Introduction
Launched in recent weeks, Juno, the 10th version of OpenStack, brings a slew of new features and even more bug fixes.
According to Jonathan Bryce, executive director of the OpenStack Foundation, Juno's successor, called Kilo, will add functionality based around managing bare metal platforms, helping telcos save costs using Network Functions Virtualization (NFV) and maximising performance out of high-performance applications. Thanks to OpenStack's bi-annual release schedule, Kilo will become available in April 2015.
We caught up with Bryce at the OpenStack Summit in Paris to discuss issues ranging from downtime to staff training, switching from commercial to open source solutions and what it means to be a platinum member of the OpenStack community.
TechRadar Pro: Can you take us through some of the big features in Juno?
Jonathan Bryce: Versus the initial releases, Juno is interesting becuase it's focused on operations, and things that make it easier to build, manage and scale an OpenStack cloud.
There were 10 times as many bugs fixed as there were features implemented in Juno. In the previous release, Icehouse, there was a feature that allowed for rolling upgrades in an environment that basically let you upgrade an environment over time while you moved workloads.
Juno enhanced that upgrade capability so that you can do live upgrades and upgrade your cloud underneath the running workloads. That's big for people running it in production. There are also little details that come directly from user feedback. We love to get users to the summit and speak, because we have more developers here than we ever get in one place. We do that every six months.
TRP: Can you name any features that have come off the back of user feedback?
JB: One is Rescue Mode which is in Juno. If you run a data centre and a server crashes, in the world you would connect a monitor and keyboard to it to see errors on the screen saying what went wrong. A disk may have crashed, but you don't want to reboot the server becuase you might lose data. In the virtual world you can't go in a data centre and connect a monitor to the server, but with Rescue Mode, you can remotely connect to the console of a virtual machine.
OpenStack has had that for a very long time, but operators told us that they need to be able to boot a disk rescue image and safe boot different kinds of things, so there's a ton of work around that in Juno.
Another interesting component in Juno is for doing data pocessing-as-a-service, which is basically MapReduce as a service. It's a system that supports Apache Hadoop, Cloudera Hadoop, Hortonworks' Hadoop and SPark. It automates thin provisioning and scaling of these kinds of workloads on top of an OpenStack environment, which was a very user-driven feature and something that came out of big data, which is a really common use case for OpenStack.
TRP: What will developers try to achieve with the next release, Kilo?
JB: We'll once again see more operational updates. There'll also be something on standardising how logging and configuring is done across different pieces of OpenStack. In terms of new capabilities, there's a bare metal management feature that's going to be in Kilo that has support for different bare metal platforms.
Basically, you can have a pool of servers and image them directly without a hypervisor. So for high-performance applications where there might be some specific type of hardware that you need to access - such as a GPU for rendering - there is a new set of workloads opened up that can be managed under OpenStack.
Talking NFV
TRP: There's a lot of talk around Network Function Virtualization (NFV) at the summit. What's that all about?
JR: It's similar to software-defined Networking, or SDN, which is the idea that you can control the flow of traffic on the network. NFC is a sort-of subset of that, but it's also an application on the network.
Suddenly everybody has a smartphone, tablet and laptop that can be connected to a hotspot, and LTE demand for data goes through the roof. NFV is about handling that demand, and it's something that went into Juno.
Telcos have to go to all of their datacentres and install new pieces of hardware, such as new types of routers, but they have no way of really changing the format of what gets delivered in each data centre. A cell tower has a type of mini data centre at the bottom of it, and you also have switching stations. NFV is about taking all of that purpose-built hardware that telcos buy and get stuck with for 10 years, and running its functions on standard hardware while being able to change their make-up.
TRP: Why might NFV be a big opportunity for telcos?
JB: Telcos have had to make predictions about where the market is going for the last 10 years. Back then people probably a sent a lot more text messages than they do now, as it's now it's all about Whatsapp and Facebook, etc.
That's shifted the load from SMS to data, which is bad for telcos as they bought a ton of SMS routers that they thought were going to last for 10 years but aren't being used. If those data centres were filled with servers and they ran those functions virtually, they could shut off machines where SMS load is falling while turning up some LTE data machines remotely. That's the goal.
However, it's a trillion dollar industry and that's not going to happen in the next six months or even year - it's going to take several years as lifecycles turn over. It's going to completely change the way that telcos operate their networks and respond to changes in the market.
TRP: How does NFV relate to OpenStack?
JB: OpenStack is software for managing infrastructure components, and enterprsies often have them centralised in a data centre, or in a few datacentres. A telco has those components spread out all over the place, and they're in much smaller data centres that, when combined, are treated as a cloud in the way that they orchestrate their services.
Telcos tell us that when they look at their resources, they don't want to see any difference between a cloud that's 100 servers in a single location, and a cloud that's one server in 100 locations. That requires a different model in terms of how you manage and track resources, so the updates that have been going into OpenStack and Juno will appear in the next version, Kilo.
The other point is that, if you're going to route LTE data, the whole point of 4G and 5G networks is that they're meant to be really fast, and if you have to go through several layers of virtualization, with each layer downgrading your performance a notch, then you're kind of missing the point. So a lot of hypervisors and harwdare have got to the point where there's a much greater ability to pass that raw hardware performance all the way up through the hypervisor and virtualization layer into the application.
It means that telcos can determine high performance applications and dedicate particular processors, or cores of processors, to applications, which is something we will be working on across several releases to meet some of those needs. It's cool because it will also help enterprises with high performance processing needs or distributed workloads - it's a massive opportunity.
How can OpenStack help companies disrupt industries that have been typically hard to crack - such as automotive and banking?
JB: The thing that's interesting when you look at really well-established industries is that they often have a few leaders that are really big. Industries like banking and automotive are often highly regulated too, so the incumbents face an environment that has a lot of rules about what they can do and how fast they can do things.
So when somebody comes into the digital currency market, for example, or e-payments where there are no laws because it didn't exist a couple of years ago, they're able to move much quicker and respond to customer demand. Being large used to be an advantage because you would build a castle and stay in it away from the people with pitchforks, but it doesn't really work that way. A castle is no longer is an advantage - it's better to be able to move quickly and go as fast as you can.
Security, training and making the switch
TechRadar Pro: The Heartbleed bug showed that there can be serious consequences when bugs in open source software are exploited. Did that event change how OpenStack developers approach security?
JB: Sure. Heartbleed was a very big vulnerability. The team that's responsible for OpenSSL has really smart guys, but they didn't have a huge support network around them, the type that allows you to dedicate the resources you need.
On the other hand, OpenStack has a massive community and a dedicated security team, along with companies that spend millions of dollars to test and develop on it. From the foundation's perspective, we make sure that we help to put the frameworks and systems in place to keep those groups operating, functioning and sharing information.
TRP: Does that involve things like cross-checking code?
JB: When a contribution comes into OpenStack, it's a really cool process. Following automated tests, two core reviewers have to approve the contribution (or patch), which then re-enters a testing environment to check that nothing has changed in the time that it has been reviewed. If that all works then it finally enters the source tree. It's a very robust system. During the six-month release cycle for Juno, that testing system ran 2 million jobs in that release process.
TRP: Something that has been raised frequently at the summit is the lack of staff training and people skilled in OpenStack. What is the OpenStack Foundation doing to improve education around the platform?
JB: I think that's just a sign of a rapidly-growing market; that's how it is in new markets at the beginning. Given time it'll all balance out, but we want to get there sooner. We launched a training market last year that aggregates training courses from companies all over the world. Hundreds of those courses have been delivered in a year in something like 60 cities in 20 or 30 different countries.
We're doing more to promote those kinds of training opportunities. In addition, we have something like 70 user groups around the world that do hackathons and community training, getting people together to learn OpenStack.
TRP: Another issue raised is the difficulty users are facing is downtime when upgrading OpenStack. Mirantis has claimed it is going to fix the problem. What do you think to that?
JB: Mirantis is a company that does pretty much everything upstream and push their stuff into the community. We love it when companies function like that becuase everybody is able to benefit from it.
TRP: Intel is now a platinum partner - how does that change its standing in the openstack community?
JB: It's more the other way around - they were elected there because of their standing. This is the first time that we've had an opening for a Platinum position, and the board has to elect a company to that spot. It's all about their standing in the OpenStack community and how much they're contributing to help both drive the code but also organise their related efforts, drive it back into the development process and make the software more useful. Those are some of the big things that were key in them being elected.
TRP: Why might an organisation move from a commercial solution to OpenStack?
JB: There are a number of companies who have moved some of the proprietary systems in their environment to OpenStack. It's interesting because OpenStack has the broadest support for the most technologies in the date centre.
We've heard that a really valuable piece of the OpenStack equation for a lot of companies is that they have a load of applications that run really well on VMware, so they don't want to get rid of that, but at the same time they want to have the option of using [open source hypervisor] KVM for new workloads, or workloads that don't require all of the functionality of VMware.
I think that sometimes it is about cost, and they want to cut licensing fees, but other times it's about flexibility and having the opportunity to choose and not being stuck on a single vendor's product cycle forever.
from Techradar - All the latest technology news http://ift.tt/1qyqYK2
沒有留言:
張貼留言