by Kim Borg
These days, it can prove downright daunting to keep pace with the constant strides storage vendors are making in the industry. This is especially true when it comes to flash storage, a technology that has not only emerged as a major disruptor but also spawned an array of products – many of which add unnecessary complexity to enterprise-level storage environments.
Enter Coraid, which has upgraded its
EtherDrive Ethernet SAN platform to support an all-flash configuration called
EtherFlash. This new solution provides high performance, low-latency access to
flash drives sans the bottlenecks and limitations inherent with legacy storage
and Fibre Channel networks. I spoke last week with Kevin Brown, Coraid’s CEO,
about how his company is breaking new ground in the SAN space and the beauty of
being able to offer an enormously flexible “one tier for all” storage product.
Kim Borg: I
know Coraid can mix SSDs, SATA and SAS drives into a scalable, single-tier storage
pool. Perhaps more importantly, Coraid’s SAN technology can handle nearly any
enterprise requirement. Can you expand on exactly what your company’s SAN
technology can handle in terms of enterprise requirements?
Kevin Brown: Coraid has come up with kind of a new type of SAN called Ethernet SAN so, instead of using traditional old Fibre Channel or lower performance iSCSI, this actually uses ‘Raw’ Layer 2 Ethernet. This type of Ethernet delivers performance faster than Fibre Channel at about one-fifth the cost. This is a new architecture for constructing a SAN, one which layers in all of the traditional enterprise SAN capabilities but at a much more attractive complexity level and cost level.
So we handle SATA, SAS and SSD drives and, in fact, we just launched our EtherFlash solution. There’s been a lot of talk about flash drives and how the economics and the performance are getting more attractive, so we launched our solution with our new CorOS 6.0 software, which allows us to get nearly 200,000 IOPS out of a single shelf and do it for less than $10 per gigabyte. This [cost] is tremendously lower than most of the flash array specialists that are coming in at $20 to $30 [per gigabyte].
Kim: Flash storage has become
known as a disruptive technology in the storage industry - especially in the
application areas. With the dizzying array of flash products available today,
what makes Coraid’s EtherFlash the logical choice in your eyes?
Kevin: Today, many companies and startups are coming up with specialized software to take advantage of flash and we think that’s a smart thing. But, what people are struggling with in the storage industry today is that they already have five or six tiers of storage, and adding a new flash array from a startup with 15 customers - they go from six tiers to seven tiers of storage.
What we are doing is something that involves a very different architecture. We’re collapsing all those complicated tiers and we have 100 percent mix-and-match capability. So, with a single storage array building block, we can configure it with all SATA drives for your archive or backup project; we can do a mix of flash and SAS and SATA for your VDI projects. We can also go all the way up to an all-flash configuration with that same array to handle high performance databases and web applications. So there’s a real contrast versus the way the storage industry has built the controller-based storage frames.
We have basically come out with something that gives one tier for all and offers extreme flexibility with configurations, so that’s a big contrast to needing a specialist array from a startup that doesn’t have a lot of experience.
Kim: Let’s talk performance
and cost. Coraid uses massively parallel 10 gigabyte Ethernet and its CorOS
scale-out operating system to deliver performance superior to what high-end
Fibre Channel arrays can deliver. Would you please expand on Coraid’s OS and
just what the EtherFlash solution comprises?
Kevin: One of the things that make it possible for us to get these superior price/performance storage economics is the distributed operating system we developed. If you look at older vendors that have tried to retrofit, it can take years and years and, still, they struggle to turn their old-style storage arrays into a distributed system.
Our CorOS is built as a distributed, scale-out system from the ground up, starting at the operating system level. This has really big implications for how you manage the whole cluster of systems as a single entity and it also has performance implications.
CorOS [is now on] version 6.0, so this is a big advantage for us – that we have a very stable and mature distributed operating system. We combine the layer of innovation on the storage arrays with another layer at the networking level - this is what we call ‘ATA over Ethernet’ or ‘AoE.’
[AoE] has been an open standard and native in the Linux kernel since 2005 and is something that we basically invented and then made available to the community. AoE has been native to allow a Linux server to connect directly to disks over Ethernet and has been used by a number of Asian manufacturers as well as ecosystem partners like Caringo (Editor’s note: Coraid recently partnered with Caringo, an object storage solutions provider, to unveil the industry’s first unified storage infrastructure for scale-out block and object storage).
[AoE] basically provides a simple way to connect servers to storage and has some really important properties. Number one: It’s very simple – the whole spec is 12 pages long, as opposed to hundreds of pages long, [as is the case] with older protocols.
Number two: Instead of being a connection-based protocol [such as Fibre Channel and iSCSI, both of which have a single connection of data that can flow through the network, and which has to be manually configured], AoE provides a massively parallel connection-less way to connect storage. This eliminates the configuration required for storage features such as multi-pathing and port bonding.
Essentially, by going from a single stream of data to a massively parallel set of data over 10 gigabyte Ethernet, we can get a tremendous boost in performance. So, not only is it simpler but it’s less expensive and it’s much faster.
The combination of AoE at the networking level with our CorOS at the operating system level is something that is really very differentiated compared to what the other storage companies out there are offering.
Kim: Can you touch on the
ease of use in regards to the configurability of Coraid’s EtherDrive storage
arrays – especially in VMware environments?
Kevin: Because of the way our SAN operates, we make the entire storage network look like a direct attached disk. In other words, the entire SAN might be two petabytes with no single point of failure, replication and mirroring, but it still looks like your C: drive. This brings a tremendous amount of simplicity to the way that a server or applications will interact with the SAN. The result of this – especially in VMware environments – is that a very large percentage of our business centers on people with virtualization environments.
So, instead of waiting around three weeks for someone to configure your SAN, as soon as you plug in a new storage array, your company’s hundreds of terabytes are now available to hypervisors, which can configure them as if they were configuring their local attached disk. That’s a big deal.
Capital expense and operating expense are a big deal, and in storage they’re also a big problem. In fact, storage [expenditures] today account for 40 to 50 percent of the capex budgets for a lot of folks in IT. We can bring down that capex tremendously because we’re coming in at about one-fifth the cost of an expensive Fibre Channel array.
Operating expense is the next thing people look at. This means: “How many administrators do I need to run it and how hard is it to upgrade or expand?” With traditional systems, when you need to expand, it involves a forklift – literally.
In our case, it’s just plug in another brick and go. So, in addition to capex and opex, though, IT departments are increasingly being changed by consumerization. People go to Amazon and their experience is ‘click, click, done.’ In a few minutes, they can configure a whole system and have computing, networking and storage configured and running. Then they come to work and it takes three weeks. So, the agility that people get out of cloud – that whole self-service idea – is really missing in the enterprise.
So what we’re enabling is essentially the ability to not only save on capex and opex but now to very quickly get your teams to work. It sounds kind of simple, but just the act of having to wait around for weeks to start a project is a big business problem. So, that’s where people have been flocking to the cloud and if IT departments want to provide that level of service, Coraid gives them a level of self-service and simplicity to make that a reality.
Kim: What do you think is the
main draw when it comes to your solution? Is it the price-performance factor or
the simplicity and flexibility?
Kevin: Certainly, price/performance and simplicity are huge pain points in the storage market. But, in the end, our customers realize that [Coraid’s offering] has the most value because it is just a much more flexible architecture for building out their data centers. I would compare this to VMware. When VMware got started, they let you go from 20 servers to four servers and that server consolidation not only had a great business value but also helped out with power, cooling and space. That [capability] started to catch on like wildfire as soon as people realized it was robust enough to run their enterprise applications.
In the end, what has become more important is that not only can you save money – which helps justify the project - but also you now have a much smarter way to operate a data center at the computing level. You’ve got VMotion so you can move applications around, you’ve got replication and you’ve got all sorts of tools that are built into the hypervisor that really have changed the way that you run computing.
That’s what has made VMware a very large and important company and that’s exactly what we’re doing at the networking and storage layers. So they fixed compute, but networking and storage today in the data center are still very broken - especially if you look at the storage networking piece. It’s still running on these old, mainframe-era Fibre Channel networks.
So, what we’re doing is bringing in an architecture that can essentially modernize and create elasticity and flexibility that has never been possible before. This allows people to scale out storage the same way they scale out servers, without having to constantly re-architect and reconfigure a complicated SAN. That, for us, is where people really see a long-term value to working with our company - and it doesn’t hurt that we can hand them back major parts of their budget to start.
company recently acquired Yunteq, a cloud orchestration company. Would you
please share with our readers how Yunteq’s offerings will complement those of
Kevin: Yunteq has specialized in providing policy-based automation for large cloud customers – what we call cloud orchestration – and so what we’ve done, as I’ve just described, is really simplified the architecture for building a storage network and continuing to scale it and operate it and really drive out capex and opex.
We’ve been tracking this company for about 18 months and had known the founders and we saw them starting to get some great traction and we said: “This is some of the functionality we were planning on building ourselves, so why don’t we see if they want to join the Coraid team?”
This is not what your typical storage company does. You don’t always think: “How do I provide automation and tie it into the whole cloud stack?” But this is exactly where IT needs to go and what the Yunteq technology allows us to do. If you look from the bottom up of the stack at the storage and networking and security policies – that’s where Yunteq have really specialized in providing automation. They have a policy engine that allows you to say if you’re running out of room, and then go and create a new container out of these assets. Essentially, this is the same code that Amazon and Google had to write themselves over 10 years, but now we’re going to be offering it in a way that really helps automate the bottom of the stack.
There’s a ton of companies that are doing automation up at the virtual machine layer but there’s no one that’s been able to attack the bottom of the stack because it’s impossible to automate Fibre Channel. The legacy storage stack has five different vendors, it’s got these complicated protocols and provisioning and it’s been a real challenge to turn that into a cloud architecture. But with our Ethernet SAN, we are perfectly positioned to do that and what Yunteq provides on top of that is a policy engine and automation that can not only speed up and drive even more operating expense but also provide an even better level of self service for those types of cloud architectures.
We’ll be unveiling our integrated product set in the first part of 2012.
For more information about
Coraid’s advances with scale-out virtual storage, check out www.coraid.com.
Kim Borg is the Editorial Manager at Computer Technology Review. She can be reached at email@example.com.