00:00.0
00:03.0
After sustential research, the detection team is ready to advance.
00:03.0
00:07.4
All we need to strengthen our research capabilities
00:07.4
00:11.7
is just a frozen index so we can conduct retrospective research.
00:11.7
00:15.9
It will greatly speed the cadance™ and improve the precision.
00:15.9
00:20.9
Alright, so just spin up a new ElasticSearch cluster.
00:30.0
00:33.7
Spinning up a new cluster is not that simple.
00:33.7
00:36.1
It's a 3 month project.
00:53.0
00:59.0
Everyone who's been in the company for more than two years, stay. Everybody else, leave now.
01:13.1
01:14.5
What is this!?
01:14.5
01:17.4
I can't comprehend what you're saying.
01:17.5
01:23.6
We are an org of 100 engineers, we have been using ElasticSearch in production for years!
01:23.6
01:27.6
Almost every team has a dedicated cluster
01:27.6
01:31.4
which has a full freaking copy of the entire dataset of the company!
01:31.4
01:33.9
We have split the clusters so issues won't impact cross-team workloads
01:33.9
01:38.4
But still dev-velocity is almost unusable
01:38.4
01:41.0
It's red if we are not super duper careful.
01:41.2
01:43.0
we're doing our best to monitor and --
01:43.0
01:46.1
Monitoring is not enough! The clusters are just too loaded
01:46.1
01:47.0
There's a new
01:47.0
01:48.5
cluster, ES-Analytics
01:48.5
01:51.9
It's only used for Sigmund and it crashes every other day. 170 node cluster!
01:51.9
01:54.2
It's too much!
01:54.2
02:01.6
We need to be able to spin up clusters more easily! Obviously they are too loaded. Everyone needs a cluster
02:01.6
02:03.7
So just have some decent infrastructure
02:03.7
02:08.4
so we could click to deploy a cluster for anyone who requires it.
02:08.4
02:14.9
Without it we would never be able to proceed with our deeper research needs,
02:14.9
02:16.7
our competitors will outsmart us.
02:16.7
02:21.8
Velocity is the key selling point of this company
02:27.7
02:29.5
We have outstanding researchers here
02:29.5
02:31.8
the engineers are fine also.
02:31.8
02:34.3
But first we are human beings.
02:34.4
02:36.6
We need to have quality tools to be able to work.
02:41.4
02:43.0
The alternative of just creating a new cluster
02:43.0
02:47.9
is dead on arrival. Precision Eng don't have the capacity to handle it
02:47.9
02:53.0
and other teams don't have the knowhow.
02:53.0
02:56.0
Making RE do it without proper infrastracture is a suicide mission
02:56.1
02:59.4
There's no chance they'll be able to handle it. It's a cluster 5x big compared to ES-Analytics
02:59.4
03:02.8
It'll comsume all of their time.
03:05.2
03:05.5
We will be fine.
03:05.5
03:07.9
We're hiring people to help... we just can't find them
03:14.8
03:16.7
There's no choice.
03:16.7
03:23.7
We will just have to invest the time.
03:23.7
03:26.9
FPGAs are a long, long way ahead.
03:32.0
03:35.6
We'll have to double down on Elastic.
03:41.0
03:47.3
We'll just have to make sure this infra project is better managed. Agile.
03:47.3
03:48.9
To the point.
03:54.1
03:55.6
So it wouldn't end up like past infra efforts.
