Feiyue Zhai
Papers
Overview
Topics
Knowledge
Graph
Tools
GitHub
PD Disaggregation Simulator
Client / Request
Concurrency
Total
=10xConc
ISL
OSL
Range
Prefill (DEP) + Router
Nodes
TP
DP
EP
PF TPS
tok/s/rank
PrefixCache
%
ChunkSz
PFMaxReq
TxCost
%
Decode (DEP)
Nodes
TP
DP
EP
MaxReq/Rank
TPOT
ms
MinTPOT
ms
Simulation
Speed
0.25
0.5
1
2
4
8
16
32
64
8x
Pause
Reset
In-Flight
PF Queue
KV Xfer
Prefill Running
Decode Running
Prefill TPS
Decode TPS
Input Length (ISL) Distribution
Output Length (OSL) Distribution
Statistics
Sim Time
0:00
Client Pool
0
Client In-Flight
0
Completed
0
Throughput
0
PF Queue
0
Transfer Queue
0
DC Running
0
PF Utilization
0%
PF Bubble
0%
DC Slot Util
0%
DC Bubble
0%
E2E Avg
-
E2E P50
-
E2E P99
-
Avg TTFT
-
E2E Cost
-
E2E PF TPS (t/gpu)
-
E2E DC TPS (t/gpu)
-
E2E TPS (t/gpu)
-
DC Max Cap
-