercomputing, by increasing the number of experiments/simulations that can be performed in any time period. This has the potential to lead to a bevy of improvements for humanity — better crops, better drugs, better weather prediction.
So, what gives, HP-Hynix? Why not take the risk of cannibalisation and clean up in the long term? At the end of the day, everyone but Hynix's storage rivals stand to benefit from the introduction of the technology.