EOS Tribe Q3 Infrastructure Upgrades

eostribe (55)in #eos • 6 years ago (edited)

We are pleased to announce an important upgrade to the EOS Tribe infrastructure.

After running our servers in two data centers for several months post-launch, we came to the following conclusion:

It’s far better and economical to run a single geographic location node with strategically distributed end points and provide a faster/lower latency connection for the network instead of spreading ourselves thin across multiple data centers.

Hence we are consolidating all bare-metal servers to our Utah Data Center, on a dedicated low latency 10GB connection with a 1GB Switch:

The cloud is based on VMWare 6.5U2 and patched to a fully PCI compliant configuration.

The new infrastructure has many benefits. The cloud is married to an MPLS backbone, protecting the cloud from outside direct access.

The portal maintains this through a coupled AES entry point with 2 key authentications and then utilizes 2 factor recognition.

Each server is comprised of 80 cores and has a combined 1.5TB RAM with over 48TB of redundant ZFS backed storage.

The switches are low latency state of the art Quanta and provide robust 10G 40G and 56G capabilities, all while using state of the art silicon to drive extremely low latency.

The ZFS appliance is configured to yield the best performance. We use vDEV mirrors with 12 hr snapshots and packet consistency to ensure that NFSv4.1 runs perfectly. We also ensure that storage uses the latest encryption method to ensure data is sound. We can tune, size, latency, and performance all while on the fly.

VMWare images come from a lot of hard work and take the headache out of maintaining the v-Library. Once installed in the cloud, they run rock solid. Tuned and dialed for VMW performance and unified with ZFS, they are error and trouble free for the end users.

We have moved into NVME/SSD Storage - it is much faster with sustained sequential read/write at over 500MB p/s and upward reading of 2GB ps. We have 256GB Ram to go into the Cache controller of the SC on the Oracle and that will improve our current load performance by 4 times.

Our new EOS Blockchain Deployment Diagram now includes 4 load balanced Full Nodes synced over bridge connection into Trusted WireGuard Seed Node, which in turn synchronizes primary and backup Block Producer nodes running on Trusted WireGuard network:

We are currently working on adding MongoDB support after stabilizing the MongoDB plugin.

Fully synchronized MongoDB plugin will allow us to provide additional API functionality to our REST endpoint and removes load from single threaded full nodes into a multithreaded web application.

Our long-term goal is to completely replace MongoDB with Kafka Queue and Hadoop cluster.

This our first step towards an infrastructure that is Bigger, Faster, Stronger and more Secure. We aim to build and maintain a infrastructure that is ready for the mass adoption of EOS.