Introduction
>> How about common backend, and as I understand it you have billions and billions of objects on that backend. How do you manage to keep that scaled in terms of it's growing? Are you using a cloud at all? In other words, using external providers to provide you with storage and compute power?
>> Sure, yeah. As you mentioned, we are now rapidly approaching 10 billion items in our container system - that's photos, videos, text items that people are sharing around. A lot of people use the word cloud to mean a lot of different things. We use edge caching through the optimize service to host all of our... to really just be an edge cache on front of our photo storage.
>> So it basically means you're going to get it delivered faster to the end user?
>> That's right. They have a really strong global presence here - Asia, Europe, all of the different places where our service is big, and a request goes first to them to us, and they cache it for as long as it takes; and our miss rate on that is really below 1 percent, that we'll actually get a call in later that we've already served to them earlier.
>> And in terms of the data centers you're using, how are you configuring them? How many servers do you run?
>> You know it's actually impressive what we're able to do on a fairly... on a relatively small amount of servers. We're at about 40 database machines, and a couple 100 1U boxes for doing our... our business logic side. The way we've managed to do that is actually to really have purpose built servers. So we have 1 particular server just for serving up X amount to all of our different slideshowages. That allows us to scale as needed per service, because what we see is there is spurts of growth... growth with each application, so it may be that we find some new truth in the... how do people express themselves with photos, and there's a spike in the amount of image uploads that we see. So our uploading service is separated from our XML service, and by making those services on commodity low end hardware, we're able to scale those up according to demand.
==== Transcribed by Automatic Sync Technologies ====
