Opscode GA’s its offerings. It’s time for it to start some automation knife fighting.

Opscode, the commercial company around the automation project Chef, had a bundle of what I’d call “1.0″ announcements today: firming up their product offerings for general availability:

• They’ve fully baked their hosted Chef suite, renaming it Opscode Hosted Chef (née Opscode Platform).
• They released Opscode Private Chef for all those folks who can’t stomach running their automation suite in the cloud.
• You can also see them staking out a claim in the next generation IT management space with their “Cloud Infrastructure Automation” term, used to describe running IT in a cloud-centric and inspired way, e.g., by using Chef.

Also, there’s Windows support, which is sort of a minority interest at the moment but damn fine to start hammering out now.

But does anyone care?

There’s plenty of momentum for Opscode as their numbers-porn slide shows:

Segment Context & Such

Indeed, I continue to see interest in Chef, particularly from developers and ISV types – see their Crowbar partnership with Dell, putting together OpenStack, Chef, and Dell hardware for quick-clouds on the cheap. Check the small Opscode partnering mention in El Reg’s piece on Calxeda today as well.

I tend to hear more interest in Puppet from IT types (see coverage of their offering here), something the Private Chef offering might help address. IT folks have been skittish about using cloud for their software, and, why not? As one admin told me last year, “well, if the Internet goes down, I’m dead in the water,” he has no tools. And, despite the fact that you’re probably dead in the water in all cases where “the Internet goes down” and that ServiceNow seems to be doing just fine, that’s some FUD that doesn’t deserve much scorn.

On the broader front, I’m still not seeing much regard from the traditional automation vendors for these model-driven automation up-starts, Chef and Puppet. But they should be paying attention more: both are classic “fixing a moribund category that sucks” strategies that seem to be actually working in removing The Suck by focusing on speed.

Downloadable PoCs

And if you’re on the other end of the stick – buying and/or using automation software – checking out these new approaches is definitely worth your time. The reports are sounding similar to the early days of open source, where the CIO is stuck in multi-month PoCs and license renewals, while a passionate admin somewhere just downloads Puppet or Chef, does a quick-n-dirty PoC, and then gets clearance to take more time to consider these whacky, new methods.

More

• My coverage of the initial, beta launch of The Opscode PlatformOpscode Hosted Chef.
• Rich Miller’s coverage.
• Stacey Higginbotham over at GigaOm write it up.
• David Strom at RWW covers another Opscode item, their QuickStart bundles of common stacks – more developer targeting.

Disclosure: Opscode and ServiceNow are clients.

A new startup around the scala language launched today, Typesafe it’s called. James wrote up a nice piece. Here’s my brief not from chatting with them:

Scala has steadily risin as a new, parallel-friendly language in recent years and the creators of it are forming a company around it, Typesafe, funded by Greylock in the first round for $3M. (See the official press release for more details.) Scala and the various platforms around it are used primarily to enable high scale applications – things that need to process a lot of changes all at once: this ranges from financial applications analyzing various market scenarios to social applications like Twitter and Foursquare: from investment and risk decisions to updates on sandwiches and cats.

Overall there’s a positive, if only curious view about Scala from many developers. One developer I spoke with said simply, “Scala? Heard it’s good.” And this point I mean as a positive one.

On the business side, Typesafe is starting with the standard list of offerings: certified stacks, developer tools, training and certification, and consultative services. As with most open source companies now-a-days, you’d expect to see them offer products around the core stack: management, reporting, and other “support” products.

There’s another angle to consider on the business side. Companies like Cloudera seem to be monetizing open source under slightly different conditions than the past round of OSS money-makers. A Big Data sale is more about enabling the business to do something new (analyze more data, more fast, for example) which theoretically motivates non-development centric people to seek out a “product,” or “solution” to buy. Contrast this with, say, application server sales where you’re trying to motivate development teams to spend money on something or, at best, operations people who want certified stacks and management tools. If people like Typesafe and others in the cloud-big data-high-scale-changes-everything market can convince business buyers that they can buy IT that improves their business, there’s a slightly different market there than in the past. We’ll see.

Disclosure: Cloudera is a client, as is Lift.

Brief notes are summaries of briefings and conversations I’ve had, with only light “analysis.”

The venerable Toad database tool line launched a “cloud” version last year, allowing users to work with NoSQL and cloud-based databases such as SimpleDB, Cassandra, SQL Azure, and Hadoop among others. In the relational database world, Toad has always been an good choice for messing around with databases so it makes sense for Quest Software to extend into the NoSQL world.

While I still don’t feel like there’s massive “mainstream” adoption of NoSQL databases, interest in new types of databases (“NoSQL” for unprecise shorthand) is certainly high and there’s enough “real” uses in the wild. RedMonk has certainly been fielding a lot of inquires on the topic as well and in-depth research notes on selecting NoSQL databases for various clients.

Thus far, Toad for Cloud Databases has 2,000+ “active users,” which is pretty good given the level of “real” NoSQL usage we’ve been anecdotatly seeing at RedMonk. As Christian Hasker (Director of Product Management) said, Hadoop tends to lead the pack, and then there’s a “sharp drop-off” to other database types.

In addition to tooling, Quest is building itself up as a “trusted voice” in the NoSQL-hungry world with community efforts like the NoSQLPedia, which actually has been doing a good job cataloging all the new databases, as in their survey of distributed databases.

For Quest, it of course makes sense to chase tooling here. They’ve maintained a huge install-base for their relational database tools and as new types of databases emerge and become popular, keeping their community (paying customers and non-paying users) well-tooled is important. Also, applying my cynical theory of “make a mess, charge to clean up the mess,” the rest of Quest has and could have plenty to sell when it comes to managing all those “cloud databases” in the wild. As an early, non-Quest, example of a janitor here, we’ve been talking with Evident Software of late about their the NoSQL support (for example, Cassandra) in their ClearStone tool for application performance monitoring.

Disclosure: Cloudera is a client, as are some other “NoSQL” related folks.

“i” is for “integration”

I like this narrow focus on integration instead of just a general PaaS. It’s what you’d expect from MuleSoft’s background is in ESBs (Mule), Tomcat support (they seem to have built a new business around their “Tcat” server), and the open source world around the kind of tasks and applications (Java, mostly) you build around a bus.

Cloud Creep

I spoke with Ross Mason and Mahau Ma today, going over the announcement but mostly about the types of applications and work-loads they’re seeing people use in cloud-based projects. Some key take-aways:

• From the customer/user point of view, everything is wrapped up in iON. While it runs on a public cloud, the end-user interacts with the iON platform without needing to go muck with the underlying cloud.
• Last year, MuleSoft started providing Mule Cloud Connect (see an interview on the topic here) to help integrations between cloud and on-premise (and cloud-to-cloud, I guess). With all of the SaaSes out there (and Salesforce in particular, it seems), Ross said there’s been a big demand for integrating data between services. As he put in on the Mule blog a short while ago: “One of the drivers for Mule Cloud Connect was that we are seeing that traditional three-tiered application architecture is outmoded – Web applications today require integrating multiple data-sources and services, both in the cloud and behind the firewall, and presenting rich data to the browser in real-time.”
• This need for integration hits on something I’ve been starting to see: cloud-use is infectious in your architecture and even how you run your project. I’ve had a handful of conversations with people who’ve mentioned that once they use cloud for one part of their application (if only delivery), using it for more parts (all the way down to build and ALM) starts to look more attractive. To build on Ross’ comments from that blog: if you’re integrating your on-premise application with data in Salesforce, chances are good that you’ll start to wonder why you don’t move that on-premise thing into the cloud as well.
• As ever, Conway’s Law probably fits especially well here: you’d expect remote, distributed teams that are not centralized to build architectures that are equally “cloud-y.” For example, you could see how virtualized teams who work on cloud-deployed applications would really like moving all their version control to the cloud as well to places like GitHub.

And, indeed, that last point and the resulting transition is what Mule looks to be going after.

Disclosure: both MuleSoft and Salesforce are RedMonk clients, as is GitHub.

ExtraHop sells a hardware appliance for Application Performance Monitoring (APM) from the inside of the firewall, monitoring network traffic to find applications slowdowns and other problems, firing off alerts to admins after diagnosing these problems. If you’re interested in hunting down alerting on poorly performing application and diagnosing what’s wrong with them, these guys have something to offer. Recently, ExtraHop announced a partnership with Keynote Systems: Keynote’s web transaction monitoring would observe web apps from the outside of the firewall, while ExtraHop would do the behind-the-firewall, “back-end” monitoring. See their blog entry on the partnership as well.

There’s two appliances available, the EH2000 and the EH5000 which, as you can guess, do less and then more amounts of analysis, gated by gigabits (3 then 10) of traffic and number of devices (300 then 1,000).

Some key take-aways:

• Most people are skeptical of “magic” performance boosts from APM appliances like this. But, because of the ease of getting up and running – you essentially plug the ExtraHop pizza box in, it auto-discovery different applications and nodes, and then goes to work monitoring and alerting, pulling the info it needs from network traffic – a two week or so PoC is pretty good at winning people over.
• To that point, ExtraHop’s Jesse Rothstein alluded to how “it works better” is possible for the ExtraHop devices: in addition to the software they write, advanced in multi-core and cheaper storage allow them to do real-time network analysis that wasn’t possible previously.
• SaaS providers who, of course, need good APM for their network delivered applications are good customers for ExtraHop.
• Check out the brief write-up of Alaska Airlines use from Beth Schultz.

Puppet Labs released a major product last week, Puppet Enterprise, an integrated bundling of the open source automation platform, Puppet, and several of the other tools needed to get up and running fast. The Puppet Enterprise release is firming up Puppet Lab’s efforts in the mainstream world of enterprise datacenters and IT.

While they have plenty of cases of round-corner cool kids using their stuff – Zynga, Google, etc. – the majority of customers (as distinct from “users” who may or not pay) for any IT Management space are far from anything thrilling that ends in “.com.” I say this as someone who actually covers this space and likes it: personally, I don’t think it’s “boring,” but it’s not iPhone vs. Android, if you catch my drift.

Still, you out there know, dear readers, the model-driven automation space that Puppet helped pioneer is actually very exciting: there seems to be genuine innovation that leads to better IT Management going on in that market.

Polishing Puppet

The idea of Puppet Enterprise is to bundle up SLA’ed support options along with a certified, bundled stack of all the stuff you need to get up and running with Puppet. Having this “polished,” productized version of the open source platform is welcome, of course. It’s one thing to download the open source bits and cobble them together for free, and a nicer, though not-free experience, to have things all arranged for you.

The promise of Puppet and other next generation, model-driven automation tools on the market now is to speed up the time it takes to provision and update IT resources. Get new servers running, manage the software on desktops, and otherwise get a machine from bare-metal up and running with all the software and configuration needed to start using it. When talking with Puppet’s CEO on this release, he told me that one customer had seen Puppet made their configuration rates 53 times faster. If you can speed up the process of configuration management that much, you, the IT folks, buy yourself a lot of time to (a.) do other, more important things, and, (b.) amaze The Business at how fast you can actually spin up new resources for their use.

Closed source conversions, but open source-y developers types are interested in Chef

Pricing is for a 12 months subscription and starts at $2,475 for basic support and 25 nodes ($99/node), goes up to $215,950 for 2,500 nodes and 24/7 support ($86.38/node), and then goes to “call us.”

Momentum

I’ve been covering Puppet, and ReductivePuppet Labs for a long time, since it was a two man shop. They continue to have good momentum and, anecdotally, it seems to be growing in the enterprise space. Here, the projects are to replace older, more expensive, automation suites from BMC (BladeLogic – by coincidence, El Reg reports today on a recent cloud-y win for Blade) and HP (OpsWare). Indeed, Luke Kanies (Puppet Labs CEO) went over a few cases from customers who’d converted, usually following the bottom-up approach to open source: while going multi-month evaluations and bake-offs for closed source automation suites, a admin would try out Puppet and have everything up and running before the ink was dry on the PoC (a dash hyperbole on my part there, of course).

One of the interesting enterprise-y side efforts of a model-driven automation approach, Luke said, is getting a much more accurate CMDB, or at least asset database: because the tool (here, Puppet) strictly enforced the configuration each node has (the “model”), you sort of know exactly what each node has on it. Drift will occur, of course, but that’s what the “enforce” part is about. No doubt, this is not perfect, but I suspect it’s “more perfect.”

On the other hand, and also anecdotally, I’ve noticed that some developers and other round-corner cool kids have been evaluating Chef to replace their Puppet installs. There’s been interesting ISV and cloud vendor interest in Chef vs. Puppet as well. To some extent, this is the non-paying customer pool (versus “enterprise” who pays if there’s something to buy, like Puppet Enterprise), but it is a challenge for Puppet Labs.

More

• Colleen Miller over at Datacenter Knowledge covers the release.
• Official press release on Puppet Enterprise, including exciting quote from yours truly.
• We’ve done several Puppet user profiles: Slideshare, RedHat, Google, and ShopZilla. Each goes over how the organization is using Puppet, how they introduced it to IT, and how it’s working out.
• Episode 17 of DevOps café – a good interview with Puppet Labs CEO Luke Kanies, including history, how it works, and current news and happenings.
• There’s a few RedMonk TV videos with Luke: from 2007, an overview of what Puppet does and a brief update on the company from August of 2009.
• A good wrap-up of Puppet in 2010 from Puppet Labs.
• Puppet versus Chef – an old post on the topic, but a lively comments section that’s kept going (last comment in Dec, 2010) with plenty of back-and-forth between the two communities.

Disclosure: Puppet Labs is a client, as are several of their competitors such as OpsCode, IBM, Microsoft, and some others.

Nitrosphere promises to increase the WAN performance of SQL Server databases with technology that fixes inefficiencies in the communications stack (the NitroAccelerator product). These same, or similar, enhancements are also used to do better compression and backups of SQL Server databases (the NiroCompressor product). There’s also the potential for performance enhancements like this to degrade performance so, as Mark Wright said, during the briefing, they check in on performance to avoid being more of a problem than a help. The scenarios here are having distributed SQL Server databases, over different, perhaps far-flung locations, or just having remote workers who need to access a SQL Server database over the network: perhaps insurance agents working from home, Fred Johannessen said. Pricing starts at $2,499 per server with some clients free.

Clearly, there’s good prospects in Azure and other cloud-based SQL Server applications that are look for quick and easy performance boosts. Early cloud developers often complain about network latency, even within nodes “in” the same cloud. Having a quick way to speed things up is always nice. As James mentioned during the call, the vast world of SharePoint has to be some sort of alluring hills as well.

The area of APM can be a shifty one – monitoring end-to-end transactions and application workflows can be technically difficult as you move across tiers, collect together un-integrated data sources, and otherwise try to track what went wrong between the user’s mouse click and the web of infrastructure the application is spread over. Solarwind’s MO, thus far, has been to provide easy to access tools (downloadable and self-installable) at affordable rates. The tools aren’t free, but they’re low enough barrier to entry that many admins who, for example, use the free Spiceworks suite also use Solarwinds. You find Solarwinds tools paired up a lot like that. See this 2009 RedMonkTV video with Solarwinds Josh Stephens for a quick, if older, overview.

While discussing the APM release with them, Solarwinds noted that APM tools are typically expensive, and perhaps exotic to their customer base. Their hope, then, is to do what Solarwinds does best: commoditize tools just enough to get wider access, but not too much that there’s not enough good business in them for the company itself. In the IT Management space, that’s a welcome trend for most (compared to the development middleware space where barbs of not open sourcing and, thus, totally making free the application would be thrown about…in past years at least).

Speaking more broadly, adding in APM and buying Hyper9 gives Solarwinds the portfolio needed to start making the claim that it’s a general IT Management shop. While Hyper9 hasn’t been a break-away success (compared to, say, Splunk), their technology has always been interesting: their depth in understanding virtualized data centers, searching, and then reporting over that lump of IT has always seemed good when I’ve been walked through demos in the past. If Solarwinds can (even lightly) integrate together their strong base in network management, storage, APM, and now servers/virtualization, they’ll be a compelling portfolio at, no doubt, competitive price points. That portfolio gardening and integration is key: Solarwinds has always been a good bucket of tools and as the company ages, they have to keep their portfolio from becoming a big katamari ball of fun that you’re destined to roll around for eternity.

Disclousre: Splunk is a client, as is Spiceworks. Solarwinds has been a client in the past.

The long road to adoption

While OpenStack was launched last year to much fanfare and attention (see the interesting, public community tracking they do), parts of the open source project is still in development so there haven’t been uses in the field (other than at Rackspace) of the stack. While Internap’s use is just of one part of the Open Stack, it’s a “mile stone,” as they say for the highly regarded stack…that hasn’t seen much production use as of yet.

By comparison, Eucalyptus which is often seen as the market-share foe of OpenStack recently released a press release saying “one in five of the Fortune 100 started a Eucalyptus Cloud in 2010.” Eucalyptus being bundled in Ubuntu certainly don’t hurt either.

When launching OpenStack, Rackspace admitted that they wanted to get the project started before the product was fully finished to get more people involved in the community – a move RedMonk thought was good for open spirit. With plenty of business in hand, Rackspace is looking at the long-term play for becoming “The New Linux” as El Reg so graciously dubbed the project in a rare charitable moment of headlinery.

Build a cloud business

I spoke with folks from Internap and Rackspace last Friday. Internap’s Scott Hrastar said they’d been working with the OpenStack crew for the past 4 months to get this offering up and running, integrating to the Internaps back-end billing and customer systems, I’d guess, given that those systems are not part of the general OpenStack offering. He also said they had some help from some of the original Nova/NASA coders.

Scott said part of the appeal of OpenStack was the ability to build differentiating services on-top of the basic offering – presumably, the open source nature will allow Internap to do this more easily than with proprietary offerings. For existing hosting companies like Internap, “cloud” brings a major fear of being “dumb infrastructure” (akin to the “dumb pipes” and “stupid networks” telcos fear becoming): swappable IT services that have no way of differentiating other than (low) price. While we didn’t go over what these unique services would be, presumably Internap things it can build additional value on-top of the default Object Storage offering, helping them gain and retain customers.

More

• Nancy Gohring at IDG coveres the announcement, including: “There has also been strong interest from financial services companies, he said. They tend to have very large operations so the deployments take time, he said. OpenStack expects to make more announcements about such large users in the coming months.”
• Cade Metz covers the story for The Register, adding: “Internap has not actually contributed to the OpenStack project, but it intends to do so. It’s also evaluating the Nova codebase, but at this point, Hrastar says, the company has no firm plans to offer a public service that serves up processing power.”
• The official press release.

Disclosure: Rackspace is a client, as is Eucalyptus, VMWare, IBM, Cloud.com, Microsoft.com, and many others working in this space.

CloudBees has purchased Stax Networks (see their write-up) to build out their ambitions to become the leading Java PaaS. Thus far, CloudBees has been known as the Hudson in the cloud company, running the continuous build tool in the cloud (on Amazon) for it’s beta users. Doing a build in the cloud is one thing, but tooling all of the activities around the build-test-deploy-run-repeat cycle is a bigger pie to eat from.

(As a minor note, by “Java PaaS” I mean “any VM-based language,” not just the Java language.)

Cloud ALM

There are many efforts underway to do “cloud ALM” though, wisely, no one calls it that. “ALM” has long been thought of as more of a bureaucratic-hell for developers than something useful. Still, for any sane organization the checks-and-balances and quality-through-process that ALM drives towards is required. It’s one thing for some slick .com to eschew ALM, but all those Toyota owners out there probably appreciate the mounds of process paperwork that ALM and higher paper-pushing practices stocked up.

As the diagram above shows, CloudBees is looking towards Stax to help them fill out the “deploy to production” part of that cloud ALM vision. I’d suggest that this “production” (the running of the code) is the area that needs the most innovation in the cloud space and is, therefore, the most difficult nut to crack. The idea of push button deployments to production is great, but the actual reality of deploying, diagnosing problems, rolling back, and so on get messy. And, indeed, solving those problems is (or should be) the value that a PaaS brings.

As a thought-experiment, I’d suggest that the ultimate PaaS would allow you to fire your entire ops team (as needed by the applications in that PaaS, at least) and just rely on the developers to run all that. Whether that’s a good idea or not is yet to be seen – I doubt developers want to strap a pager to their belt every night.

In essence, you need “fully automated provisioning” as some of the dev/ops crew would describe it.

It’s getting crowded

As a comparison, the Code2Cloud crew is tackling the problem from a tools perspective, where-as folks like CloudBees are going for an approach to build and own the entire ecosystem, not just service it. Another example: you can see this general idea being worked out in the mobile space by PhoneGap/build. I come across “build in the cloud” folks all the time now-a-days, and I expect to see even more of vendors moving the entire software development and deployment tool-chain into a cloud – if only “cloud-like” (CloudBees said they have an on-premise version for the weak-kneed) – environment.

More

• Announcement page from CloudBees and, also, Post from CloudBees’ Sacha Labourey on the topic, including: “Stax Networks was the first company to deliver a full fledge Java PaaS back in 2008. Since then, more than 3’000 applications have been deployed to it, with some customers such as Lose It! running more than 5’000 transactions a minute on it.”
• Paul Krill reports on the announcement for IDG.
• Coverage from Scott Kirsner on November 29th, 2010 of CloudBees’ $4M funding and plans.

Disclosure: CloudBees, TaskTop, and VMWare are clients.