Microsoft Project Silica finds cheaper glass, boosts write speeds but trims capacity

Microsoft Project Silica research has identified cheaper glass media with accelerated writing speed but lower capacity than its original fused silica glass.

Up until now, Project Silica has used fused silica glass, quartz glass media measuring 75 × 75 × 2 mm (2.95 × 2.95 × 0.08 inches) and written data at up to 100 levels in the glass using polarization-based patterns – volume pixels or voxels – with femtosecond laser pulses. Voxels vary by x, y, z position, orientation, and size. Orientation is used to encode a color and the size is varied by changing the power of the laser pulse. Write speed is less than 1 Mbit/s and capacity is 1.75 TB/square inch and 5.15 TB per tablet.

Project Silica researchers have published a paper in Nature saying they have discovered new methods to read and write archival storage data, and found an alternative borosilicate glass medium. They achieved 2.02 TB capacity in a larger, 120 mm square, 2 mm thick piece of this glass, but stored 4.8 TB in a fused glass tablet. 

The write speed was 25.6 Mbit/s-¹ in each case, but with borosilicate glass they reached a faster 65.9 Mbit/s-¹ by using four laser beams. They say writing with 16 or more beams should be possible, implying 263.6 Mbit/s-¹ with 16 beams. That's 33 MB/s, which is much slower than LTO-10 tape's 400 MB/s uncompressed data write speed (1,000 MB/s compressed).

Borosilicate glass, made from silica and boron trioxide, is stable through rapid low-to-high temperature changes and used in chemical laboratory glassware and kitchen cookware. It's simpler to make and more widely available than fused silica glass. It also has a 10,000-plus-year lifetime, approximately the same as the quartz glass media. 

In this second-generation Project Silica system, there are two types of voxels: phase voxels in borosilicate glass relying on isotropic refractive index (RI) changes, and birefringent voxels (elongated nanovoids) based on anisotropic changes in fused silica glass, in which a voxel's property changes with the direction of measurement. Such voxels can store more than 1 bit, helping to give fused silica glass its greater capacity.

With phase voxels, the phase change of the glass is modified instead of its polarization, and only a single laser pulse is needed to make a phase voxel. The researchers say: "Our new pseudo-single-pulse regime shows the formation of elongated nanovoids with just two pulses, improving on previous work. We split each pulse into two: one that forms a void (seed pulse) and the other that elongates a previously formed void (data pulse)."

The writing speed is further accelerated with multiple beams per laser and reading is improved with machine learning-based decode to account for noise and inter-voxel cross-talk. High levels of three-dimensional inter-symbol interference in phase voxels can be mitigated with a machine learning classification model.

In the Nature paper's Project Silica schematic diagram above, the researchers labeled sections thusly: "a), Data are received from a user. These are prepared as a stream of bits, for example, using compression, encryption and FEC. b), The bits are encoded as symbols. One symbol corresponds to a modulator configuration. c),d), The glass sample is loaded into a write subsystem and the modulator settings are changed in time as the laser beam is moved relative to the glass. The symbols are written layer by layer, from the bottom up, to fill the full thickness of the glass. e), The data can be stored securely in the glass for more than 10,000 years. f), To read, we use an automated microscope with a camera to capture images of each 2D layer of voxels. g), Images are passed to a decoder to recover user data."

In detail, "user data is received as a stream of bits, to which additional bits are added using forward error correction (FEC) (Fig. 1a above). This ensures data integrity despite stochastic errors during writing and reading. Bits are then grouped into symbols (Fig. 1b). Each symbol corresponds to a voxel and stores more than one bit. The laser energy or polarization is modulated as the beam is moved relative to the glass sample (Fig. 1c, platter). Voxels are written in two-dimensional (2D) planes and stacked into three-dimensional (3D) volumes (Fig. 1d). In each plane, voxels are organized into sectors based on the field of view (FOV) of the read system. Sectors are stacked vertically into tracks. We read using wide-field microscopy (Fig. 1f). Symbols are inferred from images using convolutional neural networks."

A blog by Microsoft Partner Research Manager Richard Black says: "The reader for the glass now needs only one camera, not three or four, reducing cost and size. In addition, the writing devices require fewer parts, making them easier to manufacture and calibrate, and enabling them to encode data more quickly." The Gen 2 system also uses parallel writing. Here is a schematic diagram of the new writing subsystem:

See a full-size image here.

The researchers say: "A key system choice for a future Silica system is whether to use birefringent or phase voxels... Efficient formation of birefringent voxels can be achieved only in high-purity silica glasses, whereas phase voxels can be written in potentially any durable transparent media, for example, borosilicate glass as demonstrated here."

They make no choice, concluding: "We demonstrate that Silica is a viable storage system by fully recovering user data using FEC (Forward Error Correction), and show through accelerated aging experiments that our modifications last more than 10,000 years at room temperature. In short, our results demonstrate that Silica could become the archival storage solution for the digital age."

So is productization getting nearer? Black says: "The research phase is now complete, and we are continuing to consider learnings from Project Silica as we explore the ongoing need for sustainable, long-term preservation of digital information." It would seem not.