4 ms·

I agree with slow disk I/O, but how do you think Netflix is doing so well with EC2? Adding more machines than they need to?

by gsharma 16y ago

tt 16y ago
They put in great engineering effort to ensure high availability despite random instance failures. I really like their Chaos Monkey. See: http://techblog.netflix.com/2010/12/5-lessons-weve-learned-using-aws.html http://techblog.netflix.com/2010/12/5-lessons-weve-learned-u...
- gfodor 16y ago
  One could argue this is how all systems should be designed anyway.
- dongle 16y ago
  It's cool that Netflix pulled that off, but it seems like a company (like Dotcloud) that could prevent you from building infrastructure on top of infrastructure would be a boon.
tedjdziuba 16y ago
I don't have any data points about Netflix's migration other than their public statements about it, which undoubtedly would conceal any major issues they are struggling with. From what they have said publicly: - They had to re-architect their entire system To me, this seems like a poor business decision. I don't know the specifics of their CapEx, but it seems unlikely that the money saved by cotenancy at Amazon is worth the risk of changing known-working, business critical code. - Machine failure is such a strong expectation that they have a system that purposefully brings machines down. While I applaud the pedantic reasoning behind this, I suspect it's driven more by technical masturbation than business needs.
cpeterso 16y ago
As a high-profile EC2 client, I imagine Amazon gives Netflix tons of "complementary" support. A Netflix failure on EC2 would be a big black eye to Amazon PR.
lsc 16y ago
the outages@outages.org mailing list says that Netflix was down today. I don't know if it's ec2 related or not... I was actually hoping someone might have details. https://puck.nether.net/pipermail/outages/2011-March/002786.html https://puck.nether.net/pipermail/outages/2011-March/002786....

I agree with slow disk I/O, but how do you think Netflix is doing so well with EC2? Adding more machines than they need to?

by gsharma 16y ago

I agree with slow disk I/O, but how do you think Netflix is doing so well with EC2? Adding more machines than they need to?

tt 16y ago

They put in great engineering effort to ensure high availability despite random instance failures. I really like their Chaos Monkey. See: http://techblog.netflix.com/2010/12/5-lessons-weve-learned-using-aws.html http://techblog.netflix.com/2010/12/5-lessons-weve-learned-u...

gfodor 16y ago
One could argue this is how all systems should be designed anyway.
dongle 16y ago
It's cool that Netflix pulled that off, but it seems like a company (like Dotcloud) that could prevent you from building infrastructure on top of infrastructure would be a boon.

tedjdziuba 16y ago

I don't have any data points about Netflix's migration other than their public statements about it, which undoubtedly would conceal any major issues they are struggling with. From what they have said publicly: - They had to re-architect their entire system To me, this seems like a poor business decision. I don't know the specifics of their CapEx, but it seems unlikely that the money saved by cotenancy at Amazon is worth the risk of changing known-working, business critical code. - Machine failure is such a strong expectation that they have a system that purposefully brings machines down. While I applaud the pedantic reasoning behind this, I suspect it's driven more by technical masturbation than business needs.

cpeterso 16y ago

As a high-profile EC2 client, I imagine Amazon gives Netflix tons of "complementary" support. A Netflix failure on EC2 would be a big black eye to Amazon PR.

lsc 16y ago

the outages@outages.org mailing list says that Netflix was down today. I don't know if it's ec2 related or not... I was actually hoping someone might have details. https://puck.nether.net/pipermail/outages/2011-March/002786.html https://puck.nether.net/pipermail/outages/2011-March/002786....