The Technical Side of Travel

The Deem IT /OPS and INFOSEC team have been hard at work, improving Deem’s systems and infrastructure.

Haarith

“We’ve completed more than a hundred optimizations, just this year,” said Haarith Devarajan, Vice President, Dev Ops, Deem.

The infrastructure upgrades mean Deem Work Fource runs more efficiently – saving travelers more time when they’re searching and booking their trips – and more securely.

“In the last six months, we have scaled our capacity almost three times,” said Devarajan. “Our new infrastructure design follows best practices and exceeds the speed and reliability goals we set for our team.

“Every customer facing change was extensively tested including specifically tuned buffering and compression settings meant to optimize the Deem ecosystem. We now save thirty to forty percent on average transfer times and have reduced the number of drops/reconnects by twenty percent. These are significant improvements for the system!”

For the full list of technical improvements, keep reading (note: unless you are an IT person, this may sound like a foreign language to you!):

  1. Upgraded underlying network infrastructure to 10G Top of the Rack switch with line rate uplink and aggregation layers allowing significant capacity transfer through the network.
  2. System auto tunes and optimizes when it recognizes mobile device handshake. Optimizations run from Layer 4 and Layer 7 - this includes different compression settings, TCP mobile optimized stack with extended congestion notification flags and larger buffers on client-side. Also supports MPTCP for LTE-Wifi seamless handover.
  3. LS record sizing is standard to help match TCP and reduce packet reassembly on TLS layer. As part of compliance, all connections are moving to TLS1.2 sessions.  
  4. Improved SSL session handling and compression managed by a dedicated ASIC processing unit that is significantly faster and can handle stricter standards. This also protects the ADC CPU during high traffic events allowing effortless scaling with traffic spikes.
  5. Highly tuned buffers for internal critical services to maximize transaction efficiency and reduce the stress of excessive connection handling on the server side.
  6. Latest OS and Hypervisor system with newer software. These systems significantly reduce the time to recover on failures compared to an equivalent RAID system.
  7. Complete system templates and prototyping that can launch an instance of webserver within seconds.