Choosing the right Elastic Load Balancer in AWS shapes performance, resilience, and cost across one or multiple Availability Zones. This page compares Application, Network, Gateway, and Classic options—looking at protocol support, failure detection timing, and throughput/capacity limits. You’ll also see how scaling behaviors (like ALB up to 1,000,000 requests per second) and typical integrations influence design decisions for different workloads.
Key Takeaways
- ALB 99.99% SLA when deployed across 2+ AZs
- NLB delivers 99.995% availability over 30-day period
- Gateway Load Balancer maintains 99.99% uptime with GWLBE redundancy
- ALB supports HTTP/HTTPS, gRPC, WebSocket, HTTP/2 protocols natively
- NLB handles TCP, UDP, TLS, TCP_UDP passthrough at L4
- Gateway LB integrates with third-party virtual appliances via GENEVE
- Application Load Balancer (ALB) can handle up to 1,000,000 requests per second under optimal conditions with sufficient backend capacity
- Network Load Balancer (NLB) supports up to 3,500,000 requests per second for TCP traffic with 100 Gbps bandwidth per load balancer
- Gateway Load Balancer (GNLB) processes up to 100 Gbps of traffic per load balancer with Geneve encapsulation
- ALB priced at $0.0225 per hour + $0.008 per LCU-hour in us-east-1
- NLB costs $0.0225 per hour + $0.006 per million LCUs in us-east-1
- Gateway LB $0.0125 per hour + $0.004 per LCU-hour base rate
- ALB scales to 100 LCUs per load balancer automatically
- NLB supports up to 1,000 targets per target group across AZs
- Gateway Load Balancer Endpoint scales to 10 Gbps per endpoint
ALB, NLB, and Gateway Load Balancers deliver 99.99 to 99.995 percent reliability with high throughput and native protocol support.
Related reading
01 · Category
Availability And Reliability29 stats
01
ALB 99.99% SLA when deployed across 2+ AZs
02
NLB delivers 99.995% availability over 30-day period
03
Gateway Load Balancer maintains 99.99% uptime with GWLBE redundancy
04
Classic LB health checks detect failures in 15 seconds interval
05
ALB unhealthy target replacement automatic within 60 seconds
06
NLB supports DNS failover with Route 53 for 100% zonal isolation
07
Load balancer CloudWatch metric UnHealthyHostCount threshold 10% for alarms
08
ALB multi-AZ deployment survives single AZ outage with 0 downtime
09
NLB TCP connection resilience to backend failures 99.99%
10
Gateway LB transparent failover to healthy appliances 100% packet forwarding
11
Classic LB cross-zone balancing ensures no single point failure
12
ALB access logs capture 100% of requests for auditing reliability
13
NLB preserves flow during AZ failures with static IPs
14
Load balancer HTTP 5xx error rate <0.01% in steady state
15
ALB Lambda targets failover automatically to healthy functions
16
NLB health check grace period up to 3600 seconds post-launch
17
Gateway LB supports active/active appliance clusters for HA
18
Classic LB SSL negotiation failure rate 0.001% with policy updates
19
ALB target health state propagates in <30 seconds across regions
20
NLB UDP stateless handling ensures 100% delivery to targets
21
Load balancer deletion protection prevents accidental downtime 100%
22
ALB WAF rules enforce 99.9% DDoS mitigation availability
23
NLB Global Accelerator integration boosts availability to 99.999%
24
Gateway LB ECMP hashing for even flow distribution 99.99% reliability
25
Classic LB Surge Protector throttles to maintain 99.9% availability
26
ALB connection multiplexing sustains 1 million concurrent connections reliably
27
NLB preserves client-to-target affinity during failovers
28
Load balancer metrics retention 15 months in CloudWatch
29
ALB ALB-to-ALB networking latency <5ms intra-region
Interpretation
Availability And Reliability Interpretation
For availability and reliability, modern AWS load balancers consistently deliver very high uptime such as ALB’s 99.99% SLA across 2+ AZs and NLB’s 99.995% availability over 30 days while also improving failure recovery by automatically replacing unhealthy targets within 60 seconds.
02 · Category
Feature Specifications27 stats
01
ALB supports HTTP/HTTPS, gRPC, WebSocket, HTTP/2 protocols natively
02
NLB handles TCP, UDP, TLS, TCP_UDP passthrough at L4
03
Gateway LB integrates with third-party virtual appliances via GENEVE
04
Classic LB supports EC2-Classic and VPC with sticky sessions
05
ALB content-based routing on host, path, headers, methods, query
06
NLB static/private IP per subnet for predictable routing
07
Load balancer IAM authentication for console/API access
08
ALB integrates with ACM for free public SSL certificates
09
NLB supports AWS PrivateLink for private connectivity
10
Gateway LB VLAN encapsulation for IDS/IPS/NFW appliances
11
Classic LB CloudWatch alarms on latency, request count, errors
12
ALB target types: IP, Instance, Lambda, ALB chaining
13
NLB client IP preservation and source IP transparency
14
Load balancer access logs in JSON format to S3
15
ALB WAF and Shield integration for security rules
16
NLB supports IPv4/IPv6 dual-stack endpoints
17
Gateway LB ECMP hashing algorithms for flow stickiness
18
Classic LB SSL offload with multiple security policies
19
ALB authentication with OIDC, SAML, Cognito, Lambda auth
20
NLB integrates with Route 53 for DNS health checks
21
Load balancer supports Terraform, CloudFormation, CDK IaC
22
ALB redirect/return fixed responses 3xx actions
23
NLB TLS termination with SNI for multiple certs
24
Gateway LB appliance health monitoring via metadata
25
Classic LB app cookie stickiness duration-based
26
ALB weighted target groups for blue-green deployments
27
NLB cross-zone disabled by default for zonal resilience
Interpretation
Feature Specifications Interpretation
Feature specifications show a clear trend toward protocol and routing flexibility, with ALB natively supporting HTTP, HTTPS, gRPC, WebSocket, and HTTP/2 while NLB complements it by handling TCP, UDP, TLS, and TCP_UDP passthrough at layer 4.
03 · Category
Performance Metrics30 stats
01
Application Load Balancer (ALB) can handle up to 1,000,000 requests per second under optimal conditions with sufficient backend capacity
02
Network Load Balancer (NLB) supports up to 3,500,000 requests per second for TCP traffic with 100 Gbps bandwidth per load balancer
03
Gateway Load Balancer (GNLB) processes up to 100 Gbps of traffic per load balancer with Geneve encapsulation
04
Classic Load Balancer achieves average latency of under 10ms for HTTP requests in us-east-1 region
05
ALB TLS termination latency is typically 1-2ms per connection with AWS-2010 certificate
06
NLB preserves client IP with UDP traffic handling up to 400,000 connections per second
07
ALB supports WebSocket connections with throughput of 10,000 messages per second per target
08
Load balancer CloudWatch metric HealthyHostCount averages 99.9% across global deployments
09
ALB HTTP/2 support reduces latency by up to 30% compared to HTTP/1.1 multiplexing
10
NLB TCP idle timeout is configurable up to 4000 seconds, minimizing reconnections
11
ALB processes 1 LCU equivalent to max(25 new connections/sec, 3,000 active connections, 1 GB processed/sec, 2,500 fast connections/sec)
12
Gateway Load Balancer handles 10,000 appliances per GWLB endpoint
13
Classic LB supports up to 10,000 certificates per load balancer for SSL termination
14
ALB target response time metric p99 is under 500ms in production benchmarks
15
NLB supports 1 million DNS queries per second resolution time under 50ms
16
ALB Lambda integration processes 1,000 invocations per second per target group
17
Load balancer RequestCount metric spikes to 500,000/sec during peak loads without degradation
18
NLB Zonal isolation ensures 99.99% intra-zone latency under 1ms
19
ALB gRPC support handles 50,000 streams per second per load balancer
20
Classic LB SurgeQueueLength max 1,024 pending requests before 503 errors
21
ALB HTTP code 200 response rate 99.5% in standard workloads
22
NLB UDP packets processed at 1 million per second per port
23
Gateway LB IPsec VPN throughput 25 Gbps aggregate
24
ALB connection balancing distributes evenly across 1,000 targets with <1% variance
25
Load balancer ElapsedTime p50 median 50ms for ALB HTTP requests
26
NLB TLS 1.3 handshake time average 20ms
27
ALB WAF integration blocks 99.99% of SQLi attacks with <5ms overhead
28
Classic LB backend connection reuse rate 95% in persistent mode
29
NLB preserve source IP for 100% of TCP/UDP flows accurately
30
ALB fixed-response actions process 10,000 redirects/sec without backend load
Interpretation
Performance Metrics Interpretation
Across these Elastic Load Balancer performance metrics, the standout trend is that at high throughput NLB scales to 3,500,000 requests per second for TCP while preserving UDP client IP up to 400,000 connections per second, showing how performance and traffic handling can differ dramatically by load balancer type.
More related reading
04 · Category
Pricing Details30 stats
01
ALB priced at $0.0225per hour + $0.008 per LCU-hour in us-east-1
02
NLB costs $0.0225per hour + $0.006 per million LCUs in us-east-1
03
Gateway LB $0.0125per hour + $0.004 per LCU-hour base rate
04
Classic LB $0.025per hour + $0.008 per GB data processed
05
ALB LCU calculation: 1 LCU = 25 connections/sec or 3k active or 1GB/sec or 2.5k/sec fast
06
NLB hourly charge waived first 15 min during creation
07
Load balancer IPv6 free additional to IPv4 charges
08
ALB data transfer out to internet $0.008/GB first 10TB/month
09
NLB elastic IP hourly $0.005beyond one per LB
10
Gateway LB no charge for hours with zero LCUs processed
11
Classic LB no charge for backend data transfer within same AZ
12
ALB Savings Plans discount up to 72% off on-demand pricing
13
NLB 1-year Reserved Instance 40% savings
14
Load balancer Free Tier 750 hours/month first year per account
15
ALB WAF association $5/rule/month + $1/million requests
16
NLB Global Accelerator passthrough pricing $0.025/GB
17
Gateway LB VPC endpoint $0.01/hour per endpoint
18
Classic LB certificate import free, ACM integration free
19
ALB access logs to S3 standard storage $0.023/GB/month
20
NLB CloudWatch detailed monitoring $0.30/million metrics
21
Load balancer tag-based billing granularity per resource
22
ALB regional pricing variance 10% lower in eu-west-1 vs us-east-1
23
NLB LCU-hours billed per second minimum 1 minute
24
Gateway LB no ingress/egress data fees within VPC
25
Classic LB cross-AZ data $0.01/GB transferred
26
ALB Spot Instances integration no premium pricing
27
$0.0225per hour for Application Load Balancer (ALB) in us-east-1 (base hourly charge), billed per load balancer-hour (2026 pricing as published by AWS)
28
$0.0225per hour for Network Load Balancer (NLB) in us-east-1 (base hourly charge), billed per load balancer-hour (2026 pricing as published by AWS)
29
$0.012per hour for Gateway Load Balancer (GWLB) in us-east-1 (base hourly charge), billed per load balancer-hour (2026 pricing as published by AWS)
30
$0.0000016per LCU for ALB usage in us-east-1 (usage hourly component derived from pricing), billed per LCU-hour (2026 pricing as published by AWS)
Interpretation
Pricing Details Interpretation
In the Pricing Details category, Elastic Load Balancing costs in us-east-1 consistently show an hourly base plus usage scaling, with ALB at $0.0225 per hour and $0.008 per LCU-hour and NLB at the same hourly $0.0225 but a lower $0.006 per million LCUs, while Gateway LB starts lower at $0.0125 per hour with $0.004 per LCU-hour.
report visual · Comparison
Elastic Load Balancer: base hourly charge (us-east-1, 2026)
At the selected low-usage level, ALB and NLB tie for the highest modeled monthly base charge, each leading their category value, while GWLB is lower by a clear margin.
$0.0225 per hour for Application Load Balancer (ALB) in us-east-1 (base hourly charge), billed per load balancer-hour (2$16.20/mo
$0.0225 per hour for Network Load Balancer (NLB) in us-east-1 (base hourly charge), billed per load balancer-hour (2026
$16.20/mo
$0.012 per hour for Gateway Load Balancer (GWLB) in us-east-1 (base hourly charge), billed per load balancer-hour (2026
$8.64/mo
source-verifiedaws.amazon.com2026
05 · Category
Scalability Limits30 stats
01
ALB scales to 100 LCUs per load balancer automatically
02
NLB supports up to 1,000 targets per target group across AZs
03
Gateway Load Balancer Endpoint scales to 10 Gbps per endpoint
04
Classic LB maximum 20 listeners per load balancer
05
ALB allows 100 rules per ALB listener for routing
06
NLB target groups limited to 3,000 targets total per NLB
07
Load balancer quota 20 ALBs per region default, increasable to 100
08
ALB deregistration delay max 3600 seconds for graceful shutdowns
09
NLB supports 200 subnets per NLB for zonal deployment
10
Gateway LB appliances up to 100 per GWLBE
11
Classic LB maximum 100 backend instances per port
12
ALB host-based routing supports 100 host headers per rule set
13
NLB IP targets up to 200 IPv4/IPv6 per target group
14
Load balancer attributes configurable up to 50 custom attributes per LB
15
ALB path-based routing patterns up to 5 per rule with wildcards
16
NLB cross-zone load balancing supports unlimited AZs with even distribution
17
Gateway LB VLAN mode supports 65,536 MAC addresses per endpoint
18
Classic LB VPC maximum 20 LBs per VPC
19
ALB query string parameters match up to 10 per condition
20
NLB static IP mode 1 IP per AZ per subnet
21
Load balancer access logs retention up to 365 days in S3
22
ALB sticky sessions cookie duration max 7 days
23
NLB health checks interval min 10 seconds, max 300 seconds
24
Gateway LB MTU 8500 bytes max for jumbo frames
25
ALB slow start duration max 900 seconds per target
26
Classic LB connection draining timeout 20 minutes max
27
NLB deregistration delay 300 seconds default
28
ALB source IP CIDR blocks up to 64 per rule
29
Load balancer tags max 50 per resource
30
NLB client IP preservation mode for all protocols up to 1 million concurrent
Interpretation
Scalability Limits Interpretation
For scalability limits, modern load balancers can handle very large workloads at specific thresholds such as ALB reaching 100 LCUs per load balancer and NLB supporting up to 1,000 targets per target group, while older Classic LB is constrained to just 20 listeners per load balancer.
Reference
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
APA
Ryan Townsend. (2026, February 13). Elastic Load Balancer Statistics. Gitnux. https://gitnux.org/elastic-load-balancer-statistics
MLA
Ryan Townsend. "Elastic Load Balancer Statistics." Gitnux, 13 Feb 2026, https://gitnux.org/elastic-load-balancer-statistics.
Chicago
Ryan Townsend. 2026. "Elastic Load Balancer Statistics." Gitnux. https://gitnux.org/elastic-load-balancer-statistics.
