February 23, 2024

Etsy, a web-based market for distinctive, handmade, and classic gadgets, has
seen excessive progress over the past 5 years. Then the pandemic dramatically
modified customers’ habits, resulting in extra shoppers buying on-line. As a
outcome, the Etsy market grew from 45.7 million patrons on the finish of
2019 to 90.1 million patrons (97%) on the finish of 2021 and from 2.5 to five.3
million (112%) sellers in the identical interval.

The expansion massively elevated demand on the technical platform, scaling
visitors virtually 3X in a single day. And Etsy had signifcantly extra prospects for
whom it wanted to proceed delivering nice experiences. To maintain up with
that demand, they needed to scale up infrastructure, product supply, and
expertise drastically. Whereas the expansion challenged groups, the enterprise was by no means
bottlenecked. Etsy’s groups had been in a position to ship new and improved
performance, and {the marketplace} continued to offer a glorious buyer
expertise. This text and the following type the story of Etsy’s scaling technique.

Etsy’s foundational scaling work had began lengthy earlier than the pandemic. In
2017, Mike Fisher joined as CTO. Josh Silverman had lately joined as Etsy’s
CEO, and was establishing institutional self-discipline to usher in a interval of
progress. Mike has a background in scaling high-growth corporations, and alongside
with Martin Abbott wrote a number of books on the subject, together with The Art of Scalability
and Scalability Rules.

Etsy relied on bodily {hardware} in two information facilities, presenting a number of
scaling challenges. With their anticipated progress, it was obvious that the
prices would ramp up shortly. It affected product groups’ agility as they’d
to plan far prematurely for capability. As well as, the info facilities had been
primarily based in a single state, which represented an availability threat. It was clear
they wanted to maneuver onto the cloud shortly. After an evaluation, Mike and
his group selected the Google Cloud Platform (GCP) because the cloud associate and
began to plan a program to maneuver their
many methods onto the cloud.

Whereas the cloud migration was taking place, Etsy was rising its enterprise and
its group. Mike recognized the product supply course of as being one other
potential scaling bottleneck. The autonomy afforded to product groups had
brought about a difficulty: every group was delivering in several methods. Becoming a member of a group
meant studying a brand new set of practices, which was problematic as Etsy was
hiring many new individuals. As well as, they’d seen a number of product
initiatives that didn’t repay as anticipated. These indicators led management
to re-evaluate the effectiveness of their product planning and supply
processes.

Strategic Rules

Mike Fisher (CTO) and Keyur Govande (Chief Architect) created the
preliminary cloud migration technique with these ideas:

Minimal viable product – A typical anti-pattern Etsy wished to keep away from
was rebuilding an excessive amount of and prolonging the migration. As a substitute, they used
the lean idea of an MVP to validate as shortly and cheaply as potential
that Etsy’s methods would work within the cloud, and eliminated the dependency on
the info heart.

Native resolution making – Every group could make its personal choices for what
it owns, with oversight from a program group. Etsy’s platform was break up
into various capabilities, similar to compute, observability and ML
infra, together with domain-oriented software stacks similar to search, bid
engine, and notifications. Every group did proof of ideas to develop a
migration plan. The principle market software is a famously giant
monolith, so it required making a cross-team initiative to give attention to it.

No adjustments to the developer expertise – Etsy views a high-quality
developer expertise as core to productiveness and worker happiness. It
was essential that the cloud-based methods continued to offer
capabilities that builders relied upon, similar to quick suggestions and
subtle observability.

There additionally was a deadline related to present contracts for the
information heart that they had been very eager to hit.

Utilizing a associate

To speed up their cloud migration, Etsy wished to convey on outdoors
experience to assist in the adoption of recent tooling and expertise, similar to
Terraform, Kubernetes, and Prometheus. In contrast to quite a lot of Thoughtworks’
typical purchasers, Etsy didn’t have a burning platform driving their
basic want for the engagement. They’re a digital native firm
and had been utilizing a completely fashionable method to software program growth.
Even with no single downside to give attention to although, Etsy knew there was
room for enchancment. So the engagement method was to embed throughout the
platform group. Thoughtworks infrastructure engineers and
technical product managers joined search infrastructure, steady
deployment companies, compute, observability and machine studying
infrastructure groups.

An incremental federated method

The preliminary “elevate &
shift” to the cloud for {the marketplace} monolith was essentially the most troublesome.
The group wished to maintain the monolith intact with minimal adjustments.
Nonetheless, it used a LAMP stack and so could be troublesome to re-platform.
They did various dry runs testing efficiency and capability. Although
the primary cut-over was unsuccessful, they had been in a position to shortly roll
again. In typical Etsy type, the failure was celebrated and used as a
studying alternative. It was finally accomplished in 9 months, much less time
than the total yr initially deliberate. After the preliminary migration, the
monolith was then tweaked and tuned to situate higher within the cloud,
including options ​​like autoscaling and auto-fixing dangerous nodes.

In the meantime, different stacks had been additionally being migrated. Whereas every group
created its personal journey, the groups weren’t fully on their very own.
Etsy used a cross-team structure advisory group to share broader
context, and to assist sample match throughout the corporate. For instance, the
search stack moved onto GKE as a part of the cloud, which took longer than
the elevate and shift operation for the monolith. One other instance is the
information lake migration. Etsy had an on-prem Vertica cluster, which they
moved to Huge Question, altering every part about it within the course of.

Not stunning to Etsy, after the cloud migration the optimization
for the cloud didn’t cease. Every group continued to search for alternatives
to make the most of the cloud to its full extent. With the assistance of the
structure advisory group, they checked out issues similar to: how one can
cut back the quantity of customized code by transferring to industry-standard instruments,
how one can enhance value effectivity and how one can enhance suggestions loops.

Determine 1: Federated
cloud migration

For instance, let’s take a look at the journey of two groups, observability
and ML infra:

The challenges of observing every part

Etsy is legendary for measuring every part, “If it strikes, we monitor it.”
Operational metrics – traces, metrics and logs – are utilized by the total
firm to create worth. Product managers and information analysts leverage the
information for planning and proving the expected worth of an thought. Product
groups use it to assist the uptime and efficiency of their particular person
areas of accountability.

With Etsy’s dedication to hyper-observability, the quantity of knowledge
being analyzed isn’t small. Observability is self-service; every group
will get to determine what it needs to measure. They use 80M metric sequence,
masking the location and supporting infrastructure. This can create 20 TB
of logs a day.

When Etsy initially developed this technique there weren’t quite a lot of
instruments and companies in the marketplace that would deal with their demanding
necessities. In lots of instances, they ended up having to construct their very own
instruments. An instance is StatsD, a stats aggregation device, now open-sourced
and used all through the {industry}. Over time the DevOps motion had
exploded, and the {industry} had caught up. Numerous modern
observability instruments similar to Prometheus appeared. With the cloud
migration, Etsy might assess the market and leverage third-party instruments
to scale back operational value.

The observability stack was the final to maneuver over resulting from its complicated
nature. It required a rebuild, somewhat than a elevate and shift. They’d
relied on giant servers, whereas to effectively use the cloud it ought to
use many smaller servers and simply scale horizontally. They moved giant
components of the stack onto managed companies and third celebration SaaS merchandise.
An instance of this was introducing Lightstep, which they may use to
outsource the tracing processing. It was nonetheless essential to do some
quantity of processing in-house to deal with the distinctive situations that Etsy
depends on.

Migration to the cloud-enabled a greater ML platform

A giant supply of innovation at Etsy is the best way they make the most of their
Machine studying.

Etsy leverages
machine studying (ML) to create customized experiences for our
tens of millions of patrons all over the world with state-of-the-art search, advertisements,
and suggestions. The ML Platform group at Etsy helps our machine
studying experiments by creating and sustaining the technical
infrastructure that Etsy’s ML practitioners depend on to prototype, prepare,
and deploy ML fashions at scale.

Kyle Gallatin and Rob Miles

The transfer to the cloud enabled Etsy to construct a brand new ML platform primarily based
on managed companies that each reduces operational prices and improves the
time from thought era to manufacturing deployment.

As a result of their assets had been within the cloud, they may now depend on
cloud capabilities. They used Dataflow for ETL and Vertex AI for
coaching their fashions. As they noticed success with these instruments, they made
positive to design the platform in order that it was extensible to different instruments. To
make it broadly accessible they adopted industry-standard instruments similar to
TensorFlow and Kubernetes. Etsy’s productiveness in creating and testing
ML leapfrogged their prior efficiency. As Rob and Kyle put it, “We’re
estimating a ~50% discount within the time it takes to go from thought to stay
ML experiment.”

This efficiency progress wasn’t with out its challenges nonetheless. Because the
scale of knowledge grew, so too did the significance of high-performing code.
With low-performing code, the shopper expertise might be impacted, and
so the group needed to produce a system which was extremely optimized.
“Seemingly small inefficiencies similar to non-vectorized code may end up
in a large efficiency degradation, and in some instances we’ve seen that
optimizing a single tensor move rework perform can cut back the mannequin
runtime from 200ms to 4ms.” In numeric phrases, that’s an enchancment of
two orders of magnitude, however in enterprise phrases, it is a change in
efficiency simply perceived by the shopper.

What had been the challenges of the cloud?

Etsy needed to function its personal infrastructure, and quite a lot of the platform
groups’ expertise had been in methods operation. Shifting the cloud allowed groups
to make use of a better abstraction, managed by infrastructure as code. They
modified their infrastructure hiring to search for software program engineering
expertise. It brought about friction with the prevailing group; some individuals had been very
excited however others had been apprehensive concerning the new method.

Whereas the cloud definitely decreased the variety of issues they needed to
handle and allowed for less complicated planning, it didn’t absolutely get them away
from capability planning. The cloud companies nonetheless run on servers with
CPUs and Disks, and in some conditions, there may be right-sizing for future
load that needs to be achieved. Going ahead, as on-demand cloud companies
enhance, Etsy is hopeful they’ll cut back this capability planning.

The stress take a look at of the pandemic

Etsy had all the time been information heart primarily based, which had stored them
constrained in some methods. As a result of they’d been so closely invested in
their information heart presence, they hadn’t been making the most of new
choices cloud distributors had developed. For instance, their information heart
setup lacked strong APIs to handle provisioning and capability.

When Mike Fisher got here onboard, Etsy then started their cloud migration
journey. This set them up for fulfillment for the long run, because the
migration was mainly completed in the beginning of the pandemic. There
had been just a few methods this manifested: they’d no capability crunch, though
visitors exploded 2-3X in a single day, as occasions had elevated from 1 billion
to six billion.

And there have been particular examples of how the cloud gave them agility
throughout the pandemic. For instance, the cloud enabled efforts to shut the
“semantic hole”, guaranteeing searches for “masks” surfaced material masks not
face masks of the beauty or costume selection. This was potential as a result of
Google Cloud enabled Etsy to implement extra subtle machine
studying and the agility to retrain algorithms in actual time. One other
instance was their database administration modified from the datacenter to the
cloud. Particularly, round backups, Etsy’s DR posture improved within the
cloud, since they leveraged block storage snapshotting as a approach of
restoring databases. This enabled them to do quick restores, have
confidence and be capable of take a look at them shortly, in contrast to the older methodology,
the place a restore would take a number of hours and never be completely
scalable.

Etsy performs intensive load and efficiency testing. They use chaos
engineering methods, having a ‘scale day’ that stresses the methods
at max capability. After the pandemic the elevated load was now not a
spike, it was now the day by day common. The load testing structure and
methods wanted to be simply as scalable as every other system with the intention to
deal with the expansion.

Regularly Bettering the platform

One among Etsy’s subsequent focus areas is to create “paved roads” for
engineers. A set of instructed approaches and equipment to scale back
friction when launching and creating companies. In the course of the preliminary 4
years of the cloud migration, they determined to take a really federated
technique. They took the “let 1000 flowers bloom” method as described
by Peter Seibel in his article on engineering effectiveness at
Twitter
.
The methods had by no means existed within the cloud earlier than. They didn’t know
what the payoffs could be, and wished to maximise the possibilities of
discovering worth within the cloud.

In consequence, some product groups are reinventing the wheel as a result of
Etsy doesn’t have present implementation patterns and companies. Now
that they’ve extra expertise working within the cloud, platform groups
know the place the gaps are and may see the place tooling is required.

To find out if the investments are paying off. Etsy is monitoring
numerous measures. For instance, they monitor traits in SLI/SLOs associated
to reliability, debuggability and availability of the methods. One different
key metric is Time to Productive – the time it takes for a brand new engineer
to be arrange with their environments and make the primary change. What
precisely meaning adjustments by area; for instance it may be the primary
web site push or the primary information pipeline working within the huge information
platform. One thing that used to take 2 hours now takes 20 minutes.

They mix these quantitative metrics with often measuring
engineering satisfaction, utilizing a type of an NPS survey to evaluate how
engineers get pleasure from working of their respective engineering environments,
and provides a chance to level out issues and counsel enhancements.
One other fascinating stat is that the infrastructure has expanded to make use of
10x the variety of nodes however solely requires 2x the variety of individuals to
handle them.

Measuring Value and Carbon Consumption

Etsy continues to embrace measuring every part. Shifting to the cloud
made it simpler for groups to establish and monitor their operational prices
than it had been within the datacenters. Etsy constructed instruments on prime of Google
Cloud to offer dashboards which give perception into spending, so as
to assist groups perceive which options had been inflicting prices to rise. The
dashboards included wealthy contextual info to assist them make
optimization choices, measured towards their understanding of what
very best effectivity ought to be.

An important firm pillar is sustainability. Etsy experiences their
power consumption of their quarterly SEC filings, and have made
commitments to scale back it. They’d been measuring power consumption in
the info heart, however making an attempt to do that within the cloud was initially extra
troublesome. A group at Etsy researched and created Cloud Jewels, an power
estimation device, which they open-sourced.

We’ve
been unable to measure our progress towards considered one of our key affect targets
for 2025 — to scale back our power depth by 25%. Cloud suppliers
typically don’t open up to prospects how a lot power their companies
eat. To make up for this lack of knowledge, we created a set of
conversion components referred to as Cloud Jewels to assist us roughly convert our
cloud utilization info (like Google Cloud utilization information) into approximate
power used. We’re proud that our work and methodology have been leveraged by
Google and AWS to construct into their very own fashions and instruments.

— Emily Sommer (Etsy sustainability architect)

These metrics have lately been added to their product dashboard,
permitting product managers and engineers to seek out alternatives to scale back
power consumption and spot whether or not a brand new function has had any impact.
Thoughtworks, who has the same sustainability mission, additionally created an
open-source device referred to as the Cloud Carbon Footprint, which was impressed
by preliminary analysis into Cloud Jewels, and additional developed by an
inner Thoughtworks group.