Scalability

The Sihope Building

With three different fiber carriers, Sihope's high levels of connectivity are very robust. Sihope can expand it's services to meet any need. We utilize extensive monitoring systems, high-availability hardware, and multiply-homed upstream fiber connections to keep the Internet completely accessible to our customers. Sihope Communications has independant Internet connections to seperate backbones over different carriers, which gives us world-class resilience to fiber cuts and network problems. When the extremely uncommon situation occurs where any of these connections fail, all traffic is automatically diverted to the remaining circuits. This configuration architecture ensures that Marvin remains connected to the Internet 24 hours a day. With access to fiber from three different carriers, additional bandwidth can be turned up quickly and easily.

The Data Center

Marvin's data center was specifically constructed for protecting sensitive systems. The data center has redundant dedicated HVAC systems controlling temperature and humidity. Every rack and chassis in our data center is protected with battery backup to maintain power through brownouts, spikes, and other temporary power hits. If there is a extended outage, our facility is equipped with a 6kVA diesel generator that automatically kicks in for seamless and uninterrupted operation. Even the very floor has been contructed of static dissipating tile, installed with copper glue, protecting against the buildup of static charges.

Marvin

Marvin was designed from the ground up to be both scalable and fault tolerant. By using hardware, software, and services that are standard in the industry, Marvin is able to grow with customer demand. The software that powers Marvin is a combination of custom programs developed in-house, and popular, well-known packages such as Sendmail. Marvin itself is written in a portable language that allows it to run on almost any platform.

	Inbound e-mail enters Marvin through servers running sendmail. Incoming messages are spread out across all of the servers automatically. If any server is busy, it will refuse new messages until it's caught up. Refused messages are given to another, less busy, server. The number of servers accepting mail can be increased quickly and easily as the volume of mail increases.
	Once mail arrives, it is passed on to the Marvin processes beginning with Anti-virus. If no virus is detected, it is sent on to the Anti-spam processes beginning with Whitelist. The e-mail goes through the spam processes in this order: Whitelist, Blacklist, SpamAssassin, Razor, DNSBL and then the Delay. If the message is flagged as spam during any of the processes, it is sent to a virus/spam database where it can be retrieved if desired. Marvin was written to allow any number of servers to work in parallel. With it's scalable design, new servers can be added in just a few hours. Multiple servers can run the same spam checks. There could be any number of servers that only check against the Blacklist. Marvin is designed to be flexible as it grows.
	Once the messages have gone through all of Marvin's processes, it reaches the mail servers which perform final delivery, sending the messages to the intended recipient. As mail volume increases, more delivery servers can be added to maintain delivery speed. The messages are evenly distributed between the delivery servers as they come out of the Marvin processes.