After carefully considering all hardware solutions, evaluating industry benchmarks and participating in community discussion, we have selected what we believe to be the best topology for our BP network architecture:
Utilizes two locations with matching hardware to achieve geographical dispersion.
Dual data centers are fortified within military grade secure subterranean facilities in Israel, 15 meters underground and with distinct physical separation between hosting rooms and support system rooms.
In each location, Block Producers aren’t directly connected to the Internet, and participate in block production through two full nodes.
Full nodes are connected to the Internet behind a firewall with two different Internet providers.
The public endpoint for the system is behind a Google Cloud Load Balancer and will also utilize BGP routing. This provides the highest level of DDoS mitigation and protection against DDoS attacks.
ESTIMATED LAUNCH PLAN by June 3, 2018
Bare Metal Servers
Two Separate Locations with matching hardware
- LAN Network Storage
- Intel Xeon 16 cores 2.4 GHz 128GB RAM - 2 Full Nodes
- Intel Xeon 16 cores 2.4 GHz 128GB RAM - 1 Block Producer
- Backup UPS for 72 hours
- Hardware firewall
- 10GB internal switch
- 1GB external Internet switch
Each location will include:
- BP network can scale servers and RAM as needed to maintain optimal throughput as the EOS network grows.
- Utilize world-class VHDL developers to build and support the creation of open-source hardware acceleration for EOS. This includes public code for accelerating low-level virtual machine (LLVM), hashing math and eventually an EOS processor, to increase bandwidth and transactions per second on EOS.
- Advance open-source development efforts in custom Kernel and LLVM optimizations in order to achieve higher TPS on the same hardware.
- Maintain maximum flexibility and constant communication with the EOS community to integrate ever-evolving industry best practices.