How is the port chosen when none of the three rates meters transfer?

What unmetered means

Unmetered means that after a tier is selected, the bill is not settled by a transfer pack or a price per GB. A machine that runs the port full for a month still pays that tier’s monthly rent times the months in the term. It does not mean the port has no rate ceiling. 250M is about 250 Mbps, 1G about 1 Gbps, and 2G about 2 Gbps.

Whether a month of downloads adds another charge, and whether pages slow at the peak, are two questions. The first does not arise on these three tiers. The second depends on how much data the workload pushes in the same second, and on whether the far network keeps up.

The three monthly prices

The add-on joins the monthly rent, then the term is prepaid. The 128GB machine on 250M is $455 a month. At 1G the monthly rent is $703, and the quarterly amount is $2,109. At 2G the monthly rent is $1,202. The 256GB machine’s base rent is $612, and the bandwidth add-on is the same. Those figures leave out added disks, extra addresses, and Windows.

Choose by the peak, not by the month’s total

A content site with many people pulling the same file at once is judged by the exit in that moment, not by the month’s total transfer. About 250 Mbps suits a modest number of connections and small responses. Images, installers, and video segments sent to many clients at once fill the port first. Pages slow. The end of the month does not open a second bill for the byte count.

Database replication, object-storage sync, and site backups that share one window also fill the port. Moving the backup into a quiet hour is sometimes the better step before a jump to 2G. If the quiet hour still cannot finish the transfer, and the workload cannot wait, then 1G or 2G is the next step.

Address count is a different line

Bandwidth selects the port rate. IPv4 selects how many addresses. One IPv4 is already included. Each extra address is $2 a month, up to 255 more. Another address does not turn 250M into 1G. Several sites on one port share that same rate.

The machine is deployed in Tokyo, Japan. The port tier does not change SSH or RDP. It changes the transfer ceiling of this machine. Address and port are both chosen at order time and written into the same monthly rent.

What to check after the machine is online

1. Read the negotiated link rate

   In the system, read the speed the NIC negotiated and confirm it is not below the purchased tier. When the negotiated value is higher, the actual exit is still limited by the purchased port.

2. Sample the busy hour

   Record send and receive rates for a few minutes. If the curve often sits on 250 Mbps and response time grows, then evaluate 1G.

3. Separate backup from live traffic

   When a backup fills the port, live requests slow with it. Shift the window first, then decide whether to raise the tier.

Raising the tier changes the bandwidth line inside the monthly rent. It does not change direct-attached disks, and it does not change the core count of the EPYC 7543P. There is no setup fee and no contract. The term is still monthly rent times the months.

How to tell whether the port will fill

Split traffic into two kinds. One is requests from visitors or clients at any hour. The other is backup, image sync, and batch release. The port curve is the sum. A month’s total misleads: many bytes, spread out, can still sit inside 250M. A modest monthly total can still fill the port for ten minutes a day, and requests in those minutes slow down.

A rough estimate multiplies the number of transfers in flight by the data each one needs per second. Many people and small responses often still fit in 250M. Fewer people, each pulling a large file, fill the port first. After it is full, new connections are not billed by transfer. They slow, or they fail inside the application’s own timeout.

1G over 250M adds $248 a month and is about four times the rate. 2G over 250M adds $747 a month. If the peak is only a little above 250M and only during the backup window, move the backup to a quiet hour before buying 2G. If the peak still sits on the ceiling after that move, step up to 1G. 2G is for a sustained large release, distribution, or replication.

Several virtual machines or services share this one port tier. Giving one virtual machine its own address does not give it a private 1G. Further limits inside the system are the workload’s own queues. They do not change the facility-side ceiling, which is the 250M, 1G, or 2G chosen on the order.

After a tier change, sample the busy hour the same way. Look at negotiated rate, application timeouts, and how long the backup takes. A NIC that negotiates 10 Gbps while transfer stays near 250 Mbps means the limit is the purchased port, not the NIC. Compare that with the bandwidth line on the order and confirm that tier is the one that was bought.

Sites, database replication, and backup on the same machine share this port. Binding one site to an extra IPv4 does not give that site a private 1G. If the backup must not crowd daytime access, move the backup time before buying a higher port. If the daytime curve still sits on the ceiling after that move, raise the port one tier and recompute quarterly or annual prepay from the new monthly rent: the add-on times the months, with no separate discount.

Note the direction while sampling. Many sites send more than they receive. Backup is often the reverse. Watching only inbound understates the load. On Linux, read interface counters before and after the peak, divide the difference by the seconds, and compare the approximate rate with 250, 1000, or 2000 Mbps. On Windows, use the system network counters for bytes sent in the same interval. The comparison is a rate, not the bytes accumulated in the month. When that rate is often above about four-fifths of the purchased tier, the next configuration includes the higher port, rather than waiting until pages already time out. Raising the tier changes only the bandwidth add-on in the monthly rent. Disks, memory, and address count stay put unless those lines are changed in the same order.