Skip to main content

ubiquiti 10gbe upgrade - network infrastructure overhaul

finally upgraded the home network.

10gbe backbone. no more bottlenecks.

the problem
#

running 4 Dell servers, 3 workstations, NAS, and market data feeds.

old setup: 1gbe switch, consumer router.

bottleneck was obvious. timescaledb replication lagging. market data feeds competing for bandwidth.

the new setup
#

core switch: Ubiquiti USW-Pro-24-POE

  • 24 ports gigabit with POE
  • 2x 10gbe SFP+ uplinks
  • layer 3 switching

aggregation: Ubiquiti USW-Aggregation

  • 8x 10gbe SFP+ ports
  • connects all servers at 10gbe

router: Ubiquiti UDM-Pro

  • 10gbe WAN/LAN
  • built-in controller
  • IDS/IPS capability

total cost: $2,400

worth every penny.

the installation
#

                    [ISP 1gbe]
                        |
                   [UDM-Pro]
                        |
              [USW-Aggregation]
               /    |    |    \
          Dell1  Dell2  Dell3  NAS
          10gbe  10gbe  10gbe  10gbe
                        |
              [USW-Pro-24-POE]
               /    |    \
          WS1   WS2   WS3
          1gbe  1gbe  1gbe

servers on 10gbe backbone. workstations on gigabit.

performance gains
#

timescaledb replication:

  • before: 45ms average, 200ms spikes
  • after: 8ms average, 15ms spikes

market data ingestion:

  • before: occasional packet loss during high volatility
  • after: zero packet loss

NAS backup:

  • before: 112 MB/s
  • after: 950 MB/s

inter-server latency:

  • before: 0.4ms
  • after: 0.08ms

configuration details
#

VLANs for network segmentation:

VLAN 10: Trading (servers, data feeds)
VLAN 20: Management (controller, monitoring)
VLAN 30: Guest (wifi, IoT)
VLAN 40: Personal (workstations, media)

trading VLAN gets priority. QoS rules ensure market data never competes with backups or media streaming.

monitoring
#

unifi controller gives decent visibility.

added custom grafana dashboard for:

  • per-port bandwidth utilization
  • packet loss metrics
  • latency between VLANs
  • switch CPU/memory

lessons learned
#

  1. 10gbe cables matter - use CAT6A or DAC cables. CAT5e won’t cut it
  2. SFP+ modules are finicky - stick with Ubiquiti branded for compatibility
  3. VLAN planning first - redesigning VLANs after deployment sucks
  4. document everything - port assignments, cable runs, IP schemes

ROI
#

$2,400 upfront.

but: faster replication = less slippage. zero packet loss = no missed signals.

one avoided bad fill pays for the upgrade.

plus the NAS backups are 8x faster which means I actually run them daily instead of weekly.


2:18am friday. network infrastructure overhaul complete. 10gbe backbone with ubiquiti gear ($2,400 total). timescaledb replication: 45ms → 8ms. packet loss: eliminated. NAS speed: 112 MB/s → 950 MB/s. worth it for zero packet loss during high volatility.

-AK

Related

upgrading chicago colocation to 10gbe - latency improvements
chicago colocation server needed upgrade. 1gbe connection = bottleneck. current setup # location: chicago datacenter (equinix CH1)
thinking about server upgrades
current setup is working but thinking about upgrades. also might be procrastinating dealing with trading losses. current setup # home server:
timescaledb optimization - 3 million rows per day
been putting this off for months. timescaledb getting slow. finally fixed it. the problem # my options flow data pipeline ingests about 3 million rows per day.
vacation algo management - what i learned from 8 days offline
took 8 days completely offline. first time since 2020. what I found was about managing algos during vacation. the pre-vacation protocol # T-5 days: stop opening new positions >5 DTE
chicago colocation 18 months - ROI analysis and lessons
18 months with chicago colo. time for ROI analysis. the setup recap # december 2023: deployed chicago colo
redis caching optimization - 40% latency reduction for market data
optimized redis caching during honeymoon downtime review. 40% latency improvement. the problem # before optimization: market data fetch: 180ms avg