When designing an email infrastructure to send high volumes of email, you need to determine your email throughput needs. Throughput refers to how much data can be transferred within a specific time frame. It is a practical measure that is influenced by many factors including server power, network speeds, concurrent connections and more. This article will explain some of the factors that you can control to help you design an email sending infrastructure that fits your business needs.
To Determine Email Throughput, Identify Email Use Cases
The primary thing to consider when planning your email infrastructure is your email use cases. Simply, what types of emails are you going to be sending? Are they time sensitive? Heavy in images or content? Sent to large mailing lists? These are just some of the questions to be considered when determining email throughput needs.
For example, some transactional emails like password resets are extremely time sensitive, but are not sent in bulk. Other emails, like newsletters or marketing emails may be less time sensitive, but are not unique to each individual recipient.
Email Sending Rate
Throughput and email sending rate are similar, but not exactly the same. The sending rate refers to the maximum sustained rate at which you can deliver email to the server, have it processed, and delivered to the recipient’s mail servers. Each of these steps is impacted by several factors that may impact the speed of your email sending.
First, let’s look at the speed by which emails are delivered to your server. The following factors all impact the speed:
- Email sending program used
- Proper use of concurrent connections
- Network speed between the sending program and LuxSci
- Server power (CPU cores, memory, and disk speed)
- Use of SMTP pipelining
- Message size
Next, the speed by which the messages can be processed (i.e., encrypted, tracked, etc.) may be impacted by the following:
- How frequently your sending email program is pushing new messages to the server
- Server power (CPU cores, memory, and disk speed)
- The size of your messages
- Rate limiting or sending rules (i.e., for IP warm up)
Finally, the speed by which the messages can be delivered to recipients. Some of these factors may be outside of your control because they depend on the recipient’s mail server:
- How much your sending email program is pushing new messages to the server
- Server power (CPU cores, memory, and disk speed)
- Message size
- Recipient server and DNS performance
- Block listing and greylisting
Sustained vs. Burst Rate Email Sending
For some businesses, the speed of sending is very important. There are two significant modes of sending, Sustained Rate and Burst Rate.
Sustained Rate
Making one or more connections to your server and then sending at a consistent rate for a long period of time. E.g. 20,000 messages/hour for minutes or hours. Sustained rates are relevant when sending an email marketing blast to a large subscriber list or when offloading a large number of messages from an external application through your sending server.
Burst Rate
A “burst” refers to sending a bunch of messages in a very short time; e.g. 25 or 50 messages within a second or two. Burst differs from “sustained” in that “bursts” are short duration events. When you are not “bursting,” you are infrequently sending only a few messages. Bursting is relevant for transactional email messages when you are sending them “as needed” but require the ability to send many at once should that need arise, as it will.
FACTORS AFFECTING YOUR SUSTAINED SENDING RATE
Most people focus on sustained sending when determining their email throughput needs. There is a theoretical maximum number of messages/hour that any server can manage, but the email throughput is affected by a number of factors.
1. Server Power
More powerful servers can handle more concurrent connections, process more messages at once, deliver more messages at once, and thus give you higher sending rates. The number of CPU cores is the primary factor governing performance; however, memory and disk speed are also important.
|
Dedicated Server Size |
Maximum Messages/Hour |
Maximum Messages/Day |
Recommended # of Sustained Connections |
| 1 CPU | 20,000 | 450,000 | 1-2 |
| 2 CPU | 40,000 | 900,000 | 2-3 |
| 4 CPU | 80,000 | 1,800,000 | 4-6 |
| 8 CPU | 160,000 | 3,600,000 | 8-12 |
For increased capacity, high burst rates, or to account for other factors (see below) like large messages, lack of pipelining, etc., we recommend choosing a larger dedicated server or load-balanced multi-server clusters.
2. Email Sending Program
Different email marketing/sending programs have different capacities for how efficiently they can send high volumes of email. Simply switching from one program to another can increase your throughput significantly.
3. SMTP Pipelining
Use an email marketing program that supports SMTP Pipelining. With this feature, the program opens a single authenticated connection to the server and sends multiple messages in succession over that connection. This is significantly faster than opening a new connection and authenticating again for every single message to be sent.
4. Concurrent Connections
If your sending program can open multiple connections to our SMTP servers at the same time and send multiple messages at once, then you can delivery messages to our servers more quickly. Depending on how powerful your server is and how slow your network connections are, you can use more and more concurrent connections to achieve higher and higher sending rates. Combined with SMTP Pipelining, an appropriate number of concurrent connections goes a long way to maximizing your throughput potential.
Under most circumstances, sending programs can deliver messages to our servers much more quickly than the servers can process those messages (i.e., encrypt them, track them, log them, etc.) and then deliver them to the recipients’ mail servers.
The recommended maximum number of concurrent connections to use for sustained email sending is given in the table, above. It is important to note that while exceeding that recommended value will get your email to your server faster, sustained sending at that rate will overtax the server and cause the overall rate of processing and delivery of your messages to your recipients to slow down. Stick to the recommended number of concurrent connections for sustained sending.
5. Network Speed
Your messages must be transmitted from your email sending program to our servers. The speed of your network, any packet loss, and the distance between your sending computer and our servers can significantly impact the maximum rate that you can send messages.
6. Message Size
Like network speed, the larger your messages are, the more time it will take to transmit them from your email sending program to our servers. For example, on a fast network, increasing message size from 5KB to 1MB decreased throughput by 50%. The actual impact of size will depend greatly on your network speed and other factors.
7. Recipient Email Server and DNS Speed
In order for our servers to deliver messages to your recipients, the servers have to (a) look up where to deliver the messages to (DNS), and (b) talk to those servers to deliver the messages to them. If the DNS lookup as slow or the recipients are in domains that do not exist, this step can take a while. If the recipient’s email servers are overloaded, on a slow network, or are having other issues, then the delivery phase can be slow.
Delivering many messages to recipients with these issues can slow down your entire sending process as other messages have to wait for the slower ones to be sent.
8. Block listing and Greylisting
If your email content or sending patterns have gotten your server block listed or greylisted, then delivering email messages to recipient servers that are now blocking you or deferring you can slow down your delivery rates.
This can also cause a backlog of messages that are waiting to be accepted by these recipient servers. The need to Retry the backlog slows the process of delivering all other messages (though it does not block their delivery).
9. Recipients per Message
Most email marketing and transactional email messages are sent individually to each recipient. This allows easier tracking of opens, views, etc. However, if you send each message to many recipients (e.g., 100s of recipients), then your sending program has significantly fewer messages to deliver to our servers. The delivery of your messages to our servers can happen 10s or 100s of times faster than if you sent in the usual one recipient/message pattern.
Additionally, if each message contains only recipients in a specific domain, then actual delivery to these recipients can be expedited by this efficiency. E.g. If you are sending to 500 AOL users, instead of sending 500 individual messages, send one message with all 500 AOL users listed as “BCC” recipients. Delivery can be almost 500 times faster.
10. SecureLine Email Encryption
In the example sending rates, we assumed that all messages would be sent using SecureLine TLS. If PGP, S/MIME, or Escrow are used, that can also impact email processing speed.
11. Outbound Email Processing
If you are using other email processing features that come with High Volume Secure Sending (e.g. content scanning, recipient restrictions, size restrictions, etc.), this processing may affect your maximum sending rate.
12. Throttling
LuxSci has features that can be used to rate limit the processing of outbound email. This is usually used when “warming up” IP addresses to assist senders in sending slowly and establishing their IP address reputation.
If rate limiting is enabled, your maximum sending rate will be capped.
Need Help Determining Email Throughput Needs?
The experienced team at LuxSci can help you determine the appropriate configuration to suit your business emailing needs. Please contact us to learn more about Secure High Volume Email Sending.
