Ben Golub, CEO of open source file system software maker Gluster, shares his thoughts on scale-out storage, cloud storage gateways and appliances and the idea that commodity hardware will beat out proprietary systems in this week's episode of Cloud Cover TV.
- GlusterFS open source file system
- Factor of ten reduction in price of storage by making hardware a commodity
- Gluster users sit on top of Amazon EC2 and EBS
- Encryption coming to Gluster, but it is POSIX compliant today
- Dedupe technology gets you another 30%-40% price reduction, but is not supported in Gluster
- Open source means Gluster has thousands of developers, QA people
- Ten times as many people using the open source version versus the professional version
- Cloud storage appliances and gateways complimentary to Gluster
- Cloud as backup but more compelling is cloud for disaster recovery
- Cloud for bursty workloads
- Gluster lets you have the apps and storage in the cloud and in your on premise data center, replication between the two
- Gluster is part of OpenStack
- Bandwidth issues getting data from point A to point B
- Long term market trends favor commodity storage
Read the full transcript from this video below:
Gluster CEO discusses best cloud storage options
Jo Maitland Jo: Hi. Welcome to Cloud Cover TV, our weekly show on
all the juiciest news in the computing market. I'm Jo Maitland, here in
San Francisco and this week I have Ben Golub from Gluster on the show.
Ben Golub : Thank you.
Jo Maitland: Tell us about Gluster FS.
Ben Golub : We're used both by several customers who are in
Amazon and who are using us in association with Amazon web services,
as well as cloud service providers who are either competing with Amazon
or trying to come up with complementary storage offerings.
Jo Maitland: Can you name a few? Tell us about them
Ben Golub : Yeah. So Dedipower, for example, is one. SingleHop is another.
But, in general, what we are finding, whether they are on premise
customers or in the cloud customers, is that people have sort of
recognized that the paradigm needs to shift in terms of storage. The
same way it changed in terms of computing. Nobody goes out and buys
big mainframes anymore and yet, for the most part, storage is still
delivered as large proprietary monolithic boxes. What we aim to do is
give you the ability to have storage that's created by pulling together lots
of, essentially, cheap-boxes, whether those are boxes you own or boxes
that you rent from Amazon, into one highly available, high-performance,
highly scalable storage system.
Jo Maitland: So do you compete with something like Amazon S3 or are you . . .
Ben Golub : No, in fact we are part of Amazon. We are deployable
as an Amazon machine image within Amazon. People who are
using Gluster tend to be a little bit more demanding in terms of people
who are normally using S3. So Gluster sits on top of EC2 instances as
well as Elastic Block Storage. Essentially we give you the ability to take
existing applications that you would be running on premise, requiring
high performance and high availability, and allow you essentially, with
no effort, to move them up to the cloud with all the benefits that
come from scalability, flexibility. . .
Jo Maitland: Right. So are there other use cases besides file serving?
Ben Golub: In general the use case is fairly broad. So any type of
application that you're running where you have a need for storage which
is file based as opposed to block based, but where you need performance,
where you need availability and where you need the ability to scale up or
scale down, we work well. So that's media applications, health care,
bioinformatics, consumer web, you name it.
Jo Maitland: Good. Does the product support encryption? I guess that's a
big question these days with cloud storage.
Ben Golub : So, we are compatible with a lot of encryption systems. Gluster
was designed to be what's called "POSIX compliant." So basically it works
with 95% of the applications that are out there. We are, though, adding
At - rest encryption capabilities in the near future.
Jo Maitland: How about things like D-Dupe and other optimization
technologies? Any plans there?
Ben Golub : Yeah. So our view was that the first thing we needed to
do was to do a factor of 10 reduction in the price of storage. So, by
making the underlying technology commodity based -- buy only what
you want, when you want -- that was sort of the big economic savings
we were able to drive. And then, from there, things like deduplication
get you another 30 or 40% in terms of the amount that people can save.
Jo Maitland: So, I'm curious how you guys make money since the
software's open source. What's the business model?
Ben Golub : The business model is fairly traditional open source,
in that we have an open source version and then we have an enterprise
version which comes not only with professional support but also with
additional advanced features.
Jo Maitland: Okay. It's interesting. I've noticed lately that, that's a real
trend with companies. Start with an open source, particularly in the
cloud market. Once they get any amount of traction, there's a closed
source version for commercial reasons but also for additional support.
You've got to . . .
Ben Golub : Right.
Jo Maitland: Right. That requires you guys to hire people.
Yeah, it's expensive.
Ben Golub : Yeah. You can't build a big fantastic software system
of any kind if you don't have some revenue stream. Of course the
exciting thing about being open source is that we're really not just
limited to the 30 or so developers we have on staff, but we have
thousands of developers, thousands of QA people and, what's very
exciting for me, thousands of marketers and thousands of sales people
who happen just not to work for the company but they're part of the
Jo Maitland: Right. So what's the split, would you say, between the
people using the open source version and the people using the closed version?
Ben Golub: Yeah. There are probably 10 times as many people
using the open source version as are using the professional version
but what we tend to see, especially since this is storage, is that
people will download the open source version, they'll try it out in
their environment and once they get to a certain size, they recognize
that they want to have support. This is sort of the nature of storage as
opposed to things that may be higher up in the stack. A failure in storage
is a pretty catastrophic failure as opposed to "Okay your application
stopped working, you have to restart it." If storage fails, you can lose
data, which is a bad thing.
Jo Maitland: What do you make of the interest in all the cloud
gateways -- the appliances now for storage? Riverbed has the
Whitewater product, there's Nasuni, there's a whole bunch of
companies. I would say, five to ten options now in the market and
they seem to be really taking off. What's so compelling about those
and how does that affect your business?
Ben Golub : Oh, we're sort of complementary. I mean, a lot of those
companies are selling you an appliance which will sit in your data center,
which will make it easier to get data to the cloud and will make the cloud
a storage target. So it will compress it, it will dedupe it, it will optimize
things across the WAN. In our case, what we enable you to do is to
have both your storage and your applications in the cloud running
seamlessly. So they tend to work very well together.
At least, what we are starting to see is that there really are a bunch
of different use cases of the cloud. One is "cloud as back-up"
which is what a lot of these companies are doing, but what we're
finding even more compelling than that is the ability to use the cloud
as a SAAS recovery site. So that you are able to do both compute and
storage in the cloud, as well as using the cloud for busy workloads –
so you need to run a bunch of tests, store a bunch of data and then you
want to tear it all back down again. That is really where the cloud is
powerful and something like Gluster, that lets you have both the
applications and the storage happening in the cloud as well as in
your on premise data center, is really a pretty critical thing.
Jo Maitland: So, the two service providers that you mentioned earlier
that are using the product, how are they? What's the . . .
Ben Golub : So, what they are using Gluster for is -- actually it's more
than that, those are just the ones I can name -- they're using it to create
value-added storage in the cloud offerings. For their customers who
want to be able to run applications in the cloud without rewriting, with
the ability to tie between the public cloud and the on premise data
center, with the ability to have high performance that's the sort of
thing that we do.
Jo Maitland: So the tie-in between on premise and -- how does your
product do that? Because I think I missed that part.
Ben Golub : Okay. So actually, at this point over half of Gluster
customers are actually on premise data centers. Regardless it
sort of looks the same; it's our software on top of commodity storage.
Jo Maitland: Any hardware.
Ben Golub : Then what we enable is replication between the two.
so that essentially lets you have private connected to public or
connected between multiple public data centers or use one as the
primary, one as the backup etc.
Jo Maitland: So, right now you have a relationship with AWS
around S3. Obviously there's a whole host of other service providers.
Ben Golub : Whole host of others. Right.
Jo Maitland: What's your plan for other service providers?
Ben Golub : Well, we recently announced that we are part of
OpenStack. As you may be aware, it's an entire cloud stack,
if you will, that's designed to work both on premise and private
clouds as well as in public clouds like Rackspace, one of the
most well-known sponsors of that. So our plan basically is to be
agnostic. We want to support lots of public providers as well as
almost any hardware provider at this point; because we think the real value
comes in being flexible software.
Jo Maitland: So what are you guys contributing to OpenStack?
What have you contributed so far?
Ben Golub : Our initial contributions have been on the nature of
bug fixes. What we talked about at the last OpenStack conference
was really giving people the ability to have a flexible combination of
both file and object storage in the cloud, to sort of combine the two
different styles of storage into one which is compliant with OpenStack.
Jo Maitland: On this idea again of gateways and sort of connecting
private and public clouds. One of the biggest problems there is the
initial upload from whatever you're doing in house to S3 or to whoever
your cloud provider is. I hear that's painful for all customers really. I mean,
even companies that have a lot of bandwidth but generally most
people are restricted on the pipe and so, that initial. What's going
to solve that problem because I think that seems to be a major
Ben Golub: At the end of the day you have to get a lot of data from
point A to point B, so you can certainly try and compress the data,
which is what a lot of these gateways do but we think the far more
important thing long term is to make it very efficient to keep the two in
sync so that you're not doing a backup in a snapshot and trying to move
huge amounts of data all the time from one place to another, but in fact
you just keep the changes in sync. That's important because, while you
can move a virtual machine from one place to another place pretty quickly,
if you've got petabytes of data that clearly isn't possible over a tiny pipe.
Jo Maitland: I don't know how much you've paid attention to the whole
iCloud announcement but Apple has been building data centers. The
whole iData data center thing. They're popping up everywhere, these
giant buildings. They bought, so the story goes, petabytes of Isilon
scale-out NAND storage from EMC. I'm curious just to get your thoughts on
why they would necessarily pick that. I mean, there are a lot of very
smart people at Apple obviously but it's going to be a very expensive
solution coming from EMC. What are your thoughts on that?
Ben Golub : Well, I rather they had chosen Gluster but besides
that I think that certainly what they get by choosing Isilon is that they
get scale-outs. They get the ability to grow incrementally rather than
buying a large box which is very old style. We've got plenty of customers
who look very similar in terms of trying to build out large centers for
storing large amounts of media, where they have very little ability to
predict how much customers are going to be storing two months in the future,
let alone two years in the future.
Ultimately I think, right now for a smaller open source company
like Gluster, our biggest competitive pressure, if you will, is to
convince people that going with software that's open source, on
top of any hardware, is a better answer than buying a self-contained box.
Some people like one, some people like the other.
Jo Maitland o: what do you think, in terms of the market trend, where
do you think the momentum is at the moment?
Ben Golub: Well, I just look at what has happened in computing
over the past ten years. I think that what you see is that, at the
end of the day Linux beat up Solaris. At the end you see that
commodity x86 clusters have beaten out big iron; you've seen that
scale-out has beaten scale-up; you've seen that cloud and virtualized
has beat out monolithic and un-virtualized and so, I think it's pretty
likely that the same thing is going to happen in storage, especially
since you're combining that trend with the fact that the amount of data
being created is growing by 60% a year. I don't think you can have
data growing by 60% a year and have storage prices only coming
down by 15% a year and have that being a sustainable situation.
Jo Maitland: Are there other open source distributed file systems
like Gluster or are you guys the only ones in the game?
Ben Golub: There are some academic projects that are out there,
but right now a file system is not an easy thing to build. You have to
get a number of things right. When Gluster started a few years back
people thought that we were crazy by doing storage as software, they
thought we were crazy by doing it as open source, and then they
thought we were crazy because of a number of technical decisions
that we made. So fast-forward a couple of years and half the world
still thinks we're crazy and the other half really likes what we're doing
so I guess, if you have lots of people telling you you're crazy, either
you're a genius or you're very crazy. We'll find out.
Jo Maitland: Yeah. What do you think the likelihood is of Google ever
open sourcing their file system?
Ben Golub : Well, they certainly have open-sourced what was called GFS.
There's a lot of the stuff that's associated with the large map produce stack
that they've open-sourced. I think what Google was able to do was that
they were able to take an approach to storage that was very similar to ours,
but they wrote their application specifically for that. So they made
their storage really simple because they have 10,000 of the world's
best engineers and they can make their applications really smart. What
we're essentially trying to do is make that same Google style of storage
available to all the enterprises of the world without requiring them to hire
10,000 of the world's best engineers. That's a hard thing to do.
Jo Maitland: Cool, thanks very much for coming in, Ben.
Ben Golub : My pleasure. Thank you.
Jo Maitland: Cheers. Bye.
Ben Golub : Bye-bye.
Jo Maitland: This has been Cloud Cover TV. Thank you for
watching and tune in next week for more insider news and interviews
in the cloud computing market.