Re: Google maps / Unresponsiveness

From: Stefano Mazzocchi <>
Date: Fri, 24 Jun 2005 11:13:11 -0400

Oliver Cole wrote:
> On Fri, 2005-06-24 at 15:22, Stefano Mazzocchi wrote:
>>you are too kind :-) (but, yes, our egos like these comments a lot ;-)
> ;-)

eheh :-)

>>I personally fear (since I'm also the one doing the sysadmin on that
>>machine ;-) of having all requests go to the simile server, it would be
>>pretty easy to bring it down to its knees if PB keeps taking up like it has.
> How is it doing? I'm surprised I haven't seen this on Slashdot!

slashdot rejected it :-) (probably because they already linked us when
we released Welkin and they don't think it's that cool anymore... oh
well, it made it to which, believe me, generates
more load than slashdot these days)

Take a look at

after almost a month, the downloads are pretty steady. What's
interesting is to compare the slashdot load of web pages over the one
generated by PB. Very different shape, but the total number of traffic
generated is a lot more... well also because PB is far more appealing
than Welkin for end users.

> Have you had any non-technical users?

yes, but PB feels too 'beta' for them. They try it out, they uninstall
it and wait for a more solid release (faster and crisper and one that
doesn't slow their firefox to their knees). We are fully aware of this
and we'll make all possible efforts to make sure this happens. We
believe in PB and want to make it happen for the general public, means
that it won't be just another cool yet unusable research platform.

>>>>That would provide a central place to update it, and could even allow
>>>>you to provide a GUI warning that the mapping is currently broken while
>>>>you work out the new version.
>>Well, that is a better idea. We could have PB check the web site once a
>>day for "update needed!" and provide a sort-of modular patch way for PB
>>to update itself without requiring a restart (this is possible in things
>>that don't rely on java code directly).
> My first thought was to hash the profile ID to produce a time of day at
> which the installation will check for updates, in order to spread the
> load, but this will of course mean that some don't get updated.
> Spreading the load on updates in that way is an interesting problem in
> itself.

well, some have argued that the best way to do that is to use the web
architecture itself: meaning to have a load-balancing DNS and tons of
machines. This is what google does. We just don't have the monetary
resources to achieve a similar thing, so we need to think of something else.

> On the other hand... I imagine all we need to update is a couple of
> functions? That shouldn't require massive amounts of bandwidth.

Exactly. If we can keep the 'patches' small enough, even having a few
distribution servers shouldn't create that big of a problem.

>>The downside of this is that it makes security of our servers a huge
>>deal (I can't even imagine somebody putting up a trojan in there!) but I
>>guess that we have the same problem with the XPI anyway.
> Well... mozdev will certainly provide another hosting platform. If you
> can collect a few more mirrors like that that you can trust, eg, a
> Sourceforge page, then you are home dry.

We would like to avoid that. Bandwitdh is not really an issue here at
MIT, it's more sysadmin resources. But don't worry, we just bought a new
beefy server to spread the load, so we are preparing for the next wave.

>>>>I also wanted to ask about the slightly adverse affect the PB extension
>>>>has had on my Firefox.
>>I thought we had an issue in JIRA, but I can't find it, so I created a
>>new one. Add your comments/suggestions to that one if you want.
> Have added your bit about Deerpark. I guess we have to wait and see. (Or
> install Deerpark... but I don't have time for that.)

I plan to work more on this issue next week. I'll post my results here.

Stefano Mazzocchi
Research Scientist                 Digital Libraries Research Group
Massachusetts Institute of Technology            location: E25-131C
77 Massachusetts Ave                   telephone: +1 (617) 253-1096
Cambridge, MA  02139-4307              email: stefanom at mit . edu
Received on Fri Jun 24 2005 - 15:11:00 EDT

This archive was generated by hypermail 2.3.0 : Thu Aug 09 2012 - 16:39:18 EDT