Google's SPDY will speed up downloads

Enlarge

(PhysOrg.com) -- As part of its effort to speed up the Web, Google is experimenting with SPDY, a new application layer protocol, that it hopes will speed up the conversation between browsers and Web servers and enable Web pages to download up to twice as fast.

According to software engineers Mike Belshe and Roberto Peon in Google's research blog SPDY (pronounced "speedy") protocol was tested in the laboratory by downloading each of the top 25 websites 10 times. The test environment simulated home network connections, and used a special web server and a Google Chrome browser prototype. The tests indicated SPDY gave page load times up to 55% faster.

SPDY does not replace HTTP, but rather creates a session between the HTTP application layer and the TCP (Transmission Control Protocol) transport layer and basically augments HTTP. It speeds up web page downloads by using several techniques, including allowing multiple simultaneous HTTP requests per TCP session, request prioritization, and compressing the data to cut the number of packets. It does not replace HTTP headers but it overrides HTTP's data transfer formats and connection management features.

HTTP allows browsers and servers to understand each other and translate data from a Web publisher to a page displayed in a browser. It became the web standard in 1996, but Google's development team think it needs the update to make it faster.

SPDY's performance in the real world still needs to be evaluated. Web pages would not need to be changed for the SPDY protocol to work, but Web servers and browsers would. If it is to be widely used, the new protocol would therefore need to be supported by browsers other than Google's own Chrome, such as Firefox and Internet Explorer. While Mozilla may be willing to support it, convincing Microsoft to do the same may be more troublesome. Previous developments aimed at improving download speeds, such as SST (Structured Stream Transport) and SCTP (Stream Control Transmission Protocol) have not become widely used.

© 2009 PhysOrg.com

   

• Bob_Kob - Nov 16, 2009
  • Rank: 4.8 / 5 (5)
  If chrome becomes any faster, webpages will open before i actually click on them!
  • report abuse
  • • Current rank
    • 1
    • 2
    • 3
    • 4
    • 5
• maxmc - Nov 16, 2009
  • Rank: 5 / 5 (3)

  If chrome becomes any faster, webpages will open before i actually click on them!

  yea Google is in on the whole LHC bein a time machine thing. Hey at least we'll already have physorg up before we sit down!
  • report abuse
  • • Current rank
    • 1
    • 2
    • 3
    • 4
    • 5
• hizon - Nov 16, 2009
  • Rank: 5 / 5 (2)
  It's not all about altruism, of course, said by some < href=http://www.research-service.com/">paper writers. Improvements like those Google is seeking with SPDY, the Go! programming language, and Native Client will all greatly benefit the Chrome browser and Chrome OS. After all, an OS which relies heavily on the cloud for access apps and data will certainly perform better with an improved protocol powering the client and server. It's probably a safe bet that Google would roll SPDY on their own servers early on to give apps like Google Docs and Picasa Web a performance boost on the Google platform.
  • report abuse
  • • Current rank
    • 1
    • 2
    • 3
    • 4
    • 5
• Husky - Nov 16, 2009
  • Rank: 5 / 5 (1)
  How about a look ahead buffer/ branch predictor that will cache/download (part) of the links on your webpage before clicking/while you are still reading, for near instantenious display. maybe its already being done, i dont know...
  • report abuse
  • • Current rank
    • 1
    • 2
    • 3
    • 4
    • 5
• jonnyboy - Nov 17, 2009
  • Rank: 4 / 5 (1)

  How about a look ahead buffer/ branch predictor that will cache/download (part) of the links on your webpage before clicking/while you are still reading, for near instantenious display. maybe its already being done, i dont know...

  
  
  I think this might be why Google Chrome is already so fast. LOL
  
  • report abuse
  • • Current rank
    • 1
    • 2
    • 3
    • 4
    • 5
• keijo - Nov 17, 2009
  • Rank: 5 / 5 (1)

  How about a look ahead buffer/ branch predictor that will cache/download (part) of the links on your webpage before clicking/while you are still reading, for near instantenious display. maybe its already being done, i dont know...

  
  
  As far as i can remember firefox had a plug-in that would do just that. But I don't think it did much predicting, it just loaded all the links from the page.
  
  • report abuse
  • • Current rank
    • 1
    • 2
    • 3
    • 4
    • 5
• Hunnter - Nov 17, 2009
  • Rank: not rated yet

  How about a look ahead buffer/ branch predictor that will cache/download (part) of the links on your webpage before clicking/while you are still reading, for near instantenious display. maybe its already being done, i dont know...

  
  Actually, this is an option in Chrome from the very beginning if i remember correct.
  DNS-prefetching in the Under The Hood tab in Options.
  
  Glad they made this a very visually findable option, very useful if you are on the move and don't want prefetching on whatever probably limited connection you are using.
  • report abuse
  • • Current rank
    • 1
    • 2
    • 3
    • 4
    • 5