Bandwidth
Bandwidth is all about max data transfer across a network connection, measured in bps or bytes per second. It determines data flow between client and proxy server per time unit. Without knowing bandwidth, you can't optimize throughput. Crappy network capacity? You're stuck with bottlenecks that drag down connection speed and cut your request success rates. You'll also be limited on how many simultaneous proxy sessions get to run smoothly.
Quick Facts
- Also known as
- Network capacity, data transfer limit, throughput
- IP source
- Applies across residential, datacenter, and mobile proxy pools
- Detection risk
- Low , high-bandwidth usage patterns can trigger rate limits on target sites if unmanaged
- Typical use
- Web scraping, streaming data collection, large-scale ad verification, bulk SEO monitoring
- Price range
- $0.27–$0.79/GB, with proxy server bandwidth allocation priced as low as $0.27/GB at scale
How a bandwidth works
Requests routed via proxy servers? Data starts at the origin, hops through the proxy node, then back to the client. Each of these jumps eats up some network capacity. Proxy server bandwidth allocation decides how much network juice a session or user gets, and it affects connection speed and data transfer limits. Geonode meters bandwidth in gigabytes so you only pay for what you actually use, not for connections sitting idle. TRADEOFFS: Bandwidth scaling isn't free. Residential nodes have wild uplink speeds, typically 5–50 Mbps per node, versus 1–10 Gbps in datacenter gear. This inconsistency means a 2 MB HTML page might load in 400ms on one node, yet take 3s on another. Geonode will avoid the slow ones, but higher latency variance (50–500ms) is a given with residential IPs compared to datacenter IPs (5–20ms). The catch? Datacenter IPs face aggressive blocking from Cloudflare and Akamai compared to residential ones. For high-volume scraping, you're looking at 5–10 Mbps max on a single residential session; need more? Split the load across several sessions. Bandwidth caps matter: $0.27/GB means a 1 TB/month budget gives roughly 3.7 GB/hour nonstop. Fine for most mid-scale jobs, but if you're crawling a huge site catalog (100M+ pages at ~50 KB each), that's a ~5 TB scenario, plan ahead.
Bandwidth vs. Latency
Bandwidth's about total data volume moved over a connection in a time unit (like pipe width), whereas latency is the lag between sending a request and getting the first response byte. High bandwidth paired with high latency? Still slow for things that need quick responses, like real-time price tracking. Low bandwidth and low latency feels better but bottlenecks big data shifts like mass HTML scrapes across Geonode's 2.5M+ residential IP pool.
Why this is different
Advantages
- 10 Mbps bandwidth per session pushes out ~1.25 MB/s sustained throughput. It's enough to grab an average 50 KB product page in under 50ms at the network layer. Fast enough.
- Requests spread across Geonode's 2.5M+ residential IP pool lower per-IP request pressure from ~50 req/s on a single IP to virtually unlimited at the pool level. You push rate-limit triggers way above what a single session would normally hit.
- Residential IPs mimic normal browser session traffic patterns. Anti-bot systems at sites like Cloudflare and Akamai see it as organic, not crawler traffic. That's how you stay under the radar.
- Per-GB billing means you're only paying for the bytes you actually use. A 100-request test pulling 200 MB costs $0.05 at $0.27/GB. No session fees burning cash on idle connections.
Tradeoffs
- Residential IPs cost 3–5x more than datacenter IPs per GB. Geonode's residential bandwidth starts at $0.27/GB at scale. It's higher than typical datacenter pricing under $0.10/GB because this infrastructure runs on actual consumer devices with real uplink costs.
- Residential nodes offer variable uplink speeds (5,50 Mbps) compared to datacenter hardware (1,10 Gbps). This means latency varies from 50,500ms per hop on residential versus 5,20ms on datacenter. For things like real-time financial feeds, where a 100ms delay messes up timestamps, you'll want datacenter options.
- Running a large crawl job eats up bandwidth budgets quicker than you might expect. Scraping 1M pages at an average response size of 150 KB uses ~150 GB. At $0.79/GB (entry tier), that costs $118.50 before you count retries on failed requests.
- High concurrent session counts multiply bandwidth consumption non-linearly. If you have 500 sessions pulling 1 req/s each against a 200 KB average page size, you'll need 800 Mbps of sustained pool capacity. This works with a 2.5M+ IP pool, but you need to calculate your size, not just hope.
Examples in practice
Real-world deployments of Bandwidth , where it works and where alternatives win.
E-Commerce Price Scraping
A retailer watching 50,000 Amazon ASINs at one request per ASIN every 15 minutes produces roughly 55 req/s sustained. An average response of 120 KB per page amounts to 6.6 MB/s, or about 53 Mbps. This is just to maintain the refresh cycle. Operations of this size typically eat through 50,200 GB per day. You'll need IP rotation across Geonode's residential pool to dodge Amazon's per-IP throttle, which kicks in around 30 req/min.
Financial Data Collection
Hedge funds grabbing tick data from Yahoo Finance, Bloomberg's public endpoints, or exchange APIs need bandwidth that doesn't flinch under load. A feed collecting 500 symbols at 1-second intervals, each response averaging 2 KB, runs at only 1 Mbps. Bandwidth isn't the issue. Latency is: a residential node at 300ms round-trip introduces jitter that throws off timestamps across feeds. For this, stick with low-latency datacenter IPs at 5,20ms, even if they carry a higher detection risk.
SERP Rank Tracking
An SEO platform monitoring 10,000 keywords across Google's 195+ country indexes fires queries from local IPs in each target region to get the real deal on localized results. At 1 query per keyword per day per country, that's 1.95M requests/day, or around 22 req/s. Google's HTML responses average 80 KB, keeping sustained bandwidth demand at about 14 Mbps. The real issue is IP freshness: reuse the same residential IP for over 10,15 Google queries per hour, and you're looking at CAPTCHAs. So, pool size matters more than raw throughput.
Ad Verification Campaigns
Ad fraud detection services like DoubleVerify or White Ops verify ad placements by loading publisher pages as if they're real users using real IPs. A campaign checking 10,000 publisher domains twice every day, with an average page weight of 1.5 MB including ad assets, uses 30 GB per day. If you don't allocate enough bandwidth at the proxy layer, you get timeouts before ad tags fire, leading to false negatives in the report. It'll look like the ad wasn't verified, but really, it just loaded too slowly to capture.
Large-Scale Web Crawling
Common Crawl's monthly web archive consumes petabytes of data across hundreds of millions of domains. A comparable enterprise crawl aiming at 100M pages with a 50 KB average response size needs 5 TB of bandwidth. At Geonode's $0.27/GB scale pricing, you’re looking at $1,350 in bandwidth cost. It's the main factor in crawl budgeting, not compute. Spread the job across 1,000 concurrent sessions at 5 Mbps each to get 5 Gbps aggregate throughput. This knocks out a 5 TB crawl in about 2.2 hours of wall-clock time.
Streaming Content Geo-Testing
Netflix and Disney+ hand out different content libraries per country. QA teams checking regional catalog availability need to load actual stream manifests, not just the homepage, from local IPs. A manifest request for a 4K title returns about 500 KB of playlist data. Testing 200 titles across 50 countries means 10,000 requests at 500 KB each: 5 GB per test run. At $0.27/GB, that’s $1.35 per full catalog sweep. So, daily automated geo-checks aren't crazy expensive at scale.
Common misconceptions
Common myths about Bandwidth , and what is actually true.
| Myth | Reality |
|---|---|
Bandwidth and speed are the same thing. | Bandwidth is capacity; latency and throughput determine how fast a given request actually feels. |
More bandwidth always means faster scraping. | Scraping is often limited by target rate limits and latency, not by your link capacity. |
Per-GB proxy billing counts only useful data. | It typically counts all transferred bytes, including headers, retries, and assets, so trimming requests saves money. |
Need Bandwidths?
2.5M+ residential IPs, 195+ countries, from $0.27/GB.
