Skip to content

Latest commit

 

History

History
104 lines (74 loc) · 3.46 KB

README.md

File metadata and controls

104 lines (74 loc) · 3.46 KB
Raw

Filecoin Optimized Sealing

This repository is under active development and is not expected to work in a general setting. It should be considered view only.

This respository optimizes the filecoin sealing process for throughput per dollar. It is a companion to https://github.com/supranational/filecoin_pc1 and will read sealed PC1 column data from NVMe devices under SPDK to generate tree-r and tree-c files that are compatible with the existing sealing software. The current software is configured to seal 64 sectors in parallel.

Like filecoin_pc1 this is an initial prototype release.

What's coming:

  • Integrate filecoin_pc1, filecoin_pc2, c1 into a single repository
  • Support for pipelined and simultaneous jobs
  • More seamless support for sealing a variety of parallel sector counts
  • Simplified configuration

Performance

Using the reference platform outlined below this software generates PC2 tree-r and tree-c files for 64 sectors in approximately 70 minutes.

Reference Platform

We will specify a reference configuration with the final release of the software. Our current testing configuration consists of:

  • Threadripper PRO 5975WX
  • ASUS WRX80E SAGE Motherboard
  • 256GB Memory
  • 15 Samsung 7.68GB U.2 Drives
  • 4 Supermicro AOC-SLG4-4E4T NVMe HBA
  • 2 NVIDIA GeForce RTX 3080

Building

Enable Huge Pages (1GB):

Though huge pages aren't strictly necessary for PC2, the software is configured to use them since it is expected to run alongside PC1.

sudo vi /etc/default/grub
GRUB_CMDLINE_LINUX_DEFAULT="default_hugepagesz=1G hugepagesz=1G hugepages=128"
GRUB_CMDLINE_LINUX="default_hugepagesz=1G hugepagesz=1G hugepages=128"
sudo update-grub
sudo reboot

You can confirm huge pages are enabled with:

grep Huge /proc/meminfo

# Look for:
HugePages_Total:     128
HugePages_Free:      128

Additionally you may need to enable huge pages after boot using:

sudo sysctl -w vm.nr_hugepages=128

Build SPDK in the Parent Directory:

git clone --branch v22.09 https://github.com/spdk/spdk --recursive spdk-v22.09
cd spdk-v22.09/
sudo scripts/pkgdep.sh
./configure --with-virtio --with-vhost
make
sudo env NRHUGE=128 ./scripts/setup.sh

SPDK is used for directly controlling the NVMe drives.

Update NVMe Controller Addresses in src/column_reader.cpp:

SPDK can be used to identify attached NVMe devices and their addresses with the following command:

sudo ./scripts/setup.sh status

For more extensive information about attached devices:

sudo ./build/examples/identify

This will show the NVMe disks (controllers) along with their addresses, which will resemble 0000:2c:00.0. The address list in src/column_reader.cpp must be updated to reflect the addresses that will be used.

In addition, if you have a different number of drives than the reference configuration, then you must update NUM_CONTROLLERS in /filecoin_pc1/src/sealing/constants.hpp

The drive configuration for PC2 must match PC1 for correct functionality.

Build the PC2 Binary

./build.sh

Running

PC1 should be run prior to running PC2 to populate the DRG data on the NVMe drives. See https://github.com/supranational/filecoin_pc1 for related instructions.

Once this is completed, run:

sudo rm results/* # Delete any existing output files
sudo ./pc2

This will create a "results" directly that stores the PC2 output files.