AWS Re:invent announcements day 2
Other Re:invent 2021 content:
- Full list of announcements with links to docs and blog posts
- Day 1 highlights
- Day 2 highlights
- Day 3 highlights
- Day 4 highlights
I’m back at it again with the most interesting announcements from the second day of AWS Re:invent:
Serverless Data Tools
These all got announced on one slide during Adam Selipsky’s keynote:
- Redshift Serverless (blog post)
- EMR Serverless (blog post)
- MSK Serverless
- Kinesis “On-Demand” (blog post)
I’m curious how pricing for these services look compared to our current set up (particularly with Redshift 3 year RIs), but more services with simplified operations and scaling is always welcome.
A year after announcing Graviton2, AWS has launched their next generation ARM instance, the Graviton3 — starting with the c7g instance class. According to AWS, these instances:
provide up to 25% better compute performance, up to 2x higher floating-point performance, and up to 2x faster cryptographic workload performance compared to AWS Graviton2 processors. Graviton3 processors deliver up to 3x better performance compared to Graviton2 processors for CPU-based machine learning workloads, with support for bfloat16 and fp16 instructions. Graviton3 processors also support pointer authentication for enhanced security in addition to always-on 256-bit memory encryption available in AWS Graviton2.
Liz Fong-Jones has been running them at Honeycomb for a while now:
+35% performance increase over #Graviton2. We were able to run 30% fewer instances of C7g than C6g serving same workload, and with 30% reduced latency.
These are available in preview now.
A bunch of s3 announcements dropped during the evening yesterday. A couple of my favorites:
This seems like a huge deal — you can now centralize your backup management of S3 buckets, in addition to RDS, DynamoDB, EFS, etc.
Initial backups are a full snapshot of the bucket, and then it will track incremental changes. You are able to restore the entire bucket or specific objects into a new or existing S3 bucket.
To enroll in the Preview, send your account ID to firstname.lastname@example.org.
S3 was the first AWS service launched, which means that it pre-dates AWS IAM. This has meant that it’s had a strange brew of its own access controls (known as S3 ACLs) and IAM access control. For the first time, we can now disable ACLs on a bucket level — hopefully account-level and org-level support are coming soon!
You can now retrieve files from Glacier instantly by using this new storage class. Along with this, the old Glacier class becomes Glacier Flexible Retrieval with 10% cheaper storage and free bulk retrievals. S3 Intelligent Tiering supports the new Instant Retrieval storage class for up to 68% savings.
Late night lambda pre-announcement
There was a definitely lack of Lambda news yesterday — I’m hoping for some bigger announcements later in the week (maybe during Werner’s keynote). However Julian Wood posted a nice improvement for Lambda that should be coming soon — a 20x increase in available ephemeral storage (/tmp) of up to 10GB.