Port Speed 1 Gbps: Metered vs. Unmetered Bandwidth Explained for Dedicated Servers

Port speed on a dedicated server defines the maximum rate at which data can be transmitted between your server and the upstream network. At AlexHost, the default allocation for Dedicated Servers is 300 Mbps, included in the base plan. For workloads demanding higher throughput, a 1 Gbps upgrade is available in three distinct configurations: Shared, Dedicated Metered, and Dedicated Unmetered — each with fundamentally different performance guarantees, bandwidth caps, and cost structures.

Choosing the wrong tier is one of the most common and costly mistakes in server provisioning. A shared port may be perfectly adequate for a low-traffic application, while a high-volume media platform or backup infrastructure can exhaust a metered allocation within days. This guide breaks down every technical and operational dimension you need to make the right call.

What "Port Speed" Actually Means at the Infrastructure Level

Port speed refers to the physical link speed of the network interface card (NIC) connecting your server to the top-of-rack (ToR) switch inside the data center. A 1 Gbps port means the NIC and switch port negotiate a 1000BASE-T or 1000BASE-SX link — the theoretical ceiling for data transfer is 1 gigabit per second in each direction (full-duplex).

However, the physical port speed and the guaranteed available bandwidth are two entirely different things. This distinction is the foundation of the metered vs. unmetered debate.

Key infrastructure concepts to understand:

• Uplink contention: Multiple servers can share a single uplink from the ToR switch to the core network. Even if your NIC is connected at 1 Gbps, the upstream pipe may be oversubscribed.
• rx + tx accounting: Bandwidth consumption is typically measured as the sum of inbound (receive) and outbound (transmit) traffic. A 99 TB monthly cap means 99 TB total across both directions, not 99 TB each way.
• Burst vs. sustained throughput: Shared connections may allow short bursts to full port speed but throttle sustained transfers when the shared medium is congested.

The Three 1 Gbps Port Options at AlexHost

Option 1: 1 Gbps Shared (Included in Plan)

The 1 Gbps Shared tier is bundled into the base price of the dedicated server plan. Your server is physically connected to a 1 Gbps port on a rack switch that is also shared with other co-located servers in the same rack or switch segment.

How it works technically:

The rack switch aggregates traffic from multiple servers onto a shared uplink. During off-peak hours, you may achieve speeds close to the full 1 Gbps line rate. During peak periods — when neighboring servers are running backups, bulk transfers, or high-traffic applications simultaneously — your effective throughput degrades proportionally.

Practical implications:

• Throughput is non-deterministic and varies with neighbor activity
• No SLA-backed speed guarantee
• Suitable for: development environments, low-traffic websites, staging servers, internal tooling
• Not suitable for: production databases with replication, high-volume CDN origin servers, or latency-sensitive real-time applications

Important fallback behavior: If you exhaust your traffic allocation on a metered plan, AlexHost automatically downgrades your connection to this shared tier until the next billing cycle begins. This is a critical operational detail — your service does not go offline, but performance becomes unpredictable.

Option 2: 1 Gbps Dedicated / Guaranteed Metered — 50 EUR/month

The Dedicated Metered option provides a guaranteed, uncontended 1 Gbps connection with a hard bandwidth cap of 99 TB per billing cycle (rx + tx combined).

What "dedicated" means here:

Your server's port is provisioned on a dedicated uplink segment, meaning you are not competing with other servers for bandwidth. The switch and upstream routing infrastructure reserve the full 1 Gbps for your exclusive use. This is enforced at the network layer, typically via traffic shaping or port-level QoS policies on the ToR switch.

Bandwidth math — does 99 TB fit your use case?

Overage behavior: Once the 99 TB threshold is reached, the connection automatically falls back to the shared 1 Gbps pool. There is no hard cutoff or service interruption — but guaranteed performance is no longer available until the billing cycle resets.

Best for: Production workloads with predictable, moderate-to-high traffic where a guaranteed speed floor is essential but bandwidth consumption is manageable within the 99 TB envelope.

Option 3: 1 Gbps Dedicated / Guaranteed Unmetered — 400 EUR/month

The Dedicated Unmetered tier delivers a guaranteed 1 Gbps dedicated connection with no bandwidth cap whatsoever. You can saturate the port continuously, 24/7, for the entire billing month without incurring overages or experiencing throttling.

Theoretical maximum throughput:

At full 1 Gbps sustained utilization (both rx and tx), the monthly data transfer ceiling is approximately:

• 1 Gbps = 125 MB/s
• 125 MB/s x 86,400 seconds/day x 30 days = ~324 TB/month per direction
• Full-duplex (rx + tx combined): up to ~648 TB/month

In practice, no real workload sustains 100% line rate continuously, but this illustrates the scale of headroom the unmetered tier provides.

Cost-efficiency analysis:

At 400 EUR/month, the unmetered option becomes cost-effective once your projected traffic exceeds roughly 100–120 TB/month. Below that threshold, the metered plan at 50 EUR/month is the rational choice. Above it, the unmetered tier eliminates both overage risk and the operational overhead of monitoring consumption.

Best for: High-volume content delivery, large-scale data replication, media streaming infrastructure, continuous CI/CD pipeline artifact distribution, or any workload where traffic is either very high or highly unpredictable.

Side-by-Side Comparison

Critical Technical Nuances and Edge Cases

The rx + tx Accounting Trap

Many administrators underestimate bandwidth consumption because they only track outbound (tx) traffic. If your server is both serving content and receiving large uploads — for example, a backup destination or a two-way sync node — inbound traffic counts equally against your cap. Monitor both directions independently using tools like vnstat, iftop, or nethogs to build an accurate consumption baseline before committing to a metered plan.

Billing Cycle Timing and the Fallback Window

The automatic fallback to shared bandwidth upon cap exhaustion is not instantaneous in all implementations — there may be a brief window where traffic continues at the dedicated rate before the switch-level policy takes effect. Do not rely on this window as a buffer. Build your traffic monitoring and alerting around 80–85% cap utilization to give yourself time to react, either by optimizing traffic or upgrading the plan mid-cycle.

Latency Is Not a Function of Port Speed

A common misconception: upgrading from shared to dedicated bandwidth does not reduce network latency. Latency is determined by routing path length, physical distance to the upstream peering points, and queuing depth at congested nodes — not by your local port speed. If your application is latency-sensitive (real-time gaming, financial APIs, VoIP), focus on the data center's peering relationships and geographic proximity to your users, not just the port tier.

NIC Bonding and Multi-Port Configurations

For workloads that require both high throughput and redundancy, consider whether NIC bonding (LACP/802.3ad) is available. Two bonded 1 Gbps ports in active-active mode can deliver up to 2 Gbps aggregate throughput while also providing failover. This is a separate discussion from the metered/unmetered tier but is worth raising with the AlexHost support team if your architecture demands it.

Choosing the Right Tier: A Decision Framework

Work through these questions in order:

1. Is your workload production-critical? If yes, eliminate the shared option immediately. Proceed to question 2.
2. Can you accurately estimate monthly traffic (rx + tx combined)? If no, or if traffic is highly variable, choose unmetered to eliminate operational risk.
3. Is your estimated monthly traffic consistently below 80 TB? If yes, the metered plan at 50 EUR/month provides a strong cost advantage with a comfortable buffer before the 99 TB cap.
4. Does your workload involve continuous bulk transfers, media streaming, or large-scale data replication? If yes, calculate whether 400 EUR/month is justified by the traffic volume. At 100+ TB/month, it almost always is.
5. Are you running a VPS Hosting environment or reselling resources to end users? Unpredictable aggregate traffic from multiple tenants makes the unmetered tier the only operationally safe choice.

Monitoring Bandwidth Consumption on Your Dedicated Server

Regardless of which tier you select, implement bandwidth monitoring from day one. Recommended tooling:

• vnstat: Lightweight, persistent traffic accounting by interface. Tracks hourly, daily, and monthly totals. Essential for metered plans.
• iftop / nethogs: Real-time per-connection and per-process traffic visibility. Useful for identifying unexpected traffic sources.
• Prometheus + node_exporter: For production environments, expose network interface metrics to a Prometheus/Grafana stack and set alerting thresholds at 75% and 90% of your monthly cap.
• Netdata: Zero-configuration real-time monitoring with built-in network throughput graphs.

Set up automated alerts well before cap exhaustion. A sudden spike — caused by a DDoS reflection attack, a misconfigured backup job, or a viral traffic event — can consume weeks of bandwidth in hours.

If you are also managing email infrastructure on the same server, note that high-volume mail delivery can contribute meaningfully to tx totals. Consider offloading transactional email to a dedicated Email Hosting service to keep your server's bandwidth consumption predictable and focused on its primary workload.

Security Considerations at High Port Speeds

A 1 Gbps dedicated unmetered connection is a powerful asset — and a potential liability if your server is compromised or targeted by amplification attacks. At 1 Gbps, a misconfigured DNS resolver or NTP server can become a significant source of reflected DDoS traffic, consuming your bandwidth and potentially violating acceptable use policies.

Baseline hardening steps for any 1 Gbps server:

• Disable open recursive DNS resolution unless explicitly required
• Apply ingress and egress rate limiting at the firewall level for ICMP, UDP, and other amplification-prone protocols
• Enable connection tracking (conntrack) and set aggressive timeouts for UDP flows
• Monitor for asymmetric traffic patterns (high tx with low rx) which are characteristic of outbound amplification participation
• Keep fail2ban or equivalent active to suppress brute-force SSH and application-layer attacks that generate sustained traffic

For servers also handling HTTPS termination, pairing your dedicated port with a properly configured SSL Certificate ensures that the additional throughput is not wasted on unencrypted traffic that exposes your users to interception.

Practical Key-Takeaway Checklist

Before finalizing your bandwidth tier selection, verify each of the following:

• Baseline your current or projected monthly traffic using actual logs or analytics — include both rx and tx
• Identify whether your workload has predictable traffic patterns or is subject to sudden spikes
• Confirm whether any co-located services (backups, replication, email) contribute to your bandwidth total
• Determine whether a guaranteed speed floor is operationally required or merely preferred
• Calculate the cost-per-TB breakeven point: metered at 50 EUR/month becomes more expensive than unmetered at 400 EUR/month once you consistently exceed ~100 TB/month
• Deploy vnstat or equivalent monitoring on day one, regardless of tier
• Set alerts at 75% and 90% of your monthly cap if on the metered plan
• Review your firewall ruleset to prevent amplification attack participation before enabling a high-speed unmetered port
• If traffic unpredictability is a concern, the 8x cost premium of unmetered over metered is insurance, not overspend

For teams evaluating a full infrastructure stack — including control panel management — reviewing the available VPS Control Panels options alongside your bandwidth tier can simplify ongoing server administration significantly.

Frequently Asked Questions

What happens to my server when I exhaust the 99 TB metered cap?

Your connection automatically transitions to the shared 1 Gbps pool for the remainder of the billing cycle. The server remains online and accessible, but throughput becomes best-effort and non-guaranteed. Service resumes at dedicated speed at the start of the next billing period.

Does "dedicated" port speed mean I have a physically separate cable from other servers?

Not necessarily a separate physical cable, but a logically isolated, reserved bandwidth allocation enforced at the switch and routing layer. Your traffic is not competing with neighboring servers for the same uplink capacity, which is what matters operationally.

Is inbound (download) traffic counted against the 99 TB cap?

Yes. The cap applies to the combined total of received (rx) and transmitted (tx) traffic. A server receiving 50 TB of backup uploads and sending 60 TB of content delivery would exceed the cap even though neither direction alone crossed 99 TB.

Can I upgrade from metered to unmetered mid-billing cycle if I approach the cap?

This depends on AlexHost's provisioning workflow — contact support to confirm availability. In general, it is advisable to upgrade proactively based on traffic trends rather than reactively when approaching exhaustion, as the transition may not be instantaneous.

Is a 1 Gbps port sufficient for GPU-intensive workloads like AI model training or rendering?

For most AI training workloads, the bottleneck is compute, not network throughput. However, if your pipeline involves continuous large dataset ingestion from remote storage or frequent model checkpoint synchronization, a dedicated unmetered 1 Gbps port is the appropriate choice. Teams running distributed training across multiple nodes should also evaluate GPU Hosting infrastructure purpose-built for high-throughput inter-node communication.