Owen DeLong wrote:
>
> Actually, if the content provider simply honors MEDs, that should cover most
> of the issue. Then, the long haul is done across the content providers'
> backbone anyway.
Then you get into address aggregation issues that have already been
discussed before.
The way I see it, an intelligent content distribution scheme addresses the
vast majority of these concerns.
So, a question to the large web farms. Why is it that the largest web sites
still seem to be hosted out of a single data center?
Where do you get your data? It seems to me that the bulk of the largest web
sites with which I am familiar are located in at least two datacenters.
Alec
Owen
Senior Network Engineer
Exodus
Here's the issue. Customers like the idea of distributed architecture, but
are reluctant to do it. We've (FGC) developed an intelligent DNS product
that works like a Cisco Distributed director, and can offer it for free to
our customers. It does not cost customers anything additional to be in
two sites, but they do not want the management hassles.
One of the main issues for large sites is database replication. For
example. If you have a search engine in both Herndon VA, and Sunnyvale,
CA, the amount of cross country bandwidth to reliably do database
replication in real time is huge. Therefore it is easier for that site to
stay in one location. If you have a static site with no database,
chances are they are not large enough to need a distributed architecture.
Jason
Then you get into address aggregation issues that have already been
discussed before.
The way I see it, an intelligent content distribution scheme addresses the
vast majority of these concerns.
So, a question to the large web farms. Why is it that the largest web
sites
Owen DeLong wrote:
Where do you get your data? It seems to me that the bulk of the largest web
sites with which I am familiar are located in at least two datacenters.
Most of the busy sites I frequent are hosted on the west coast (not just
Exodus-hosted sites, but sites in general).
Alec
That's just how the internet plays, there are probably more sights
phisically in the silicon valley then anywhere else on the west coast. It's
where the busniness's are, and since they want to have their machines as
close to them as possible, they put them here.
On the other hand, the customers who view the sites are spread out
all over the country (and world). Therefore replication is a good thing
for these businesses. With your servers located all around the country
(close to private or public peering points) you will get better performance.
Of course if there are no public or private peering points located near
the site, the value would be close to nil, since distribution needs local
inbound traffic to be worthwhile.
steve@altrina.exodus.net wrote:
That's just how the internet plays, there are probably more sights
phisically in the silicon valley then anywhere else on the west coast. It's
where the busniness's are, and since they want to have their machines as
close to them as possible, they put them here.
There's some interesting logic. Don't get me wrong, I understand it (from
the customer's point of view). However, I think that it would be a good
idea for web farms to start to make it worth the customer's while to
distribute their servers. It may not be much of a network cost savings from
the web farm's point of view, but it does address some of the traffic
asymmetry issues, on top of improving performance to the end user.
On the other hand, the customers who view the sites are spread out
all over the country (and world). Therefore replication is a good thing
for these businesses. With your servers located all around the country
(close to private or public peering points) you will get better performance.
Agreed.
Of course if there are no public or private peering points located near
the site, the value would be close to nil, since distribution needs local
inbound traffic to be worthwhile.
Well, in the world of private interconnects there are private interconnect
points all over the place. Also, you can always place your colocation
facilities near the public interconnect points.
Steven O. Noble -- Sr. Backbone Engineer, Exodus Communications (EXDS)
Alec