SMTP.com gets new datacenter, ditches Rackspace

SMTP.com a provider of email relay and email delivery services has shifted its infrastructure strategy. Previously they contracted managed servers at multiple hosting providers such as Rackspace and SoftLayer, that is over. Now SMTP.com has opened its own, new datacenter.

Transactional email increase

Semyon Dukach, CEO of SMTP.com commented that the change was partly driven by an increase in the amount of emails sent based on behavioral e-commerce systems. Transactional email is one of the email types where email relay services are often used. Transactional email include sales order receipts, shipping acknowledgements, billing statements, and registration confirmations. Requiring high deliverability even when not making use of a full service ESP.

After smtp.com left, Pat Condon had some extra Rackspace to spare

Email delivery service market

SMTP.com continues in their press release: For today’s email delivery service market, service level quality is a critical requirement as new social networks, behavioral commerce systems, and other mission-critical applications rely upon email as the primary means for sending customer communications. We concur, it is very important to know how reliable your email service provider is. This of course also includes the infrastructure.

Send reputation and new IPv4 addresses

SMTP was allocated 4,096 portable IPv4 addresses, a big enough supply of IP addresses for all its customers. The move will force SMTP.com to use these new IP addresses for sending. The IP based send reputation built by SMTP will be lost, and needs to be rebuilt at the new hosting center.

SMTP.com is locating its new data center at the Hurricane Electric facility in Freemont, California, with alternative direct routing over the XO Communications fiber network. CISCO and Dell will be providing the primary router and server hardware.

About Jordie van Rijn


Jordie van Rijn is an independent email marketing consultant at www.emailmonday.com. He specializes in smart email marketing, event-driven campaigns and vendor selection. As a consultant, Jordie will help you get the most out of your email marketing efforts.

  • http://twitter.com/MarketingXD MarketingXD

    Everything at one datacenter? Do they have a Disaster Recovery site too?

  • http://twitter.com/MarketingXD MarketingXD

    Like the picture, but I hope they use rack-mount or blade servers!

  • http://twitter.com/MarketingXD MarketingXD

    BTW my question about a DR site was serious. Do they have a second site?

  • http://twitter.com/jvanrijn Jordie van Rijn

    It doesn’t seem so, based purely on the press release. They do have an alternative routing over the XO Communications fiber network, but that is not Disaster Recovery site. I’ll try and get a hold of them and ask.

  • http://twitter.com/MarketingXD MarketingXD

    It’s a bit unfair to single out SMTP over Disaster Recovery. I think a “dirty secret” about ESPs is that very few of them have true DR sites.

    ESPs are mostly based on 2000-era technologies, which means it takes a lot of effort to keep a second site ready to go, with a full copy of all the customer databases. And of course their business models require them to minimise cost-to-serve.

    I have never heard of an ESP doing a real test of DR by pulling the plug on its main site, on a busy day!

    Would make an interesting report.

  • http://twitter.com/jvanrijn Jordie van Rijn

    I wouldn’t be confident about a power outage with most ESPs, let alone a permanent disaster like fire, flood or worse. And it *does* happen, force majeure? Not really if you know it’s coming, eventually. So what is a company looking for an ESP to do? Get dirty and ask for proof of DR?

  • Richard SMTP

    At this point, we are not completely ditching Rackspace and Softlayer partly for disaster recovery reasons and partly because some customers will not want us to switch their existing IPs and rebuild reputation.

  • http://www.smtp.com Richard SMTP

    Jordie, your question to us regarding how are we warming up the new IPs
    at our new data center is a great question and very insightful to some
    of the challenges in changing infrastructure with established customer IPs.  
    We are at the beginning stages but we have developed a two-prong
    strategy to tackle this challenge. 

    One prong relies on obtaining permission from a specific set of customers to use their traffic to warm up a multitude of IPs
    (more than what a specific customer really needs).  After the IPs are warmed, we
    would then switch the customer’s traffic to their new permanent IPs which would be just one or a subset of the warmed IPs.   All
    the remaining warm IPs would now be available for us to use for additional customer migration.

    The other approach would rely upon us splitting the traffic for each of our customers between their existing IPs and the new IPs in the new data center.   Gradually, we would increase the distribution weight to the new IPs until all traffic is flowing through the new IPs only.

    We are at the early stage, so we will move slowly with only a few beta customers so we can carefully monitor and evaluate success of each approach.