You can view graphs showing the most recent site serving capacity, utilization percentage, and traffic levels for each of your embedded sites from the home page in the Partner Portal. You can also view additional details and a historical view in the Site Capacity Reports.
In this article:
- OCA Site Capacity reports
- Interpreting site capacity and utilization data
- Site capacity planning and forecasting
OCA Site Capacity reports
Reports
- Site Capacity is defined as the maximum throughput (in Gbps) that can be served from the clustered set of OCAs within an embedded site before traffic needs to shed to another site.
- Utilization is calculated by comparing site capacity to recent peak traffic served from the site.
- Capacity values are updated daily. Refer to the "last updated" date on the report to understand when the data was last refreshed. This delay is necessary because capacity and peak traffic are calculated based on the summed bits from the peak hour of traffic for each day, and highest peak can only be determined after the day ends.
There are two detailed capacity reports available:
OCA Site Capacity (1 week)
This report displays port capacity vs. average traffic levels for the previous 7 days. The data refreshes daily. There are two charts in the report: Overall capacity and traffic levels for your organization, and a per-site view. If there are some ports on OCAs in the site that are not connected, and if connecting these ports would provide additional serving capacity, the charts include an additional value for the potential capacity that could be added to the site by connecting additional ports. More information below.
OCA Site Capacity (Historical)
This report shows a similar set of data over a longer period. Use this report to look at trends or consistency over time.
What do we mean by capacity "with additional ports"?
We do some calculations to determine whether and by how much each OCA would be able to increase its serving capacity if additional ports were connected. The chart will show an additional "with additional ports" value for the site if the site contains one or more OCAs that meet the following criteria:
- The OCA reaches a high % capacity on most days
- The OCA's capacity is usually limited by the number of connected NICs
- The OCA will see a significant increase in capacity if additional ports are connected
Running the reports
-
Open Metrics > Reports from the main menu.
-
In the panel that opens, select the OCA Site Capacity (1 week) or OCA Site Capacity (Historical) report, select options, and click Submit.
The report runs and results are displayed in the right panel.
Interpreting site capacity and utilization data
Generally, we recommend that average utilization for a site should average between 50-75% of the site's capacity. If you see some sites in the report that are running consistently at or above 75%, action may be warranted. However, the impact of a site running above the typical desired capacity utilization will depend on your particular network design - whether you have sufficient backup capacity in nearby peering or embedded sites, how high overall traffic levels are in each site, and so on.
It may be useful to look at site capacity and utilization in conjunction with the high level traffic patterns displayed in the Traffic by AS Path report to determine what solutions might work best for your network.
If an embedded site is running consistently above the 75% threshold, consider these options:
- If the site capacity graphs indicate that additional capacity can be achieved by connecting additional ports, consider connecting more ports where possible. You can view connected and disconnected ports on the Overview diagram for each OCA.
- If you have additional embedded sites in a similar geographic area that have more headroom, consider modifying your BGP announcements to re-route some traffic from the over-utilized site(s) to the under-utilized site(s).
- Similarly, if you have private peering (PNI) connectivity with Netflix, review the PNI capacity reports to evaluate whether some traffic could be shifted to PNI.
- If none of these solutions are applicable, the solution might be to augment over-utilized sites with additional OCAs. To discuss this option with the Netflix team, or for assistance troubleshooting over-utilized sites, you can open a ticket using the ticket issue type "Capacity / Additional OCA or PNI".
The site capacity model
Our capacity model considers many factors such as OCA hardware and software limitations, potential offload based on catalog size for the region, and so on. Therefore, you may see less serving capacity than what might be predicted based on connected NIC ports and interface speed alone. In small sites with 1 or 2 OCAs, these limiting factors may be more pronounced.
Capacity limiting factors may include:
Hardware factors
Components such as CPU, bandwidth, spinning disks, memory, or SSD can limit the capacity of each OCA. The health of each hardware component, including whether there are disabled disks, inactive NIC ports, or disconnected SFP optics, also contributes to capacity differences.
Site/cluster density
The total number and type of the OCAs within a site can contribute to variations in site capacity.
Offload and regional popularity
Every site attracts different traffic volumes based on the regional content that it stores and localizes, which informs the overall capacity of the site.
Traffic volume
The more traffic a given OCA can attract, the more its capacity scales up proportionally, within the limiting boundaries. This factor, which may vary from the normal daily pattern due to tentpole titles or live events, is typically what causes small peaks and valleys in the Site Capacity reports.
To summarize: Capacity for each OCA or site cluster is not a fixed number. All of the factors described here can contribute to normal slight fluctuations in embedded site capacity from week to week. These fluctuations can occur due to changes in title popularity, content placement, hardware health, or encoding strategies.
Site capacity planning and forecasting
Capacity forecasting starts with demand planning, using forward-looking signals that come from the business. These demand signals are translated into forecasted traffic demand in Gbps. The specifics of these signals will necessarily vary by region, country, and site, depending on such things as local device mix, Netflix plan mix, number and projected growth of subscribers, and the new titles or local live events that are projected to contribute to traffic volumes. We apply our internal architecture standards and resiliency standards, consider lead times for augments, and produce an internal OCA site capacity forecast plan and timeline.
The Netflix Open Connect team produces these internal OCA site capacity forecasts and augment plans quarterly. If Netflix traffic levels within a site are forecasted to go above the 75% utilization threshold within 12-18 months, the Netflix team will reach out to you to discuss potential augmention of the site. As mentioned above, you can also reach out to us via our ticketing system at any time.