Web Proxy Caching: The Devil is in the Details

1
Web Proxy Caching The Devil is in the Details

• Ramon Cacere Fred Douglis Anja Feldmann
• Gideon Glass Michael Rabinovich
• ATT Labs-Research
• Florham Park, NJ, USA

2
Brief Review
clients
servers
Reply
Req.
proxy
Req.
Reply
3
Brief Review

• Client send requests to the proxy.
• If the requested document is in its cache, the
  proxy serves the request from its cache.
• Otherwise, the proxy forward the request to the
  server.
• Server replies the request through the proxy
  (proxy keep a copy of the requested document).

4
How does proxy caching improve performance?

• Reduce the user-perceived latency associated with
  obtaining Web documents.
• Lower the network traffic from the Web servers.
• Reduce the service demands on content providers.

5
Previous Work

• High level details hit ratio byte hit ratio
• Ignored exceptional cases such as connection
  aborts.
• Omitted the effect of cookies on cacheability of
  resources.

6
This paper argues that...

• Low-level details have a strong impact on
  performance, particularly in heterogeneous
  bandwidth environments.
• Aborted trasfers can contribute significantly to
  total bandwidth requirements.
• Cookies dramatically affect the cacheability of
  resources therefore, affect the latency.
• Caching TCP connections at proxy can reduce
  latency more simply caching data.

7
Simulation

• Web proxy simulator (PROXIM)
• Workload trace from ATT Worldnet
• 12 days dialup traffic on a FDDI ring
• encrypted IP addresses
• contained information on both TCP events and HTTP
  events

8
Simulator PROXIM

• Simulator Cache
• sufficiently large
• included proxy overhead in the request service
  time
• Network Connections
• zero or more open connections (cache-to-proxy
  proxy-to-server)
• Proxy closes client-to-proxy connections with 3
  minutes of idle time.
• Proxy-to-server connections are timeout after 30
  secs of idle time.

9
Simulator PROXIM (cont.)

• Document Transfer
• Packet-level delivery with TCP slow-start
• 1500-byte packets
• constant round-trip time estimate for each
  connection
• Latency Calculations
• connection setup time
• HTTP request-response overhead
• document transfer time

10
Results (Hit Ratio)

• When taking cookies into account
• Hit ratio decreases from 54.5 to 35.2.
• Byte hit ratio decreases from 40.9 to 30.42.
• Solution Techniques aimed at enabling caching
  documents with cookies are important for
  increasing hit rate.

11
Results (Bandwidth Savings)

• When the proxy is present, the bandwidth
  consumption of aborted requests is higher due to
  the bandwidth mismatch between the connections of
  client-to-proxy and proxy-to-server.
• Question how much would this be offset by the
  savings from caching?

12
Results (Latency Reduction)

• Caching has limited effect on improving latency
  (reduced the mean by 3.4, the median by 4.2)
• Solution Maintain persistent connections between
  clients and servers
• Proxy as a connection cache.
• Re-use persistent proxy-to-server connection for
  obtaining documents for multiple clients.

13
Questions

• How does the proxy manage a connection cach?
• How many simultaneous connections it should
  maintain with a server or a client.

14
Conclusion

• For dialup users
• Hit ratios is lower than those reported
  previously.
• Bandwidth savings non-exist or is negative.
• Latency reduction coming mostly from caching TCP
  connections rather than documents.