SZM.com - COMDOM Antispam Case study

Client: SZM.com
www.szm.com



The Problem:

Since the early 2000s, the volume and sophistication of spam has grown significantly. Today, more than 80% of all messages are spam. As a result, ISPs and other network operators have been forced to increase capital and operational expenditures required for running a given level
of traffic. In addition, the volatility of spam flows poses further challenges in terms of real time network stability.

The Response:
Given the costs of spam on network infrastructure, antispam developers have focused increasingly on building systems that are both accurate (in terms of false positives/negatives), and fast. Second generation statistical content filters currently lead other approaches in their ability to learn about emerging patterns of spam and classify messages without posing significant costs in terms of computational power and bandwidth.

Objective:
This case study looks at the experience at a medium sized E-mail provider switching from a first to a second generation content filter.

About SZM.com:
SZM.sk started its operations as the first search portal in the Slovak Republic in 1997, and has remained one of the most visited sites in that country. SZM.com is a full service operator focused on providing messaging services, as well as hosting solutions for home and business users.

Dušan Štefánik, senior network administrator at SZM.com highlighted how the growth and evolution of spam has influenced their operations: “We used to rely on open source SpamAssassin running on a number of servers to deal with spam. As the volumes of spam increased and became less predictable, we had to add more servers. Still we experienced frequent breakdowns and had to further devote significant administrative resources to the problem.”

”Because of the rising costs of spam, we have decided to switch from an open source to a commercial product with lower administrative and hardware demands,” says Štefánik. “After testing a number of different systems, SZM.com chose COMDOM Antispam for Servers”. 

Antispam Technology: COMDOM Antispam for Servers
By automatically learning about the preferences of end users, statistical (Bayesian) content filters have become increasingly accurate in detecting spam. COMDOM Antispam was developed to combine accuracy and speed required by ISPs facing large volumes of complex spam. Low level programming allows the system to process significantly more messages per second than other Bayesian, fingerprinting, or reputation based filters.

Adoption of COMDOM Antispam at SZM.com: 

Baseline Network Technology:

Hardware:  Dual-Core AMD Opteron 1,8 GHz, 1GB  RAM
Operating System:  Linux (Debian)

Implementation:
Mr. Štefánik of SZM.com pointed out that “during the tests we were pleasantly surprised from the very beginning by simplicity of  COMDOM Antispam installation. Thanks to its compatibility with PostFix commands, no special training was needed and we implemented the system without any difficulties.”

Pricing model:
COMDOM Antispam can be licensed on a per CPU basis, a model which allows for significant savings for ISPs and organizations with a large number of end users. In addition to the cost savings that flow from the reduced number of servers, message delays, and operational expenses, this pricing model minimizes the Total Cost of Ownership (TCO) of a messaging network. Milan Igrini, Director of SZM.com, pointed out that ”COMDOM is among the most cut rate commercial solutions available to network managers.”

Tables


Message delivery delay:

Antispam typeAveragePeak
ComDom Antispam (now) 2 sec. 5 sec. 
SpamAssasin (before) 183 sec. 3621 sec. 


Hardware resources used

 Servers usedCPU cores usedRAM per serverCPU load per server (avg) Total load index 
COMDOM Antispam (now) 1 1     350 MB 
Spam Assasin (before)  3 6 1 GB 5,5 33

Total: 90% of Hardware resources saved