Matsya - Geeknight April 2016

Slides of the talk I gave at Chennai Geeknight 2016.

Video of the presentation - https://www.youtube.com/watch?v=qeBV9JRoTOA

Transcript

1. learnings from running production infra on aws spot reliably j.mp/to-matsya-geeknight
2. Ashwanth Kumar Dev/Ops @ Indix Hindu Mythology Fan OSS Contributor ashwanthkumar.in
3. running hadoop clusters on aws (this is not Amazon EMR)
4. Typical Hadoop Slaves Setup on AWS An Auto Scaling Group (ASG) based, set of instances running TaskTrackers and DataNodes Scaling up or down is as easy as updating the Desired value (or) You can also do Auto scaling based on specific metrics from CloudWatch
5. By default you would spin up On Demand instances on AWS If you’re running 100 c3.2xlarge (15G memory, 8 cores) instances, you’re spending 100 * 0.420 = $42 per hour 100 * 0.420 * 24 = $1008 per day Problem 1 - Cost of On Demand Instances
6. use spot instances (wherever applicable)
7. AWS leases unused hardware at a lower cost as Spot instances Spot Prices are highly volatile But, highly cost effective if used right Spot’s “Demand vs Supply” is local to it’s Spot Market AWS Spot Primer
8. Problem 2 - Spot Outages Using spot is a high risk and high reward game While you get the best value in terms of cost of the same compute, there’s no guarantee on when the machines are going to be taken away There’s Spot Termination Notice, but unfortunately not all applications are AWS- aware like Hadoop in our case
9. Split the spot instances across availability zones
10. For a cluster, the spot markets can be viewed in the following dimensions • # of Instance Types, Regions and Availability Zones The number of spot markets is a cartesian product of all the above numbers. Example - Requirement for 36 CPUs per instance • Instance Types - [d2.8xlarge, c4.8xlarge] • AZs - [us-east-1a, us-east-1b, us-east-1c, …] • Region - [us-east, us-west, …] - 10 regions • Total in US-EAST (alone) => 2 * 1 * 5 = 10 spot markets AWS Spot Markets
11. We had this running for a while with good success We saw our AWS bill was gradually increasing on “Data Transfer” section Because HDFS write pipeline is not very AWS-Cross-AZ-Data-Transfer-Cost aware With HDFS on Spot, each machine going down meant replication will kick in from start Not only HDFS but in a MR job, reducers download data from each mappers Problem 3 - Cost of Data Transfer
12. Go back to single availability zone Find a way around Spot outages within a AZ
13. github.com/ind9/matsya Optimize for cost and keep the fleet running
14. You always need the full fleet (like Hadoop / Mesos / YARN) running 24x7 Ability to fallback to On Demand if needed to save costs Switch back to Spot once the surge ends Use Cases
15. Scala app that monitors spot prices and moves the ASG to cheapest AZ Meant to be run as a CRON task Can fallback to OD (if required) Posts notifications to Slack when migrating Matsya
16. How Matsya Works?
17. How Matsya Works? ASG
18. How Matsya Works? Spot ASG
19. How Matsya Works? us-east-1a us-east-1c ... Spot ASG
20. How Matsya Works? us-east-1a us-east-1c ... Spot ASG
21. How Matsya Works? OD ASG us-east-1a us-east-1c ... Spot ASG (Optional)
22. matsya { working-dir = “local_run” slack-webhook = “http://hooks.slack.com/services/foo/bar/baz” clusters = [{ name = “Staging Hadoop Cluster” spot-asg = “as-hadoop-staging-spot” od-asg = “as-hadoop-staging-od” machine-type = “c3.2xlarge” bid-price = 0.420 od-price = 0.420 max-threshold = 0.90 nr-of-times = 3 fallback-to-od = false subnets = { “us-east-1a” = “subnet-east-1a” “us-east-1b” = “subnet-east-1b” “us-east-1c” = “subnet-east-1c” } }] } Sample Configuration
23. Deployed across all production clusters for 6 months now Along with Vamana, enabled us to achieve • ~40% reduction in monthly AWS bill • ~50% of AWS Infrastructure is on Spot • 100% of Hadoop MR workloads are on Spot Matsya at Indix
24. Matsya at Indix
25. Support for other Spot products - Spot Fleet and Spot Blocks More notification systems Multiple Region support Multiple Product support Minimum number of OD instances Work In Progress Questions? github.com/ind9/matsya
26. http://serverfault.com/questions/448746/ec2-auto-scaling-with-spot-and-on-demand-instances http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/using-regions-availability-zones.html https://aws.amazon.com/ec2/spot/bid-advisor/ References