Why ViciDial server sizing matters
Under-sizing a dialer costs you dropped calls, laggy screens, and burned leads. Over-sizing wastes budget. The sweet spot depends on agent count, dialing mode, codec, recording, and your SIP trunk capacity. This guide explains practical rules of thumb to decide how many ViciDial servers you need and how to split roles for stable outbound campaigns.
Core concepts that affect capacity
- Dialing mode and ratio: Predictive and power dialing generate more concurrent calls per agent than manual or preview. A 3:1 ratio means roughly three live legs per active agent until calls connect.
- Recording and audio mixing: Recording every call and live monitoring increase CPU and disk I/O on dialers.
- Codec choice: G.711 uses more bandwidth but little CPU if no transcoding. Compressed codecs reduce bandwidth but can add CPU load when transcoding.
- SIP trunk limits: Your carrier’s maximum concurrent calls and calls-per-second (CPS) can throttle a campaign even when servers are strong enough.
- Role separation: ViciDial scales by adding dialer nodes that point to a central database and web server. Splitting these roles improves stability as you add agents.
- Storage: Call recordings grow quickly. Plan for local or network storage sized for retention policies.
Baseline hardware guidelines
Actual capacity varies by workload. These conservative specs are a proven starting point for outbound centers:
- Dialer node (Asterisk): 8 to 16 vCPU or physical cores, 16 to 32 GB RAM, SSD storage for spool and temp, 1 Gbps NIC. Keep transcoding to a minimum. Recordings can be written locally then archived.
- Database server: 8 to 16 vCPU, 32 to 64 GB RAM, enterprise SSDs. The database is latency sensitive. Avoid noisy neighbors.
- Web/reporting/recording server: 4 to 8 vCPU, 16 to 32 GB RAM, SSDs. Hosts the ViciDial UI, reports, and can handle archive jobs.
Rules of thumb for agent capacity
Use these as planning ranges. They assume mostly outbound, mix of manual and predictive, call recording enabled, no heavy transcoding, clean network, and a stable carrier.
- All-in-one server (DB + Web + Dialer on one box): typically up to 20 to 25 agents for outbound with recording. Suitable for small teams or pilots.
- Single dedicated dialer + combined DB/Web: supports roughly 30 to 50 agents depending on dial ratio and monitoring usage.
- Two dialers + dedicated DB + dedicated Web: comfortable around 50 to 90 agents. Add a third dialer if you run high predictive ratios or heavy QA monitoring.
- Three to five dialers + dedicated DB + dedicated Web/recording: scales to 90 to 200+ agents, subject to trunk capacity, network quality, and recording load.
If you run preview or manual-only with low dial ratios, these ranges increase. If you run aggressive predictive, do heavy call monitoring, or transcode compressed codecs, they decrease.
A quick sizing worksheet
Work through these steps to forecast how many dialer nodes you need.
1) Define live call legs
Concurrent outbound legs approximate to: Active agents x Dial ratio. For predictive at 3:1 with 60 active agents, expect around 180 outbound legs before connects.
2) Add headroom for connects and monitoring
When calls connect, you will have both agent and customer legs live. Add 20 to 30 percent headroom for connects, call progress detection, and occasional 3-way or transfers. In the 60-agent example, plan for 220 to 240 total legs.
3) Map legs to dialer nodes
A well tuned dialer node with recording and minimal transcoding typically handles 80 to 120 live legs without strain. Divide your total legs by 100 as a planning midpoint, then round up.
Example: 240 legs ÷ 100 ≈ 2.4, so plan 3 dialer nodes.
4) Size the database and web roles
- Database: 1 dedicated DB server supports multiple dialers for most centers up to 200 agents. Use fast SSDs and ample RAM for query cache and temp tables.
- Web/UI: 1 web node commonly supports 100 to 150 concurrent UI sessions. Add a second for redundancy or heavy reporting.
5) Check carrier and network limits
- Concurrent channels: Must exceed your peak legs with headroom.
- CPS: Predictive campaigns can burst dials. Ensure your carrier allows enough CPS or use multiple trunks.
- Bandwidth: For G.711, budget roughly 100 kbps per leg including overhead. Add margin for signaling, recording sync, and VPN if used.
Common scenarios
20 outbound agents, mixed manual and predictive, always-on recording
- Recommended: 1 all-in-one server or 1 dialer + 1 combined DB/Web.
- Why: Predictive bursts are modest and the UI load is light.
60 outbound agents, predictive 2.5 to 3:1, QA monitors 10 percent of live calls
- Recommended: 2 to 3 dialers, 1 dedicated DB, 1 dedicated Web/recording.
- Why: Recording and monitoring add CPU and I/O. Separate the DB to keep agent screens responsive.
120 outbound agents, aggressive predictive, remote teams across regions
- Recommended: 4 to 5 dialers, 1 strong DB with SSDs, 1 to 2 Web/recording nodes.
- Why: More dialers smooth bursts and isolate faults. Extra web capacity helps reports and real-time screens.
Additional considerations
- Transcoding: If codecs on phones do not match trunks, transcoding hits CPU. Standardize codecs end to end when possible.
- Recording retention: Estimate storage by average call length x calls per day x retention days. Archive to external storage if you keep recordings long term.
- Monitoring and QA: Live listening and barge add extra legs. Factor these into your leg count.
- High availability: If downtime is unacceptable, plan N+1 for dialers and consider failover for DB and Web.
- Security and latency: Keep latency low between dialers and database. Use firewalls, strong auth, and monitored VPNs for remote agents.
How FlareDial sizes and hosts your ViciDial
At FlareDial we deploy hosted ViciDial on dedicated servers sized to your campaigns. We install, secure, and maintain the stack. VoIP is delivered via our carrier partner our carrier, with capacity planning for concurrent calls and CPS. Pricing is prepaid from 80 USD per server per month.
Our typical approach:
- Start with a load estimate using your agent count, dial ratio, recording policy, and monitoring needs.
- Select a layout: all-in-one for small teams, or separate Dialer, DB, and Web roles as you scale.
- Choose server specs that leave headroom for busy hours, not just averages.
- Validate trunk concurrency and CPS with our carrier and configure safe dial limits in campaigns.
- Monitor CPU, channel utilization, queue times, and dropped call indicators, then tune or add nodes before you hit ceilings.
Next steps
If you share your agent count, target dial ratio, codec, and recording plan, we will map it to a clear Vicidial server sizing plan with a recommended node count and specs. Contact FlareDial to get a right-sized, dedicated ViciDial deployment that grows with your campaigns.