Live data from Hacker News

Dropbox’s Exodus from the Amazon Cloud

wired.com

111–120 of 252 posts

Re: Dropbox’s Exodus from the Amazon Cloud

#111

The word "cloud" loses all meaning in this article. "The irony is that in fleeing the cloud, Dropbox is showing why the cloud is so powerful. It too is building infrastructure so that others don’t have to. It too is, well, a cloud company." Wait ... so using AWS is "cloud", having your own servers is "cloud" too. Everything is cloudy!

Remember, there is no cloud, there is just someone else's computer... =)

Re: Dropbox’s Exodus from the Amazon Cloud

#112
post #102
post #2

Hi HN! A couple of us from the Magic Pocket software team are around to answer questions if anyone has some.

Does magic pocket use any non-standard error-correction algorithms, or just parity-style RAID5 or 6?

We use a variant on Reed-Solomon coding that's optimized for lower reconstruction cost. This is similar to Local Reconstruction Codes but a design/implementation of our own.

The data placement isn't RAID. We encode aggregated extents of data in "volumes" that are placed on a random set of storage nodes (with sufficient physical diversity and various other constraints). Each storage node might hold a few thousand volumes, but the placement for each volume is independent of the others on the disk. If one disk fails we can thus reconstruct those volumes from hundreds of other disks simultaneously, unlike in RAID where you'd be limited in IOPS and network bandwidth to a fixed set of disks in the RAID array.

That probably sounded confusing but it's a good topic for a blog post once we get around to it.

Re: Dropbox’s Exodus from the Amazon Cloud

#114

>> Transferring four petabytes of data, it turned out, took about a day. 46GB/s, is my math right?

We were pushing over a terabit of data transfer at peak, so 4-5PB per day.

Btw. that was a typo, I meant over "half" a terabit of data transfer. You get into bigger numbers than this when you start talking about aggregate traffic.

Re: Dropbox’s Exodus from the Amazon Cloud

#115
post #109
post #25

Earlier quoted context omitted.

Good questions, let me try to tackle them one by one. > The article makes a brief mention of Go causing issues with RAM usage. Was this due to large heap usage, or was it a problem of GC pressure/throughput/latency? If the former, what were some of the core problems that could not be further optimized in Go? The reasons for using rust were many, but memory was one of them. Primarily, for this particular project, the…

How do you do network io with rust? Thread-per-connection, non-blocking (using mio or?), or something else?

We have an in-house futures-based framework (inspired by Finagle) built on mio (non-blocking libevent like thing for rust). All I/O is async, but application work is often done on thread pools. Those threads are freed as soon as possible, though, so that I/O streams can be handled purely by "the reactor", and we keep the pools as small as possible.

Re: Dropbox’s Exodus from the Amazon Cloud

#116
post #86

I want to point out one other thing about this project, if only to assuage my guilt. And this community of hackers and entrepreneurs seems as good a place as any to clear the air. There is a necessary abridgement that happens in media like this, wherein a few individuals in lead roles act as a vignette for the entire effort. In particular, on the software team, James and I are highlighted here, and it would be easy t…

What does SRE mean?

Re: Dropbox’s Exodus from the Amazon Cloud

#117
post #86

I want to point out one other thing about this project, if only to assuage my guilt. And this community of hackers and entrepreneurs seems as good a place as any to clear the air. There is a necessary abridgement that happens in media like this, wherein a few individuals in lead roles act as a vignette for the entire effort. In particular, on the software team, James and I are highlighted here, and it would be easy t…

I think most of us figured there'd be more people involved, but it shows lots of class to come post a comment giving credit.

Re: Dropbox’s Exodus from the Amazon Cloud

#118
post #86

I want to point out one other thing about this project, if only to assuage my guilt. And this community of hackers and entrepreneurs seems as good a place as any to clear the air. There is a necessary abridgement that happens in media like this, wherein a few individuals in lead roles act as a vignette for the entire effort. In particular, on the software team, James and I are highlighted here, and it would be easy t…

What does SRE mean?

I assume Site Reliability Engineer from the last time I looked at their openings. I didn't look at the description, but I assume it's a more formal description for devops?

Re: Dropbox’s Exodus from the Amazon Cloud

#119
post #89

Earlier quoted context omitted.

Rarely do people "get" that if you have a consistent workload or resource need, AWS is more expensive than your own gear.

But presumably Amazon benefits (saves money) from making more use of hardware resources. It has to have some spare, but not as much spare as a comparable group of non-AWS users would have in total.... Also, presumably, not all the AWS users are awake at the same time ;-)

> But presumably Amazon benefits (saves money) from making more use of hardware resources

yeah, they sure do.

Re: Dropbox’s Exodus from the Amazon Cloud

#120
post #86

I want to point out one other thing about this project, if only to assuage my guilt. And this community of hackers and entrepreneurs seems as good a place as any to clear the air. There is a necessary abridgement that happens in media like this, wherein a few individuals in lead roles act as a vignette for the entire effort. In particular, on the software team, James and I are highlighted here, and it would be easy t…

What does SRE mean?

http://googleforstudents.blogspot.com/2012/06/site-reliabili...
Post reply on HN