Capacity & Throughput
Obsidian is optimized for low-latency, high-frequency data publication. Through the Silica Protocol, the system can sustain 10,000+ messages/second while data throughput scales independently of EVM execution.
Message Capacity
Per Slot
Lane count
8 parallel lanes
SM per lane
2 MB
PM per lane
2 MB
Total SM per slot
16 MB (8 × 2 MB)
Total PM per slot
16 MB (8 × 2 MB)
Combined max per slot
32 MB
Max message size
8 KB
Slot time
12 seconds
Message throughput is constrained by:
Per-lane sidecar limits (
maxSmSidecarBytes,maxPmSidecarBytes).Slot time (
SECONDS_PER_SLOT, default 12s).Lane count (default 8).
Per Day
Slots per day
86,400 / 12
7,200
Max SM per day
7,200 × 16
115.2 GB
Max PM per day
7,200 × 16
115.2 GB
Max total / day
7,200 × 32
230.4 GB
Per-lane sidecar limits determine the upper bound for messages:
slotsPerDay = 86,400 / SECONDS_PER_SLOT(default: 7,200)maxBytesPerSlot = laneCount × (maxSmSidecarBytes + maxPmSidecarBytes)maxMessagesPerSlot = maxBytesPerSlot / avgMessageSize
Realistic Throughput
Actual throughput depends on message sizes:
1 KB
32,000 (32 MB / 1 KB)
230 million
8 KB (max)
4,000 (32 MB / 8 KB)
28.8 million
Practical throughput depends on lane utilization and message distribution.
Latency
PM (low competition)
~12 seconds (next slot)
PM (high competition)
~12 seconds (if top bid in lane)
SM (lane has space)
~12 seconds (next slot)
SM (lane busy)
Variable (FIFO within lane)
Finality
~192 seconds (2 epochs)
Lane Limits
SM per lane per slot
2 MB
maxSmSidecarBytes
PM per lane per slot
2 MB
maxPmSidecarBytes
Total per lane
4 MB
SM + PM combined per lane
Total per slot
32 MB
8 lanes × 4 MB
Max headers per lane
1
maxHeadersPerLanePerSlot
Max msgs per header
1000
maxMessageCountPerHeader
When mempool is full, new messages are rejected until space opens.
Growth Projections
Data growth depends on average message size and lane utilization:
slotsPerDay = 7,200maxBytesPerSlot = 8 lanes × 4 MB = 32 MBmaxDailyBytes = 7,200 × 32 MB = 230.4 GBmaxYearlyBytes = 230.4 GB × 365 = ~84 TB
This is why sharded archives matter.
Last updated