In this article, we are going to look at different rationale for optimizing networks.
Latency Sensitivity Application
- Certain real time services are more sensitive to latency.
- Streaming data packets
- VOIP, video, RDP
- Methods for optimizing the network to help with latency sensitivity.
High Bandwidth Applications
Applications that generate a lot of network traffic. Not latency sensitive, but use a lot of resources. These include:
- Video/music streaming
- File streaming
- Cloud applications
- When people try to use cloud applications they are using remote resources, so that requires networks
- Want to track down the sources of high bandwidth uses.
Quality of Service
- Ensures priority of latency sensitive packets
- Built into network devices
Traffic Shaping
- Can delay less important traffic
- Makes the most of available bandwidth by throttling certain data streams
- May be on the user end or the ISP end.
Load Balancing
- Load balancing distributes workload across multiple devices.
- The user experience must be as transparent as possible.
- Load balancing methodology:
- Round robin between servers
- May not be the best solutions in situation where requests vary in the amount of work they require
- Least connections
- Use the server that is least busy.
- Fastest response
- Use the server with the fastest response time.
- Weighted round robin
- Devices are weighted based on operational capacity
- Round robin between servers
Uptime
Important statistic to administrators. An IT infrastructure doesn’t live in a bubble, it is there to server end users, so we need to make sure we have a large up time.
- Percentage of time services are available
- Needs vary
- Check SLA
- Standard Lease Agreement guarantees a level of uptime
- May include several computers
- Redundancy
- Want to minimize single points of failure
- Power back up
- Secondary Internet
- Backup sites
- Redundancy
High Availability
Translates into a high uptime. Uses a 24 hour, 365 day calendar to calculate:
90% uptime equals 36.5 Days of downtime/year ==> 16.8 hours/week
99% uptime equals 3.6 Days of downtime/year ==> 5.04 hours/week
99.9% uptime equals 8.76 hours of downtime/year ==> 10.1 minutes/week
Caching Engines
Devices/software that allow network to save frequently used resources to provide rapid access. These include:
- Load Balancers
- Proxy Servers
- Can act as a cache for certain websites
- Local Cache
- Computers have a local DNS cache to save time.
- Need to consider refresh time of caching with regards to resource refreshing.
Improves connection speed and reduces strain as there is less distance needed to service a request.
Fault Tolerance
- Having fail over devices available if something goes off line
- May be less robust than primary device, but maintains network connections
- Preferably automatic
- Routers, switches, cables
- UPS, Data back ups, cold/hot sites
- Configure them to send warning when failover kick in
- Need to make sure failover devices work
- Have scheduled outages to make sure backups work
CARP
- Common Access Redundancy Protocol.
- Allows multiple host share IP.
- Transparent fault tolerance.
- Each device has shared IP address and private IP address for management.
- Can be failover or load balanced.
- Need to make sure failover devices work.
- Have to have a fake outages to make sure system works.