Engineering Cheat Sheet
Essential latency, throughput, and performance numbers every software engineer should know
Latency
Time measurements for various operations
CPU L1 cache access
CPU1ns
Best case scenario
CPU L2 cache access
CPU4ns
L2 cache hit
CPU L3 cache access
CPU12ns
L3 cache hit
Memory access (RAM)
Memory100ns
Main memory access
SSD random read
Storage10μs
NVMe SSD
SSD random write
Storage20μs
NVMe SSD
HDD random read
Storage10ms
Traditional hard drive
Network round trip (same datacenter)
Network0.5ms
Local network
Network round trip (cross-country)
Network50ms
US coast to coast
Network round trip (international)
Network150ms
US to Europe
Database query (simple)
Database1ms
Indexed lookup
Database query (complex)
Database100ms
Full table scan
Throughput
Operations per second and data transfer rates
CPU instructions per second
CPU3.5 billion
Modern CPU (3.5 GHz)
Memory bandwidth
Memory25 GB/s
DDR4-3200
SSD sequential read
Storage3.5 GB/s
NVMe SSD
SSD sequential write
Storage3.0 GB/s
NVMe SSD
HDD sequential read
Storage150 MB/s
7200 RPM HDD
Gigabit Ethernet
Network125 MB/s
1 Gbps network
10 Gigabit Ethernet
Network1.25 GB/s
10 Gbps network
Database transactions per second
Database10,000
PostgreSQL on SSD
Redis operations per second
Database100,000
In-memory operations
Storage & IOPS
Input/Output operations per second and storage limits
SSD random read IOPS
Storage500,000
NVMe SSD
SSD random write IOPS
Storage400,000
NVMe SSD
HDD random read IOPS
Storage100
7200 RPM HDD
HDD random write IOPS
Storage80
7200 RPM HDD
Single file size limit
Filesystem16 TB
ext4 filesystem
Directory entries limit
Filesystem4 billion
ext4 filesystem
Inode limit
Filesystem4 billion
ext4 filesystem
Database Limits
Common database performance and capacity limits
PostgreSQL max connections
PostgreSQL100-500
Default configuration
MySQL max connections
MySQL151
Default configuration
Redis max memory
RedisUnlimited
Depends on available RAM
MongoDB document size limit
MongoDB16 MB
BSON document limit
Elasticsearch shard size
Elasticsearch50 GB
Recommended maximum
Cassandra partition size
Cassandra100 MB
Recommended maximum
Network Limits
Network performance and connection limits
TCP connection limit
TCP65,535
Per port (16-bit port number)
UDP packet size
UDP65,507 bytes
Maximum payload size
HTTP/2 concurrent streams
HTTP100
Default browser limit
WebSocket message size
WebSocket2 GB
Theoretical maximum
DNS query timeout
DNS5 seconds
Default timeout
Performance Targets
Recommended performance targets for different use cases
Web page load time
Web< 3 seconds
Good user experience
API response time
API< 200ms
Interactive applications
Database query time
Database< 100ms
Real-time applications
Cache hit ratio
Caching> 95%
Effective caching
CPU utilization
CPU< 80%
Sustained performance
Memory utilization
Memory< 85%
Avoid swapping
Availability & SLAs
Service level agreements and availability targets
99.9% availability
SLA8.76 hours/year
Downtime per year
99.99% availability
SLA52.56 minutes/year
Downtime per year
99.999% availability
SLA5.26 minutes/year
Downtime per year
Mean Time to Recovery (MTTR)
SLA< 1 hour
Incident response target
Mean Time Between Failures (MTBF)
SLA> 30 days
System reliability
Recovery Point Objective (RPO)
SLA< 15 minutes
Data loss tolerance
Recovery Time Objective (RTO)
SLA< 4 hours
Service restoration
Scalability Limits
System capacity and scaling thresholds
Vertical scaling limit
Scaling~1000 cores
Single machine maximum
Horizontal scaling limit
Scaling~10,000 nodes
Distributed systems
Database connection limit
Scaling~1000 connections
Per database instance
Load balancer capacity
Scaling~1M requests/sec
Modern load balancers
CDN edge locations
Scaling~200+ locations
Global content delivery
Kubernetes cluster limit
Scaling5,000 nodes
Per cluster maximum
Security & Compliance
Security thresholds and compliance requirements
Password entropy
Security> 60 bits
Minimum security
Session timeout
Security15-30 minutes
Idle session limit
Rate limiting
Security100 req/min
Per user/IP
Encryption key rotation
Security90 days
Recommended frequency
Vulnerability scan frequency
SecurityWeekly
Security monitoring
Data retention period
Security7 years
Compliance requirement
Cost Optimization
Resource cost thresholds and optimization targets
CPU cost per hour
Cost$0.10-0.50
Cloud compute pricing
Memory cost per GB
Cost$0.05-0.20
Cloud memory pricing
Storage cost per GB
Cost$0.02-0.10
Cloud storage pricing
Network cost per GB
Cost$0.01-0.05
Data transfer pricing
Database cost per hour
Cost$0.50-5.00
Managed database pricing
CDN cost per GB
Cost$0.01-0.02
Content delivery pricing
Monitoring & Alerting
System monitoring thresholds and alerting rules
CPU alert threshold
Monitoring> 80%
5-minute average
Memory alert threshold
Monitoring> 85%
5-minute average
Disk space alert
Monitoring> 90%
Storage utilization
Error rate threshold
Monitoring> 1%
5-minute error rate
Response time alert
Monitoring> 1 second
95th percentile
Queue depth alert
Monitoring> 1000 messages
Message queue depth
Data Consistency
Consistency models and data integrity metrics
ACID compliance
Consistency100%
Transactional databases
Eventual consistency delay
Consistency< 1 second
Distributed systems
Data replication lag
Consistency< 100ms
Cross-region replication
Consistency check frequency
ConsistencyEvery 5 minutes
Data integrity validation
Backup frequency
ConsistencyEvery 6 hours
Point-in-time recovery
Data validation coverage
Consistency> 95%
Input validation
SLA/SLO Metrics
Service Level Agreement and Objective metrics for system reliability
Error Budget
SLO1% per month
SLO error allowance
Burn Rate
SLO2x normal rate
Error budget consumption
SLO Measurement Window
SLO30 days
Rolling window period
Alerting Threshold
SLO50% error budget
When to page on-call
SLA Credit
SLA5-10% monthly
Service credit for downtime
SLO Target
SLO99.9%
Typical service target
Distributed Systems
Metrics for distributed system design and operations
CAP Theorem trade-offs
DistributedPick 2 of 3
Consistency, Availability, Partition tolerance
Consensus algorithm latency
Distributed10-100ms
Raft/Paxos round trip
Leader election timeout
Distributed150-300ms
Raft election timeout
Heartbeat interval
Distributed50-100ms
Node health checks
Split-brain detection
Distributed< 1 second
Network partition detection
Gossip protocol interval
Distributed1-10 seconds
Eventual consistency
Microservices
Microservices architecture metrics and patterns
Service mesh latency
Microservices+1-2ms
Istio/Linkerd overhead
Circuit breaker threshold
Microservices50% failure rate
Open circuit condition
Circuit breaker timeout
Microservices60 seconds
Half-open state duration
Bulkhead thread pool
Microservices10-50 threads
Isolated resource pools
Service discovery TTL
Microservices30 seconds
DNS/Consul cache
API versioning deprecation
Microservices6-12 months
Backward compatibility
Observability
Metrics, logs, and tracing for system observability
Metrics cardinality
Observability< 10,000 series
Per service maximum
Log retention period
Observability30-90 days
Hot storage duration
Trace sampling rate
Observability1-10%
Production tracing
Metric collection interval
Observability15-60 seconds
Prometheus scrape
Alert evaluation interval
Observability15 seconds
Prometheus rule evaluation
Dashboard refresh rate
Observability30 seconds
Real-time monitoring
Cost Benchmarks
Resource cost benchmarks for capacity planning
AWS EC2 (t3.medium)
Compute$30/month
2 vCPU, 4GB RAM
AWS RDS (db.t3.medium)
Database$50/month
2 vCPU, 4GB RAM
AWS S3 storage
Storage$0.023/GB
Standard tier
AWS CloudFront
CDN$0.085/GB
Data transfer out
Redis (cache.r6g.large)
Cache$100/month
13GB memory
Elasticsearch (r6g.large)
Search$120/month
3 nodes, 16GB RAM
Alerting Thresholds
Key metrics to monitor for system health
CPU alert threshold
Alerting> 80%
5 minute average
Memory alert threshold
Alerting> 85%
5 minute average
Disk space alert
Alerting> 90%
Available space
Error rate alert
Alerting> 1%
5 minute window
Response time alert
Alerting> 1s
P95 over 5 minutes
Queue depth alert
Alerting> 1000
Message queue length
Database connections
Alerting> 80%
Max connections
Security & Compliance
Security thresholds and compliance requirements
Password entropy
Security> 60 bits
Minimum security
Session timeout
Security15-30 minutes
Idle timeout
Rate limiting
Security100 req/min
Per IP address
API rate limit
Security1000 req/hour
Per API key
Failed login attempts
Security5 attempts
Before lockout
Data retention
Compliance7 years
SOX compliance
Backup frequency
ComplianceDaily
Critical data