AAPT Home > Support Information > Proxy Servers > Current Cache Configuration
AAPT's web proxy cache service is, we believe, the largest and fastest in Australia. Multiple high performance Sun Ultra multiprocessor Unix servers in both Sydney and Melbourne. These cache servers are together currently processing a total of approximately 125 GB per day with over 1,000,000 requests per hour during busy periods. Typical service times are under 300 ms per request for the majority of requests.
The proxies operate in clusters. Traffic is directed to them by load-share devices in Sydney and Melbourne. If a proxy doesn't contain a requested object it fetches it from the Internet and caches it.
Performance, reliability and scaleability have been key drivers in the design. The architecture is such that any of the servers within a cluster can fail and traffic will automatically route via an alternate server or directly from the Internet. Connect has pioneered a number of enhancements to the Squid software on which the caches are based, and is generally recognised within the Internet community as an expert in caching technology.
As our traffic volumes continue to grow, our policy is to proactively review our cache implementation and expand our cache investments so as to continue to provide high service levels to customers.
All Connect Proxies run the Squid proxy, which is based on the Harvest object cache.
The Sydney servers are dual processor Sun Ultra 250s running at 250Mhz, with 1.5 gigabyte of memory and 84 gigabytes of disk each.
One of the Melbourne servers is a dual processor Sun Ultra 250, the other a dual processor Sun Ultra 450, both running at 250Mhz, with 1.5 gigabyte of memory and 84 gigabytes of disk.
The systems run Squid and act as sibling caches, using the "proxy-only" option. They are connected by 100 megabit Ethernet to our core routers, one step (via 100 megabit Ethernet) from our border routers and our external links.
| Copyright © AAPT Limited |
|