V******y
发帖数: 53 | 1 今天收到HR的team match邮件，有人可以大概说一下哪个组比较有前途吗？大组是UFO
UFO Accelerated Cluster Turn-up (ACT)
When Google builds data centers there is a lot of manual work that goes into
the planning. The purpose of ACT is to implement a process using automation
and new workflow systems to decrease the total time from planning stage to
bringing up clusters down to one-quarter the current time, or less. Right
now it's a catch up game because when we are rolling out new technologies
the manual side is still working on the physical part.
It is new technology every year and new ways to manage the network. You will
be coming in on the ground floor on the new process of how we bring up data
centers -- essentially warehouse-sized computers. These new technologies
are on-going and will be new each year.
In terms of impact, this work is in the context of a key initiative within
Google's technical infrastructure that will change the way Google Data
Centers are developed, deployed and managed. Engineers will be able to drive
cross functionally development delivering a solution that will save Google
millions of dollars, thus having the opportunity to make a big impact on the
business. For example our software engineers have saved hundreds of
megawatts of power reducing the need for new data center development. You
have the chance to be part of a new initiative driving technology for both
Google and the world.
In terms of actual work, when Google deploys infrastructure into datacenters
there is a lot of work that goes into planning the logical, physical and
material aspects of the datacenter. Today, a bunch of ad-hoc spreadsheets
and rudimentary tools are used to do all of this planning over the course of
several months. The purpose of this team is to make planning possible
within a few hours. This is all the more challenging because Google
datacenter technology changes very often. The key things we expect SWEs to
do is to be able to design and implement:
• authoritative datastores of logical, physical and material
datacenter models, which must be as future-proof as possible
• complex distributed systems that automate the planning as much as
possible
• user interfaces to help configure the datacenter as well visibility
into the progress of automation executing those high-level configurations.
Cloud Capacity
Cloud Capacity: Creating a planning system for planning Google Data
Centers: resources needed for different cloud products, ordering, delivering
to right teams, helping with constraint checks and budget reviews, making
sure things fit in the data centers where they are scheduled to go. Making
sure they have what they need, when they need it, without having too much
excess/waste.
Team impact & perks:
• Being part of building Google Cloud. Industry is moving to
Cloud and this team will be part of making the transition a reality.
• The development environment is fast paced and light weight. They
have fast turnaround times (completing code to pushing it out).
• Direct interaction with their customers
• Freedom to impact direction of team
• Positive team environment
• Can sit in SVL or SF
RIOT (Resource and Infrastructure Optimization Technology) Capacity Planning
and Delivery
• Team is developing and delivering the fleet for Google.
• This org manages all planning and deployment aspects of Google's
massive multi-billion dollar data center fleet. The role responsibilities
include design and development of software services to make intelligent
decisions and automate rapid expansion and transformation of the fleet.
• How does this benefit Google? Development of this infrastructure
will enable faster TTM of new Google Cloud services and capabilities such as
in ML.
• Very behind the scenes Software Engineering (no customer visibility).
RIOT Continuous Fleet Transformation
How do Alphabet's product areas like Search, YouTube, and Gmail support
their 1-billion+ users? How does Blobstore store X exabytes of data?
Answer: Google's data centers, whose massive fleet of machines, storage,
network switches and more are managed by RIOT software systems.
Come join RIOT's Continuous Fleet Transformation team (Fleet Optimizer,
Fleet Transformation Service, Twister, gSpace, and others), which creates
software to manage and optimize design, purchase, deployment, and
decommissioning for Google data center resources.
The fleet uses as much power as the entire city of San Francisco, so that’s
a lot of impact to the environment and cost to Google to minimize if we do
it right!
Making these decisions isn’t easy; the quantity and varying quality of
decision-making data and the overall scale make this a challenge for humans
to get right. Our software helps us explore the tradeoffs and keeps humans
in control through simple but powerful scenario exploration and compelling
visualization.
Help realize the vision of "autonomous, intelligent infrastructure" to
continuously plan and re-shape resource capacity based on statistical models
and in response to ever changing workloads in every cluster in Google's
fleet.
本人有一年backend背景和一年data/ml. 真诚求建议！
谢谢 | n***h
发帖数: 51 | | t**********n
发帖数: 1718 | | G******8
发帖数: 1 | | k***a
发帖数: 1199 | 5 all shitholes?
UFO
into
automation
to
will
data
drive
Google
the
datacenters
of
visibility
delivering
Planning
as
visibility).
’s
do
humans
models
【在 V******y 的大作中提到】
: 今天收到HR的team match邮件，有人可以大概说一下哪个组比较有前途吗？大组是UFO
: UFO Accelerated Cluster Turn-up (ACT)
: When Google builds data centers there is a lot of manual work that goes into
: the planning. The purpose of ACT is to implement a process using automation
: and new workflow systems to decrease the total time from planning stage to
: bringing up clusters down to one-quarter the current time, or less. Right
: now it's a catch up game because when we are rolling out new technologies
: the manual side is still working on the physical part.
: It is new technology every year and new ways to manage the network. You will
: be coming in on the ground floor on the new process of how we bring up data
| V******y
发帖数: 53 | 6 真的假的？如果不好的话我叫HR再换几个team聊聊？ |
|
