hypersdk

Performance Tuning Guide

Table of Contents

1. Export Performance
2. Network Optimization
3. Disk I/O Optimization
4. Memory Management
5. CPU Optimization
6. Database Performance
7. Benchmarking

Export Performance

Parallel Downloads

export:
  parallel_downloads: 8  # Default is 4
  download_workers: 8

# CLI override
./hyperexport -vm myvm -parallel 8

Connection Pooling

connection_pool:
  max_connections: 10  # More concurrent vSphere connections
  min_connections: 2
  idle_timeout: "10m"
  max_lifetime: "2h"

Compression vs Speed

export:
  compress: true
  compression_level: 1  # Fastest (default is 6)
  # 1 = fastest, 9 = best compression

# For maximum speed, disable compression
./hyperexport -vm myvm -compress=false

# For maximum compression
./hyperexport -vm myvm -compress -compression-level 9

Buffer Size

export:
  buffer_size: 16777216  # 16MB buffer (default 8MB)
  read_buffer_size: 8388608  # 8MB
  write_buffer_size: 8388608  # 8MB

Concurrent Jobs

daemon:
  max_concurrent_jobs: 20  # Process more jobs simultaneously
  download_workers: 8

Skip Verification

# Skip checksum verification for speed (not recommended)
./hyperexport -vm myvm -verify=false

Network Optimization

Network Bandwidth

export:
  bandwidth_limit: 0  # Unlimited (default)
  # Or limit: "100M" for 100 Mbps

TCP Tuning

# Increase TCP buffer sizes
sudo sysctl -w net.core.rmem_max=134217728
sudo sysctl -w net.core.wmem_max=134217728
sudo sysctl -w net.ipv4.tcp_rmem='4096 87380 67108864'
sudo sysctl -w net.ipv4.tcp_wmem='4096 65536 67108864'
sudo sysctl -w net.ipv4.tcp_congestion_control=bbr

# Make permanent
cat | sudo tee -a /etc/sysctl.conf <<EOF
net.core.rmem_max=134217728
net.core.wmem_max=134217728
net.ipv4.tcp_rmem=4096 87380 67108864
net.ipv4.tcp_wmem=4096 65536 67108864
net.ipv4.tcp_congestion_control=bbr
EOF

sudo sysctl -p

Network Interface Tuning

# Increase ring buffer
sudo ethtool -G eth0 rx 4096 tx 4096

# Enable TCP offloading
sudo ethtool -K eth0 tso on
sudo ethtool -K eth0 gso on
sudo ethtool -K eth0 gro on

# Increase queue length
sudo ifconfig eth0 txqueuelen 10000

MTU Optimization

# Test optimal MTU
ping -M do -s 8972 vcenter.example.com

# Set jumbo frames (if supported)
sudo ip link set eth0 mtu 9000

Disk I/O Optimization

Filesystem Selection

# XFS for large files (recommended)
sudo mkfs.xfs -f -l size=128m -d agcount=32 /dev/sdb
sudo mount -o noatime,nodiratime,logbufs=8,logbsize=256k /dev/sdb /exports

# Or ext4 with optimizations
sudo mkfs.ext4 -E lazy_itable_init=0,lazy_journal_init=0 /dev/sdb
sudo mount -o noatime,nodiratime,data=writeback,barrier=0,commit=60 /dev/sdb /exports

I/O Scheduler

# Use none (noop) for SSD
echo none | sudo tee /sys/block/sdb/queue/scheduler

# Or deadline for HDD
echo deadline | sudo tee /sys/block/sdb/queue/scheduler

# Make permanent
cat | sudo tee /etc/udev/rules.d/60-scheduler.rules <<EOF
ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/rotational}=="0", ATTR{queue/scheduler}="none"
ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/rotational}=="1", ATTR{queue/scheduler}="deadline"
EOF

Read-Ahead

# Increase read-ahead for sequential reads
sudo blockdev --setra 8192 /dev/sdb

# Make permanent
echo 'ACTION=="add|change", KERNEL=="sdb", ATTR{bdi/read_ahead_kb}="8192"' | \
  sudo tee /etc/udev/rules.d/60-readahead.rules

Direct I/O

export:
  direct_io: true  # Bypass page cache

RAID Configuration

# RAID 0 for maximum speed (no redundancy)
sudo mdadm --create /dev/md0 --level=0 --raid-devices=4 \
  /dev/sd[bcde]

# RAID 10 for speed + redundancy
sudo mdadm --create /dev/md0 --level=10 --raid-devices=4 \
  /dev/sd[bcde]

# Format and mount
sudo mkfs.xfs -f /dev/md0
sudo mount -o noatime /dev/md0 /exports

Memory Management

Increase Memory

daemon:
  max_memory: "16GB"  # Limit daemon memory usage

Page Cache

# Drop cache before large exports
sudo sync
sudo echo 3 | sudo tee /proc/sys/vm/drop_caches

# Adjust swappiness (prefer RAM over swap)
sudo sysctl -w vm.swappiness=10

# Make permanent
echo "vm.swappiness=10" | sudo tee -a /etc/sysctl.conf

Huge Pages

# Enable transparent huge pages
echo always | sudo tee /sys/kernel/mm/transparent_hugepage/enabled

# Or configure static huge pages
sudo sysctl -w vm.nr_hugepages=1024
echo "vm.nr_hugepages=1024" | sudo tee -a /etc/sysctl.conf

CPU Optimization

CPU Affinity

# Pin daemon to specific CPUs
sudo systemctl edit hypervisord

[Service]
CPUAffinity=0-7  # Use CPUs 0-7

CPU Governor

# Use performance governor
sudo cpupower frequency-set -g performance

# Or on-demand
sudo cpupower frequency-set -g ondemand

NUMA Optimization

# Check NUMA layout
numactl --hardware

# Run on specific NUMA node
numactl --cpunodebind=0 --membind=0 hypervisord

Goroutine Pool

daemon:
  max_goroutines: 10000  # Limit concurrent goroutines
  goroutine_pool_size: 1000

Database Performance

SQLite Optimization

database:
  type: "sqlite"
  path: "/var/lib/hypersdk/hypersdk.db"
  pragmas:
    journal_mode: "WAL"  # Write-Ahead Logging
    synchronous: "NORMAL"
    cache_size: -64000  # 64MB cache
    mmap_size: 268435456  # 256MB mmap
    page_size: 4096
    temp_store: "MEMORY"

PostgreSQL Optimization

database:
  type: "postgres"
  host: "localhost"
  port: 5432
  database: "hypersdk"
  max_connections: 100
  max_idle_connections: 10
  connection_lifetime: "1h"

-- PostgreSQL tuning
ALTER SYSTEM SET shared_buffers = '4GB';
ALTER SYSTEM SET effective_cache_size = '12GB';
ALTER SYSTEM SET maintenance_work_mem = '1GB';
ALTER SYSTEM SET checkpoint_completion_target = 0.9;
ALTER SYSTEM SET wal_buffers = '16MB';
ALTER SYSTEM SET default_statistics_target = 100;
ALTER SYSTEM SET random_page_cost = 1.1;
ALTER SYSTEM SET effective_io_concurrency = 200;
ALTER SYSTEM SET work_mem = '10MB';
ALTER SYSTEM SET min_wal_size = '1GB';
ALTER SYSTEM SET max_wal_size = '4GB';

-- Restart PostgreSQL
SELECT pg_reload_conf();

Benchmarking

Export Benchmark

#!/bin/bash
# benchmark-export.sh

VM="/datacenter/vm/benchmark-vm"
ITERATIONS=5

echo "Running export benchmark..."
for i in $(seq 1 $ITERATIONS); do
  echo "Iteration $i..."

  START=$(date +%s)

  ./hyperexport -vm "$VM" \
    -output /tmp/benchmark-$i \
    -format ova \
    -compress \
    -verify

  END=$(date +%s)
  DURATION=$((END - START))

  SIZE=$(du -sh /tmp/benchmark-$i | cut -f1)

  echo "Duration: ${DURATION}s, Size: $SIZE"

  rm -rf /tmp/benchmark-$i
done

Network Benchmark

# iperf3 test
iperf3 -c vcenter.example.com -t 60 -P 4

# Test vSphere network
time govc vm.info -vm.path="$VM" > /dev/null

Disk Benchmark

# Sequential write
dd if=/dev/zero of=/exports/test bs=1M count=10240 oflag=direct

# Sequential read
dd if=/exports/test of=/dev/null bs=1M iflag=direct

# fio comprehensive test
fio --name=sequential-write --ioengine=libaio --direct=1 --bs=1M \
  --iodepth=64 --rw=write --size=10G --filename=/exports/test

fio --name=sequential-read --ioengine=libaio --direct=1 --bs=1M \
  --iodepth=64 --rw=read --size=10G --filename=/exports/test

fio --name=random-read --ioengine=libaio --direct=1 --bs=4k \
  --iodepth=256 --rw=randread --size=10G --filename=/exports/test

API Benchmark

# Apache Bench
ab -n 1000 -c 10 http://localhost:8080/health

# wrk
wrk -t 4 -c 100 -d 30s http://localhost:8080/status

# Custom benchmark
for i in {1..100}; do
  time curl -s http://localhost:8080/health > /dev/null
done

Performance Monitoring

System Monitoring

# Real-time system stats
htop
iotop
nethogs
nmon

# CPU usage
mpstat 1

# Disk I/O
iostat -xz 1

# Network
iftop -i eth0

Application Monitoring

# Daemon metrics
curl http://localhost:8080/metrics

# Profiling
curl http://localhost:6060/debug/pprof/profile?seconds=30 > cpu.prof
go tool pprof cpu.prof

Prometheus Metrics

metrics:
  enabled: true
  prometheus_enabled: true
  prometheus_path: "/metrics"

# Scrape metrics
curl http://localhost:8080/metrics

# Example metrics:
# hypersdk_export_duration_seconds
# hypersdk_export_bytes_total
# hypersdk_jobs_active
# hypersdk_connection_pool_size

Optimization Checklist

• Parallel downloads configured (8+)
• Connection pooling enabled
• TCP buffer sizes increased
• MTU optimized for network
• Filesystem tuned (XFS with noatime)
• I/O scheduler optimized
• Read-ahead configured
• Page cache settings tuned
• CPU governor set to performance
• Database optimized (WAL mode for SQLite)
• Monitoring enabled
• Benchmarks run and baselined

Performance Targets

Export Performance

• Small VM (< 50GB): 5-10 minutes
• Medium VM (50-200GB): 15-30 minutes
• Large VM (200GB-1TB): 1-3 hours

API Performance

• Health check: < 10ms
• Job submission: < 100ms
• Job query: < 50ms
• WebSocket latency: < 100ms

Resource Usage

• Memory: < 2GB for daemon
• CPU: < 50% during exports
• Disk I/O: Limited by network bandwidth
• Network: Saturate available bandwidth

See Also

• Configuration Reference
• Troubleshooting Guide
• System Requirements

This site is open source. Improve this page.