3 ms·

I think Google Compute Engine does the right thing here: by default it uses "persistent disks" (network-attached redundant/highly-available block devices) for a

by kentonv 11y ago

I think Google Compute Engine does the right thing here: by default it uses "persistent disks" (network-attached redundant/highly-available block devices) for all disks. The only case I've heard of where persistent disk data was lost was a few acknowledged writes occurring just before an unusual lightning-induced power outage: https://status.cloud.google.com/incident/compute/15056 https://status.cloud.google.com/incident/compute/15056 For added protection, you can take regular snapshots. You only pay to store the diff from the last snapshot (so go ahead and snapshot often), and snapshot storage is geographically distributed. (Note: I have no idea what EC2 does, maybe it's similar.)

tshtf 11y ago
Elastic storage on AWS has a fairly low AFR (https://aws.amazon.com/ebs/details/ https://aws.amazon.com/ebs/details/): Amazon EBS volumes are designed for an annual failure rate (AFR) of between 0.1% - 0.2%, where failure refers to a complete or partial loss of the volume, depending on the size and performance of the volume. This makes EBS volumes 20 times more reliable than typical commodity disk drives, which fail with an AFR of around 4%. EBS also supports a snapshot feature, which is a good way to take point-in-time backups of your data.
mwcampbell 11y ago
I've been suspicious of network-attached block storage ever since the Amazon EBS cascading failure of April 2011 and these three blog posts that Joyent did in response: https://www.joyent.com/blog/on-cascading-failures-and-amazons-elastic-block-store https://www.joyent.com/blog/on-cascading-failures-and-amazon... https://www.joyent.com/blog/magical-block-store-when-abstractions-fail-us/ https://www.joyent.com/blog/magical-block-store-when-abstrac... https://www.joyent.com/blog/network-storage-in-the-cloud-delicious-but-deadly https://www.joyent.com/blog/network-storage-in-the-cloud-del... Besides Joyent, I think DigitalOcean and Vultr do it right, with RAID-backed local storage and a backup option.
- kentonv 11y ago
  I'm having trouble unpacking useful information out of these rambly blog posts. It sounds like the basic complaint is that abstractions add complexity, which is more trouble when it goes wrong. I agree that if you have a large-scale sophisticated operation you will probably want to choose local disks and handle availability your own way (and GCE provides those options). But for small-scale operations that can't justify as much engineering, abstractions like persistent disks and automatic migration will save a ton of time and avoid data loss. Ironically, Digital Ocean seems to target this smaller scale, yet they've set the wrong defaults for their target market.

by kentonv 11y ago

tshtf 11y ago

Elastic storage on AWS has a fairly low AFR (https://aws.amazon.com/ebs/details/ https://aws.amazon.com/ebs/details/): Amazon EBS volumes are designed for an annual failure rate (AFR) of between 0.1% - 0.2%, where failure refers to a complete or partial loss of the volume, depending on the size and performance of the volume. This makes EBS volumes 20 times more reliable than typical commodity disk drives, which fail with an AFR of around 4%. EBS also supports a snapshot feature, which is a good way to take point-in-time backups of your data.

mwcampbell 11y ago

I've been suspicious of network-attached block storage ever since the Amazon EBS cascading failure of April 2011 and these three blog posts that Joyent did in response: https://www.joyent.com/blog/on-cascading-failures-and-amazons-elastic-block-store https://www.joyent.com/blog/on-cascading-failures-and-amazon... https://www.joyent.com/blog/magical-block-store-when-abstractions-fail-us/ https://www.joyent.com/blog/magical-block-store-when-abstrac... https://www.joyent.com/blog/network-storage-in-the-cloud-delicious-but-deadly https://www.joyent.com/blog/network-storage-in-the-cloud-del... Besides Joyent, I think DigitalOcean and Vultr do it right, with RAID-backed local storage and a backup option.

kentonv 11y ago
I'm having trouble unpacking useful information out of these rambly blog posts. It sounds like the basic complaint is that abstractions add complexity, which is more trouble when it goes wrong. I agree that if you have a large-scale sophisticated operation you will probably want to choose local disks and handle availability your own way (and GCE provides those options). But for small-scale operations that can't justify as much engineering, abstractions like persistent disks and automatic migration will save a ton of time and avoid data loss. Ironically, Digital Ocean seems to target this smaller scale, yet they've set the wrong defaults for their target market.