The main risk sits between “in stock” and the actual handover. CPUs may come from different generations, drives may have different wear, servers may be spread across racks and the subnet may arrive before the route is ready.
Put the conditions in one document before the invoice. Messages are useful for clarification, but the configuration, schedule, money and readiness criteria belong in the quotation or its appendix.
Define what matching hardware means
“Equivalent configuration” is too broad for a cluster. CPU generation, memory layout, controllers, network adapters and drive models can matter even when core count and total RAM look identical.
If substitutions are allowed, define the boundaries in advance: acceptable CPU models, whether drive vendors may be mixed and how much frequency or endurance may differ.
- exact CPU model and socket count in every server;
- total RAM, module type and upgrade path;
- drive model, interface, quantity and condition;
- network adapters, port speeds and redundancy method;
- IPMI or KVM, virtual media and console access.
Separate stock from the delivery schedule
Ten chassis in a warehouse do not prove that ten matching servers are ready. Confirm the number of assembled machines and state how many will be delivered in the first and later batches.
Each batch needs a date, quantity and substitution rule. If staged delivery is acceptable, confirm whether part of the fleet can start without losing the volume discount or shortening the paid term.
- matching systems ready now;
- size and date of each batch;
- what counts as a delay and when a substitute is offered;
- when billing starts for each machine;
- whether the customer may reject a substitute.
Describe the network and the /24 separately
A subnet may be assigned to one host, routed to a cluster, split by VLAN or announced through BGP. These are different designs. Record the prefix, gateway or next hop, routing method and MAC restrictions.
Port speed is also incomplete without the bandwidth model. Confirm dedicated or shared capacity, included traffic, overage rules, DDoS handling and the PTR change procedure.
- block size and usable address count;
- who routes the prefix and to which next hop;
- VLAN, allowed MAC addresses and anti-spoofing;
- port rate, committed bandwidth and traffic allowance;
- PTR, LOA, IRR/RPKI and abuse terms when required.
Price the complete paid term
Compare the fleet total rather than one monthly server price. Include addresses, bandwidth, extra ports, remote hands and one-time work. A 3, 6 or 12-month calculation should show the discount and the next-period price separately.
Define when rental starts. For staged delivery, per-server billing from the handover of a working machine is usually the clearest rule, but it still needs to be written into the order.
- first payment and every one-time charge;
- volume and term discounts and whether they stack;
- price lock for the prepaid period;
- renewal price and change notice;
- refund or recalculation if part of the fleet is late.
Accept every server against one protocol
A password email is not enough. Check hardware inventory, SMART data, memory, IPMI access, port rate, routes, assigned addresses and a controlled reboot from the management interface.
Store results against the serial number or internal asset code. This shows which system failed acceptance and when its billable period started.
NextWhat the final quotation should contain→What the final quotation should contain
The final document should connect configuration, quantity, subnet, schedule, acceptance, support and price. If BGP or routed address space is required for the workload, the server is not ready until the route works.
- ten systems and permitted substitutions;
- delivery plan and billing start;
- network design and readiness tests;
- total term price and renewal price;
- hardware replacement SLA and escalation path.