welcome to episode number 186 of the weekly
“Google Cloud Platform” podcast. My name is Mark Mandel. And I’m very excited to be
joined today by Carter Morgan. How are you doing today, Carter? CARTER MORGAN: I’m
doing well, Mark. It’s good to be here. MARK MANDEL: Yes, so glad
to have you on the podcast. It’s your first
time here, so you want to give people a
little intro to yourself and what you do? CARTER MORGAN:
Yeah, it’s actually my first time since I got back
because I was here, like, two, three years ago. MARK MANDEL: Yeah. CARTER MORGAN: Developer
advocate, though, so I’m here. And I just make
videos, and I just make videos about everything
Google Cloud related, and specifically, VMs and
GCE coming up pretty soon. MARK MANDEL: I’m just
actually trying to remember, you were on the
podcast a while ago. Like, it was a long time ago. CARTER MORGAN: A long time
ago, like Kubernetes 1.3. MARK MANDEL: 1.3. CARTER MORGAN: Yeah. MARK MANDEL: Yeah. OK, so you’re no stranger. It’s fine. CARTER MORGAN: Been
here, done that. MARK MANDEL: So today we’ve
got a really cool topic. We are going to talk
about blockchain with developer advocate
Allen Day, which I’m kind of excited about. Hot topic, but also
kind of interesting. CARTER MORGAN: I know so
little about blockchain. It’s time to learn. It’s time to figure this out. MARK MANDEL: Yeah. Yeah, it’s going to be fun. And then after that, we have
our question of the week, as per always. I’m going to ask
Carter about VMs. What are the four types
of virtual machines that exist on Google Cloud Platform? CARTER MORGAN: Oh, intriguing. [MYSTERIOUS PIANO] MARK MANDEL: Intriguing. But before we do
that, why don’t we get stuck into our cool
things of the week? [MUSIC PLAYING] We are talking about
blockchain today, so it is actually quite
apropos that there’s an article talking about
blockchain.com scaling and saving with Cloud Spanner. So blockchain.com
is a website that serves the cryptocurrency
blockchain, [DING] really unsurprisingly,
[DING] and it has helped 39 million cross-platform
wallet users in 140 countries worldwide access the
crypto ecosystem. And they’ve been using our
horizontally scalable database spanner to basically help
them with their needs in managing what is
essentially money, which is important to some people. So if you want to see
some details about what it is they’ve been doing and
how they’ve been using it, make sure to check
out the article, which we’ll have in the show note. CARTER MORGAN: OK. That was cool, but what
about winning AI records and competitions? MARK MANDEL: I suppose
it’s all right. CARTER MORGAN:
That’s pretty cool. So Google Cloud recently
was in a competition, and they set three new
performance records in the latest round of the
MLPerf Benchmark Competition. So this is like an
industry wide standard for measuring ML performance. And so it’s cool because
things like object detection or natural language modeling are
what these algorithms are used for, and Google Cloud
TPUs do it very fast, like over 84% faster than the
fastest on-premise systems in the MLPerf Close Division. So I thought that
was really cool. There’s a whole article
about it on cloud.google.com called “Cloud TPU Pod
Breaks AI Training Records.” Definitely check that out. MARK MANDEL: Very nice. So we have some updates as well
to Cloud Memorystore for Redis, our fully managed Redis service. So if you were looking
for Redis 4.0 support, that is now available. Also what we have available is
the new import-export feature, so you’re able to import and
export out of a Cloud Storage bucket of your choosing. So you can basically get your
data in and out of Redis really easily. Just two really nice
features that are super nice to talk about. CARTER MORGAN: That’s very cool. All right, features that
are nice to talk about– let’s go back to
Kubernetes and being able to customize Kubernetes
to fit your needs. So there’s this article that
came out, “To Run or Not to Run a Database
on Kubernetes.” And it basically
talked about how if you can’t find something
like a fully managed database like Cloud Spanner, or if you
don’t want to do it yourself on a VM, how you can do it on
Kubernetes using operators, so you can make basically
custom code to run and help you manage this database. So that, I think, is super cool. When I customize a database
to fit exactly what you need for your application,
you can check that out in this article. MARK MANDEL: Yeah. I really like the
flowchart in the article as well, where it
talks about, like, this is how you can easily make
the decision about where you should put your database. So I thought that
was super nice. Finally, we want to just say
congratulations to Elastifile. [COIL SOUND EFFECT] We want to announce
that Google has entered into a
definitive agreement to acquire Elastifile. They are an enterprise
file storage for the cloud, basically
lovely scalable file storage. We actually have
Elastifile file service on GCP, which is a fully
managed version of Elastifile integrated with
Google Cloud, and we have some large customers
already using it, but super excited to
have them on board and joining the
Google Cloud family. CARTER MORGAN:
Congrats, Elastifile. [APPLAUSE EFFECT] MARK MANDEL: Excellent. Awesome. Well, Carter, why don’t
we go chat with Allen and learn all about blockchain? CARTER MORGAN: Ah, I can’t wait. Let’s do this. [MUSIC PLAYING] MARK MANDEL: So very excited
to have Allen Day here, a developer advocate for
data science and verticals, joining us. Allen, how are you doing today? ALLEN DAY: Hey, I’m doing great. Thanks so much. MARK MANDEL: Yeah, thanks
so much for joining us. You’ve come today to talk to
us about blockchain, which– ALLEN DAY: That’s right. MARK MANDEL: –should
be kind of fun. But before we get
stuck into that, do you want to tell us a
little bit about yourself, who you are, what do you
do here at Google? ALLEN DAY: My name’s Allen. I’m a developer advocate
in Google Cloud. And I work on the data
and analytics products. That would be BigQuery,
and Pub/Sub, and Dataflow, and Bigtable, and any
of the computer storage resources for
processing and analyzing large volumes of
fast-moving data. My PhD is in human
genetics, so I have a background in
biological data types, like genome sequencing
and medical imaging, and that’s part of
what I do at Google. And then I also
cover blockchains and cryptocurrencies because
I’m based in Singapore, and that’s one of
the more advanced regulatory jurisdictions
for this type of technology. Singapore and Switzerland
are the two hotbeds of activity for incorporating,
and making a foundation, and so on. So I figured I could
do work locally by focusing on this industry. And it’s been going pretty well. CARTER MORGAN: That’s
exciting to hear. Wow. Very technical stuff, too. So then just a
question I have is on, like, a day-to-day
developer advocate, what do you do in these areas? Are you making videos or– ALLEN DAY: Yeah, sometimes. I used to do a lot of events. And I know that part of what
we’re going to talk about later is where I’m showing up soon. But it’s less than I used
to, being out in the field. I tend to do a lot more
writing these days. I find that it’s just better to
have my time blocked like that, with less switching cost
of moving around and not losing days during travel. I could do more
focused deep work. And then that also lets me spend
more of my interrupt time doing coordination with others. The cryptocurrency
data, which is mostly what I’m working
with, is public data that we can index, just
like Google indexes the web. And so as a result, all
of the work that I’m doing is open source development. And it’s with distributed open
source teams of contributors, where we all have some
business interests, but are also volunteering time
to build open source repos and open source data sets. And so there’s a lot
of interrupt time dedicated to team
coordination, herding cats. MARK MANDEL: So all of
that sounds amazing. But why don’t we start
at the high level and work our way
down all the way through the chain, which
sounds really cool. No, that sounds awesome, though. So like, OK, what is blockchain? Because I actually have no idea. It’s a word that gets
thrown around a lot, especially here in the Bay
Area, as where you are as well. So like, what is
this thing that we keep hearing about everywhere? ALLEN DAY: I would guess
most of the audience should be familiar with
something called BitTorrent, which is protocol [INAUDIBLE]. I see your heads are nodding. It’s a protocol for a bunch
of peer-to-peer technologies in general, which is a
type of distributed system. So these peers get
together, and they’re trying to solve some problem,
like make sure everybody has the same copy of the file. Or in the case of
this blockchain, they’re trying to make
sure everybody agrees that the latest set of
transactions that are being added on to the
end of a log file are agreed upon by
everybody that that’s really what is in the latest block. So there’s some consensus
algorithm, essentially, being reached by a group of
peers to coordinate work. Think about it like that. MARK MANDEL: How does
that kind of work? ALLEN DAY: In terms of when
we talk about data structures, that’s probably the easiest way. MARK MANDEL: Yeah. Like, do you and I just chat,
and I’m just like, it’s cool, and you’re like,
it’s cool, and then we just go on our separate ways? Or is there some
sort of contract, or what does that look like? ALLEN DAY: So there’s many
of these different blockchain implementations, but at
the most basic level, they’re using a cryptographic
function, a hashing function, right? Where the output
is deterministic based on the input, but
it’s not predictable because the outputs appear
to be uniformly distributed relative to whatever
input you get, which means that for a
given input and output, a peer on the network can
verify that the output is what it ought to be
based on the inputs. You can match them. But in order to find
a particular output based on the input is
difficult because the outputs are uniformly distributed
relative to the input. It’s random, essentially. OK, so what that
means is that if we try to reduce the entropy
of the output, like by– let’s say we have a
bit vector of 256 bits, and we’re trying to
make the first bit zero. OK, that’s really easy. That’s 50% of the time. But if you try to make the first
10 bits zero or the first 100 bits zero, it becomes a lot more
difficult to find that, given that you have to do a
search basically just by throwing darts at a board. Just because you’ve got
99 zeros in one place, you can’t move directly
next door and get 100. This could be back
to one, one zero. MARK MANDEL: Got it. ALLEN DAY: So what
these peers do is that they look
at how long it took to find a solution for a
particular number of zeros to emit a block in a
particular amount of time. So it’s basic set of some
computing power being consumed over time, which they
call a hash rate, so how many of these hashing
functions can be done [INAUDIBLE] at a time. And if a solution was
found too quickly, then the peers agree
because they’re all sharing the same code base
that the difficulty should be increased. We now need to add more
zeros to make it slower. Or if it was done
too slowly, then we make it an easier
problem to solve. And so this is how they’re
doing the agreement. Some peers are
doing this hashing to search for solutions
to make a block within a particular
amount of time. And they’re adjusting the
difficulty more or less based on how close they are
to the desired amount of time. CARTER MORGAN: OK. ALLEN DAY: OK, so now
let’s talk about what is this block all about. So what are they
trying to agree about? The block contains a
set of transactions. So each of the entities
that is participating on the– the thing about
cryptocurrencies now– this is to make it
simple, like Bitcoin– anybody can generate a key pair. This is just like the
same kind of key pair you use for PGP,
or SSH, or SSL– same thing. So you generate a key pair. And your private key can
produce a public key, so that’s deterministic. You can get the public key
directly from the private key, but you can’t get the
private key directly from the public key. So what you do is
you share activities that you’re interested
to do by signing with your private key,
which is then verifiable that it corresponds
to your public key in assigning the transaction
that you want to enter onto the ledger, this log file. So the blocks
consist of a series of transactions
that are essentially entries in a log file. All the records, which are
these transfers of value, like moving Bitcoin
from party A to party B, those transfer
operations are all grouped together in a block. And the block is a transaction. Like in SQL, where you
say begin and then commit, the commit is like that. So all the transfers
within the same block have the same timestamp. MARK MANDEL: So very much
like if I was in a bank and I was like, I
want to move money from here to here, and between
two accounts, that’s why you need multiple transactions. Well, essentially,
it’s one transaction wherein one account
is removing something, and one account is
adding something, but at the same time. ALLEN DAY: Correct. MARK MANDEL: But in a way that– ALLEN DAY: It’s [INAUDIBLE] MARK MANDEL: –there’s
a distributed, I’m guessing it’s like a
peer-to-peer system that is able to say, OK, I
know that this happened, and we all know
that this happened. And we can verify
this at a later date that we all agree on. ALLEN DAY: Yeah. And so there’s an
interesting bit here. So I was talking about
hash rates earlier, and so the incentive
for these computers to be doing the hashing
is that what they will do as part of their
attempt to find a solution to this hashing
problem on a block, right, is they will include their own
transaction to give themselves new Bitcoins as
part of that block. So if they produce
the block, their block contains a block reward that
produces new Bitcoins that go to the beta of the
block that it’s then broadcasting onto the network. And so that particular transfer
is a special type of transfer called a Coinbase
transaction, which is where if you’ve
heard of Coinbase, which is the biggest
exchange in the US for doing cryptocurrency activity. CARTER MORGAN: And
so the idea there is to incentivize
people, allowing you to use their computers to
form these hash operations? ALLEN DAY: That’s correct. Yeah, because you can
mint new cryptocurrency by doing the hashing. CARTER MORGAN: Very cool. So I got a question where
you were saying basically before, we talked
about cryptocurrency. Before that, we were
talking about basically streaming video. And you said that
you have a background in biomedical and agricultural. So can you use this
for that as well? Are there applications
in those areas? ALLEN DAY: Yes, there are
some open source projects in this blockchain
space that are working on streaming video applications,
something like a CDM, as well as doing distributed
file storage, which would be the equivalent of,
like, fully peer-to-peer Google Cloud Storage equivalent. Biomedically, but there
are some companies that are working on making
data ownerable by the user. So if your data are encrypted
according to a private key that only you control, then you
can decide who has access to it at what level and at what
time and can revoke access based on your encryption
of your own data. CARTER MORGAN: Oh, that is cool. ALLEN DAY: Yeah, there there’s
various ways you can do this. Creating new keys and
recrypting or doing multi-signature keys where
more than one party has to sign in order to get
access to some data, this is also possible. Yeah, so there’s
applications like that. For analytical applications,
which is where a lot of what GCP is doing for
biomedical kind of things, there’s not really any good way
to partition the computing– to shard it, is what these
distributed systems people call it– to make the computations
cost effective relative to a centralized cloud. And so it would be
for doing, like, a deep learning
model of retinopathy, or a breast cancer detection,
or something like that. It’s just not really a
tractable problem right now, given the architecture of
the peer-to-peer systems. MARK MANDEL: Well, it sounds
like, then, that there’s actually a wide variety of
use cases for blockchain outside of cryptocurrency,
is what I’m hearing. ALLEN DAY: It’s true. Yeah. Any place where you want to
have immutable data where you can prove that was
produced and nobody tampered with it, that’s a
good place where you want to use some of this technology. So with all this fake news
stuff that’s going on, like wow, all really
good applications. CARTER MORGAN: So then,
I mean, if there’s so many use cases for this,
it’s kind of strange because I’ve heard
of blockchain a lot, and people joke about it. Like, every startup has a new
blockchain or cryptocurrency offering. And so why is it that
this is something that people take
light of so much, something people
make so much fun of? ALLEN DAY: Yeah,
it got a bad rap. Kind of a couple
of dynamics here. So Bitcoin is the granddaddy
of all of these things, and the Bitcoin
community enthusiasts have their own
strong culture that has some political ideology
associated with it. It’s very similar to the gold
bugs, where there’s, like, finite supply, and
you can’t inflate it. And it runs contrary to how
mainstream macroeconomics work and how central banks operate. And so it’s a different
kind of thing. And because it looks
different, it gets ridiculed. The other aspect of
this is that Bitcoin has been having some
success, and Ethereum has been having
some success, which is another cryptocurrency
and smart contact platform we can talk about. But as a result of that, on
Ethereum, you can actually mint different types of tokens,
which are not a first class cryptocurrency in
their own right, but they can have
cryptocurrency-like attributes in that they can be
created and traded. And so because there were
enough users using this thing, some clever
marketers figured out they could start doing purely
financial and marketing projects that didn’t have
much technical basis or merit for solving any problems. They could just create a token
and ICO it, and buy a Lambo. And there was a bunch of
that that happened in 2017, and a lot of people
got burned by that. So that’s another reason
that things are ridiculed. MARK MANDEL: That’s fun. Definitely, I’m in the
same camp as Carter, where I’ve been told, if you
want 5% extra evaluation, make sure that blockchain’s
in your startup title. ALLEN DAY: Yeah. There’s some good– maybe we can
link to this in the description about one of these ice tea
manufacturers calling them, like, blockchain iced tea, and
the stock went up overnight by double digit percent. MARK MANDEL: Getting back to
more functional type of things, so we talked a little bit about
what blockchain is good at– and maybe it’s also kind of
relevant since we’re saying people feel like they can
apply it to anything– but what is blockchain bad at? What’s it not good for? I’ve heard of people
saying things, like, oh, we’re going to use
blockchain for databases. Is that a terrible idea? I have no idea. ALLEN DAY: It’s
an OK idea if you want an append-only database. So basically all of
these things boil down to being log files that
have some mechanism of reaching consensus about what the
contents of the log file are, using a group of peers that
may completely not trust one another or may have
some level of trust between one another. MARK MANDEL: And I’m
guessing that trust is sort of implementation detail– ALLEN DAY: That’s correct, yeah. MARK MANDEL: –depending
on how you set it up. ALLEN DAY: How you
implement that, and where you have a public
chain or a permission chain, it’s also related to trust. And so yeah. So if you have a
situation where you need a log file that
can’t be tampered with, you have a use case. But if you want to do
something like search the log file to
find old records, it’s not a use case for that
because it’s really just adding things onto the end of a list. That’s really what
the primary concern is of one of these consensus
transaction heading systems. MARK MANDEL: So I’m thinking
through communication type stuff. Like, I send you a message. You know it comes from me. You know the messages came in
the order that they came in. That would be a fit? ALLEN DAY: That
would be a fit, yeah. But realize that every
one of these messages has to be added into something
that all the peers are validating. So it could be that there
is some throughput problems with this. It’s more and more I/O. So this
is another scaling difficulty of these systems. There are some ways
around it using things like zero knowledge proofs
and other cryptographic stuff, but that’s outside the scope
of what we’ll talk about today. And it’s not clear
that that will even be deployable into production
at large scale anyway. But I want to go
back to something. So I was talking
about like, OK, it’s really good for adding entries
to the log file, right? Putting my data
science hat on, I was looking for some kind
of thing I could write about as a developer advocate to get
people excited about working on Google Cloud. And I wanted to do some
network analysis of how all these parties are
interacting with one another on the Bitcoin blockchain, or
any blockchain for that matter. Who is sending money to
whom, and when, and how much, and can we figure
anything out about it? Are there any patterns here? Can we identify exchanges? Can we identify big owners, high
throughput nodes, et cetera? These type of questions are
very similar to the kind of questions that I was
addressing in my dissertation work because genetic regulatory
networks have the same kind of characteristics. So there’s some
information in the form of structured molecules coming
into a cell that are food, basically, or parts
to help the cell run. And then there’s
some waste molecules that are being thrown out. And then everything
in between has to do with a bunch of proteins
and other molecules that are parts of the cell that allow
that highly structured material to keep the cell running
and turn into the exhaust. And so how does that
regulation work? Well, the way that
we go about analyzing it is by collecting
some data using a biochemical assay of some
sort, typically very high throughput one. And then we look at
the data over time to see what’s going on. And that type of analysis
requires all this network analysis methodology, which
could be directly applied to crypto economic networks. I’m thinking, OK,
this would be great if I can just take some
of these things I already know how to do and apply it
to this other data set that’s getting a lot of
attention right now. Because nobody’s talking
about economic network analysis of these systems. They’re just talking
about ICOing– making tokens, right? MARK MANDEL: What
does ICO stand for? ALLEN DAY: Initial Coin
Offering– it’s like an IPO. MARK MANDEL: Got it. ALLEN DAY: Yeah, so
I tried to do this. And because I
didn’t know anything about any of this stuff
about a year and a half ago, so I tried to go and get the
data from the Bitcoin node. And I was like, well,
why can’t I get the data? Why can’t I query it so that I
can do an aggregation over time to look at what is the volume
of transaction over time? There’s no way to do this. So I was like, OK, well,
I guess I can ETL this out into BigQuery and query it. And so that took a lot more time
than I thought it would take and justified writing a
blog post about just doing that, and giving that
quote away because it was really hard to do. So I did that, and everybody
was very happy about it and started using the data set. I got a lot of traction
on our public data set program as a result. And so because of
that, I haven’t really done any analysis. I work with a bunch
of analysts now. But mostly, I’m focused on
building the infrastructure to support other analysts. CARTER MORGAN: Right. OK, now, I’m curious. GCP specifically–
what kind of things do we have that help with that? You said you did
some ETL analysis, so that sounds like Stackdriver
or something like that maybe. ALLEN DAY: Yeah, so how we
have things set up right now– and this is all in GitHub if
anybody wants to grab the code to replicate this in
Amazon or wherever else, it’ll all work–
we’re using Kubernetes to run the blockchain nodes. And we have health checks
and load balances in place so that if one of
our peers falls over, we failover to the other one. And then there is a
[INAUDIBLE] composer service that is holding the
blockchain nodes periodically to get data out. When a new block shows up, we
stick it into a Pub/Sub topic. We publish to the topic. And then there’s a
subscriber to the topic that streams the data into BigQuery. So we make the BigQuery tables
available to users to query. But if they also
want to subscribe to the stream of events,
those topics are also public. So people who want to do real
time trading based on events that Google is
publishing to the topic, they could do that
if they want to. MARK MANDEL: Just so I’m
clear as off the nodes, how are they getting
that information? Is that, like, a centralized
repository of some kind, or– ALLEN DAY: The ones that
we’re running on Kubernetes? MARK MANDEL: Yeah. ALLEN DAY: So when you
turn on the Bitcoin node, it connects to its
peer-to-peer network. MARK MANDEL: Oh, I see. ALLEN DAY: So it’s a
bootstrapping process. MARK MANDEL: Right, got it. So you sort of become one
with the mesh network. ALLEN DAY: You become
one of the peers. And all the peers
start sending you old data until you’re fully
synced with all the peers. And then if you want
to, you can start producing new blocks by mining. We don’t do any mining, though. We are just receiving and
transmitting existing chain data to be a good citizen
on the peer-to-peer network, but we’re not actually
mining any cryptocurrency. CARTER MORGAN: I need to
look this up right now. That sounds so cool. ALLEN DAY: Not every day, but
it’s more than once a week, I get pinged by
somebody who’s done something interesting
with this data set, some kind of analysis. MARK MANDEL: Well, yeah. What sort of interesting
things have people found? ALLEN DAY: The most
interesting things are related to basically
characterizing entities on the network. So it’s pseudonymous, right? So any given actor
on the network has a private key
that’s private, and then they’re
signing transactions so you can verify what
their public identity is. But beyond that, you have
no idea how any given actor corresponds to any other actor. But if you start looking at
patterns of activity over time, you can begin to get
some sense of what attributes could belong
to a particular actor on the network. Let’s say there’s a node that’s
receiving a lot of inputs from many different
other addresses and is outputting to
many other addresses. This is some kind of flow node
that has a high throughput. It’s typically an exchange
where people are going to trade tokens, for example. Yeah, so this is the direction
we’re heading right now. So we’ve made the
BigQuery tables available. That was the first thing. And we’ve done this for
eight different blockchains at this point– Bitcoin, Ethereum, a
bunch of other ones. Then we made the
Pub/Sub topics available so people can subscribe to that. I’m now working on
two additional fronts. So one of them is data flows
where the Pub/Sub topics will be read by a data
flow and output to another Pub/Sub topic
for producing derivatives. So we could basically
do some kind of sliding window analysis,
run it all the time, and just publish to another
topic periodically for making, like, a candlestick
chart, if you to see something like amount of
activity on network over time. The other one is really to what
we were just talking about, analyzing the network, possibly
also in real time, where we can find anomalous
transactions, or find labels on entities
that are changing, or the entity has some
significant volume through it. And these type of events of
turning that continuous data into discrete data
are actually a way to take the graph, the
crypto and economic graph, and vectorize it. So now we have a
vectorized graph, and we can start
applying standard machine learning models onto
this because we’re now working in a vector
space as opposed to a high dimensional
graph space. So I’m planning on putting
all that in Bigtable and doing cool TensorFlow
stuff on top of that also inside of DataFlow. CARTER MORGAN: Wow. So this is such a complex topic. And something you said
earlier really shocked me. You said a year
and a half ago, you didn’t know too much about this. So you had to ramp up. Are there resources
you maybe recommend to people who are interested
in also ramping up in blockchain, crypto
applications, any of that? ALLEN DAY: Yeah,
read any of the stuff that I’m writing because
it’s all from the beginning. You can just look
at what I’m doing. Go back to my early
GitHub commits and look at my [BLEEP] code. MARK MANDEL: Excellent. Are there any other
particular people that you follow or look
for inspiration from, or online resources or courses? ALLEN DAY: Yeah, there’s
a lot of resources. You could just go, like, how
does Bitcoin work on YouTube, and you’d find all
kinds of very basic, conceptual cartoon
diagrams all the way to down-the-rabbit-hole,
cutting-edge cryptography. Other good podcasts– how
about we talk about that? There’s a good business
podcast, the “Unchained” podcast from Laura Shin. That’s a good one. OK, so another good one is
the “Off the Chain” podcast. This is from Anthony Pompliano. I was just talking about zero
knowledge proofs a moment ago. He did a great interview
with the Zcash founder a couple of weeks ago. Yeah, I highly
recommend that one. So the stuff Laura Shin is
doing is more business related, and then Pomp is doing some
technical, some business stuff. Yeah, those are both really,
really good resources. MARK MANDEL: Fantastic. All right. Well, unfortunately,
we’re starting to run out of time a little bit. But before we do that,
is there anything that you want to mention,
any events or content that you’re pushing
out that people should hear about, or know
about, or keep an eye out for? ALLEN DAY: Yeah, I published a
blog post about three weeks ago that I was showing how you
could use Ethereum, which is a smart contract
platform, special type of one of these peer-to-peer networks
to start composing not only applications where
we can read, and index, and process the data from
the blockchain on cloud, but how to call out to
cloud from the blockchain. So we were talking earlier
about biomedical analyses and not having the kind of
right architecture on chain to be able to do
the computing as efficiently as a
centralized cloud provider. But there is some way
that you can do this if you can call to the
cloud from the blockchain. So this has sort
of become circular. So we can index data from
the chain into Google Cloud, and then call Google Cloud from
the chain to get data back. CARTER MORGAN:
Absolutely mind-blowing. Allen, all day, you’ve
been blowing my mind with information. Thank you. ALLEN DAY: Yeah, so I wrote
an article about that recently about building
these hybrid cloud blockchain applications
that I think would be really interesting
if people want to see what cutting-edge stuff is going on. MARK MANDEL: Excellent. Well, Allen, thank you so
much for joining us today. I definitely learned
a lot, and now I’ll maybe make slightly less
jokes about blockchain. ALLEN DAY: OK, great. Yeah, it was a pleasure. Thanks. You asked such great
questions, so yeah, I had a good time, too. CARTER MORGAN: Thank
you so much, Allen. I learned so much. I really appreciate this. MARK MANDEL: I now feel like
I might actually vaguely know what blockchain vaguely is. Now Allen’s a great teacher. Now I have a much
better understanding. So I feel really
good about that. CARTER MORGAN: It blew my
mind just the wide range of applications it’s
used for, not just money. MARK MANDEL: Not
just cryptocurrency. Yeah, exactly. Fantastic. All right, why
don’t we get stuck into our question of the week? CARTER MORGAN: Dun
dun dun, dun dun dun. Dun. MARK MANDEL: OK, so what are the
four types of virtual machines that exist on Google
Cloud Platform? I actually didn’t realize
there were that many. CARTER MORGAN: So
Google Cloud is trying to make things
simple and efficient. So they came up with four
different types of VMs for different types
of applications. So it’s easy to
remember it if you think about the types
of applications. So they have a
general purpose one for most types of
applications, and it’s called general purpose VMs. Then they have memory optimized
VMs for large database and memory databases. They have compute optimized
for intensive gaming and scientific modeling. And then my last
and favorite type are called preemptible VMs. And so these are
really affordable. They’re going to be,
like, 80% of the cost. And they’re just a
flag that you can add to any of the
other three types to say, I don’t need
these to run in real time, so you can preempt
me at any time and run other applications, as
long as this runs at some time. And so those are
great for, like, batch rendering or
simulations that don’t need to be done in real time. MARK MANDEL: So really, don’t
we kind of have six then? [POLITE LAUGHTER AND APPLAUSE] CARTER MORGAN: Wait, wait, wait. MARK MANDEL: Because we’ve
got three of the same ones, and then they’re
each one of them can be preemptible,
rather than four. CARTER MORGAN: I mean,
that is logical, yes. MARK MANDEL: It depends on
how– look, it’s just fine. It’s fine. It’s fine. CARTER MORGAN: That was a
trick question this week. I love it. MARK MANDEL: Awesome. All right, just
thinking it through. It’s fine. Cool, but no, that’s great. Good to know that we have
different types of machine types for different
kinds of workloads. So thank you very much, Carter. CARTER MORGAN: You’re welcome. MARK MANDEL: So
actually, Carter, where are you going to be? What are you doing? What cool stuff are you doing? I know you’re doing cool stuff. CARTER MORGAN: Yeah. All right, so outside
of work, something cool is I’m in the Comedy Festival,
the Edinburgh Fringe Comedy Festival for a whole month. MARK MANDEL: Whew. CARTER MORGAN: And
Google’s letting me work from there for a month. It will be really cool. And workwise, I’m
releasing a bunch of videos about VMs and GCEs,
Cloud Security Command Center. So there’s a lot of
cool stuff coming out. Check out the GCP YouTube page. You can see a lot of it there. MARK MANDEL: Awesome. Those videos sound great,
but I think more importantly, so you’re actually doing a
show at the Edinburgh Fringe Festival? CARTER MORGAN: I am. I’m doing a comedy show. I’m terrified because it’s
my first time doing a show at this scale on this level. MARK MANDEL: Yeah. CARTER MORGAN: Yeah. But it’s a lot of fun. And again, I’m just happy I’m
at a place like Google that allows me the flexibility
to go out and try this– MARK MANDEL: That’s fantastic. CARTER MORGAN:
–while still working. MARK MANDEL: Awesome. CARTER MORGAN: Yeah,
it’s really cool. MARK MANDEL: That
sounds fantastic. Well, I’m not doing anything
nearly that exciting. I’m going to be at Tokyo Next. I will be talking at the
[INAUDIBLE] Mini Conference inside of Tokyo Next, talking
about Agones and game stuff. Then I’ll be the Open
Source in Gaming Day, just before the Open Source
Summit down in San Diego. And I’ll be at the
Open Source Summit as well, talking about open
source in games and Agones. And then I’ll be at Pax Dev,
speaking about now actually not Agones this time. I’m actually doing
two separate panels. I’m doing one panel with
a bunch of colleagues, talking about open
source in games, and if that’s something
you might be interested in, and why you might want to do
it, and that kind of stuff. And then I’m also
doing another panel with some other colleagues,
talking about dedicated game servers, why you
might want to use them and [INAUDIBLE] they’re in. And I’ll also be hanging
out at Pax West, too. CARTER MORGAN: Oh, that’s cool. I’m definitely going to
have to pick your brain about open source in games. I didn’t know that was a thing. MARK MANDEL: I’m trying
to make it a thing. I’m trying to work really hard. There’s a lot of
work that’s happening inside open source in games, but
I think there’s a lot of work still to be done. CARTER MORGAN: Mm. MARK MANDEL: Fantastic. Well, Carter, thank you
so much for joining me this week on the podcast. CARTER MORGAN: Thank
you for having me. It was a blast. MARK MANDEL: Yeah,
I look forward to having you as a regular host. And I’m sure we’ll make this
happen again at some point soon. CARTER MORGAN: Oh, great. I will see you soon, Mark. MARK MANDEL: Wonderful. Well, thank you
all for listening, and we’ll see you all next week. [MUSIC PLAYING]