Disruption in Data Center Design Part III: The Disruptor
In The Innovator’s Delema, Clayton Christensen describes the fallout when new technology (at the time, the 3.5-inch hard drive) disrupted the data storage market. Simply put, the compact hard drive stored the desired data volume in less space. If a similar disruption took place today, it may look like cloud service providers delivering services with less floor space and utilities. They would no longer need to build and operate data centers of the size and scale that they do today. Disruptive technology is indeed the catalyst for abandoned data centers in Northern Virginia and across the country. Of the potential disruptors, there is one that may forever change data center design. It has its roots in the field of Biotechnology.
For decades, experimental biologists have relied upon
commercial equipment to rapidly and repeatably synthesize and sequence deoxyribonucleic
acid (DNA), the blueprint of life. Researchers have applied this same equipment
to successfully retrieve digital data encoded in synthetic DNA molecules. DNA
is a polymer and a nucleic acid. Two polynucleotide chains form the familiar
double helix molecular structure, pictured below
(a)
(b) DNA Animation (Source: Zephyris, CC BY-SA 3.0 <http://creativecommons.org/licenses/by-sa/3.0/>, via Wikimedia Commons)
rungs of a ladder, each nucleotide incorporates the nucleobases
cytosine (C), guanine (G), adenine (A), and thymine (T), arranged in base pairs:
A-T, and C-G. The arrangement of base
pairs encodes genetic information and now, digital data. Organick describes DNA
data storage by a six-step process. The steps are Encoding, Synthesis, Storage,
Random Access, Sequencing, and Decoding
In the initial step, an algorithm encodes digital
information in DNA sequences. Effective encoding has two objectives. The first
is managing errors. Organick estimates that during synthesis, substitution,
insertion, and deletion errors occur at a rate of 1% per base. These errors are
overcome by randomizing data and adding so-called outer coding. Outer coding introduces
redundant DNA sequences according to a Reed-Solomon error correction code.
Interestingly, Reed-Solomon is a common approach for forward error correction
(FEC) in radio frequency communication links and a variety of other
applications, including DVDs and QR code readers. Second is enabling random
access. Random Access, as discussed below, is a must-have property that eliminates
the need to read out the entire database every time the user wishes to access a
particular file. For random access, the system parses data into “payloads” of a
fixed size and adds molecular components that act as memory addresses. Adding
inner coding completes the encoding step. Inner coding helps convert bits to
sequences of synthetic DNA.
The second step synthesizes
DNA sequences according to the encoding scheme. This is analogous to writing data
to random access memory. The key performance metric of the Synthesis step is latency,
or the time required to synthesize the DNA sequences with the necessary
redundancy. Latency and throughput in the write step are much greater in DNA
data storage than conventional microelectronic memory. This is the factor that presently
limits DNA to a potential replacement for magnetic tape, suitable for back-ups
and long-term storage
There are two ways to host DNA for long term storage. The first
is within a living cell or in vivo. The second and more practical method
is in vitro, such as cold storage in a frozen test tube solution
Data Center in Space (Source: Wikideas1, CC0, via Wikimedia Commons)
dark conditions of
the second Lagrange point (L2) beyond the moon, where the James Webb Space
Telescope orbits today. Leo Bear-McGuinness discusses how researchers used a
resilient bacterium to host DNA encoded with data. Dubbed “Conan the
Bacterium,” it can survive high doses of radiation and vacuum conditions. This
suggests that future space-based data repositories may host DNA in vivo
Random Access is the third and most important process step. Directly
accessing a selected file starts by identifying primer pairs attached to the
file during the encoding step. Next is cloning
sequences associated with the file. Cloning uses a polymerase chain reaction (PCR),
a wet chemical technique that results in billions of clones of selected DNA
sequences in a few hours. This sets the stage for data retrieval by sequencing.
Two commercial sequencing instruments are used to detect the
molecules of the cloned DNA corresponding to the files of interest
Illumina NexSeq 500
Sequencer (Konrad Förstner, CC BY 2.0
<https://creativecommons.org/licenses/by/2.0>, via Wikimedia Commons)
instrument from Oxford Nanopore Technologies (ONT) permits data readout in real time (lower read latency) at the expense of throughput. Organick et al. used Illumina sequencing to retrieve over 108 bytes of data encoded in vitro DNA in 2017. Decoding the data detected in DNA sequencing is the final
step. This is where error correction codes that rely on redundancy are
paramount
Certainly, DNA atomic-scale data storage is not the only
potential disruptor of today’s data center ecosystem. Quantum memory is a
strong contender. This post does not suggest that the data center boom is a
bubble waiting to burst. It does present evidence of technology and other
trends that can fundamentally change data center economics. In a few decades, these
may give rise to decentralization where compact, low power, edge-deployed data
centers displace the resource-intensive data centers of today. When that time
comes, just like the old shopping malls, city managers will scramble to find
redevelopment solutions for data center sites. Some will succeed in turning empty data
centers into profitable mixed-use space on short notice. Others will not be so fortunate. In their
haste to replace the millions of dollars in lost tax revenue, some will make
bad bets.
References
Bear-McGuinness. (2018). Is DNA the Future of Data
Storage? Retrieved from YouTube:
https://www.youtube.com/watch?v=r8qWc9X4f6k
Ceze, L. N. (2019). Molecular Digital Data Storage
using DNA. Nature Reviews Genetics, 20(8), 456-466.
Organick, L. e. (2018). Random Access in Large-Scale
DNA Data Storage. Nature Biotechnology, 36, 242-248.
Wang, S. M. (2024, February). Data Storage Using DNA.
Advanced Materials. doi:10.1002/adma.202307499
Wikipedia. (2025, 08 22). DNA. Retrieved from
Wikipedia: https://en.wikipedia.org/wiki/DNA
Wikipedia. (2025, July 28). Retail Apocalypse.
Retrieved from Wikipedia: https://en.wikipedia.org/wiki/Retail_apocalypse




Comments
Post a Comment