If you’re working in a small business or network then this issue will probably never arise. However with the growth of the internet and web enabled devices and clients it’s an issue that will almost certainly effect most network administrators. Do we just keep adding an extra proxy to expand capability and bandwidth or should you install an array.
Nevertheless the solution can be dependent on a variety of external factors. for example in the event the corporation is concentrated in a single location, just one level of proxies is a better solution. This reduces the latency as there’s only a single additional hop added by proxies, as opposed to two or more with tree structured proxy hierarchies.
Although the general rule would be to have one proxy server for every 5000 (possible, not simultaneous) users, it doesn’t automatically mean that a company with 9000 users should have 3 departmental proxies, that are then chained to some most important proxy.
Instead, the 3 proxies might be installed in parallel, using Cache Array Routing Protocol (CARP) or another hashbased proxy selection mechanism. Larger corporations with in-house programming skills may have resource to create custom solutions too which work better to a specific environment which perhaps incorporates remote VPN access to the network too. For example many larger environments have different levels of security in place and have various zones which need to be isolated, generic ‘serve all’ proxies can be a significant security issue in these environments.
This approach can also combine multiple physical proxy caches into a single logical one. Ln general, such clustering of all proxies is recommended as it increases the effective cache size and eliminates redundancy between individual proxy caches. Three proxies, each with a 4 gigabyte cache, would give an efficient 12 gigabytes of cache when put up in parallel,as opposed to only about 4GB if used individually.
Generally, some quantity of parallelization of proxies into arrays is obviously desired. Nevertheless, the network layout might dictate that depart psychological proxies be utilized. That is, it is not feasible to have all of the trafc originating from the entire company go through one array of proxies. It can cause the entire array to become a 1/ O bottleneck, even when the individual proxies of the variety have been in individual subnets. The load created by the users can be so high that the subnets leading to the proxies may choke. To alleviate this, some departmental proxies need to be deployed closer to the end customers, in order that a number of the traffic created by the users will not reach the main proxy array.
Failover? Since proxies are a centralized point of traffic it’s vitally important that there is a system in place for failover. If a proxy goes down, users will instantly
lose their access to the internet. What’s more it may be that many important applications rely on permanent internet access to keep running. They might need access to central database systems or perhaps need frequent updates or security patches. In any ways, internet access is often much more crucial than simply the admin office being able to use Amazon, surf UK TV abroad or check the TV schedules online.
Failover might be achieved in many various ways. There are (relatively expensive) hardware solutions which transparently change to a hot standby system in the event the primary system goes down.
Nevertheless, proxy autoconfiguration and CARP provide more cost effective failover support. During the time of this writing, there are a couple areas in customer failover sup port which might be improved. Users have a Propensity to detect a intermediate proxy server going down by seeing fairly long delays, and possibly error messages. A proper proxy back up system should be virtually seamless and provide similar levels of speed and bandwidth than the primary system.