Cliqr control freaks apps across many clouds
Google I Owe – and Foundation Capital, too
Moving applications between clouds is arguably not much easier than moving them from physical machines, and there are very good (if disappointed) economic reasons for that.
There will be no shortage of companies that are going to try to make a killing thwarting the proprietary nature of all IT vendors, whether they sell physical iron or fluffy virtual systems. One of them, called Cliqr Technologies, has just uncloaked and it wants you to pour your apps into its cloud blades.
Cliqr was founded in late 2010 by Gaurav Manglik, the company's CEO, and Tenry Fu, its CTO. Manglik did a stint at Oracle writing distributed database code and the two met at VMware, working on cloud controller software that eventually ended up in various parts of the ESXi hypervisor and the vSphere and vCloud Director add-ons to the hypervisor.
The company was operating in stealth mode as Osmosix, and secured under $1m in seed funding from Google Ventures, the venture capital arm of the search engine giant, and Foundation Capital, Manglik tells El Reg.
The company, which is headquartered in Palo Alto, has 10 employees, who have been building its cross-cloud templating and application performance system, which is called CloudCenter. Cliqr has filed three patent licenses already and has three more in the works. It also has four customers in production using the tool already, and is now rolling out the product in advance of raising more cash to push sales and do more development.
CloudCenter V1 is generally available starting on Tuesday, in fact, and it has the lofty goal (you'd expect that from a cloudy tool maker) of allowing a company to onboard any application to any public cloud in a day, and then to manage that cloud-deployed app in such a way that the price/performance is 5X to 10X better than what system admins could do by hand.
"One of the challenges of onboarding an application to a cloud is that a CIO has to make a big bet on any cloud," explains Dave Cope, chief marketing officer at Cliqr, adding that Netflix had a crack team of coders and it still took them two years to move the company's movie streaming systems to the Amazon cloud. "You could spend millions of dollars porting to a cloud, only to find out that it costs more and performs worse."
Manglik says that early cloud adopters have in some cases seen a horrifying factor of 10 decrease in bang for the buck on cloudy infrastructure, which is obviously unacceptable to those who work in the paneled offices.
The CloudCenter tool is designed to take the grief away and squeeze the most out of cloud infrastructure for the least amount of money. In other words, it is designed to replace a lot of the human beings who are running the infrastructure plumbing today in the data center.
It is also designed to make applications portable across many clouds, alleviating the vendor lock-in that is one of the reasons why cloudy infrastructure is not as inexpensive as it might look on the front end of a deal. Once a cloud vendor knows you are in for a penny, you are in for a pound.
The secret to cloud portability is – yes, you guessed it – some more layers of abstraction. In this case, Cliqr has created an abstraction of public clouds and cloud fabrics that are commonly used in private clouds; this abstraction layer codifies the best practices for that cloudy infrastructure as it relates to compute, storage, and network capacity and security for these three layers.
Cliqr currently has cloud blades for the public clouds run by Amazon, Hewlett-Packard, and Rackspace Hosting, and also supports OpenStack; the company has a CloudBlade SDK that will allow it to quickly add support for other clouds and also lets companies create templates that encapsulate their homegrown or ISV applications and move them to clouds as well.
Block diagram of Cliqr CloudCenter
CloudCenter also includes an abstraction layer that runs atop the CloudBlade called the Cloud Orchestrator, which is common across all clouds and which normalizes all of the compute, storage, network, and security processors for the application. It is also the tool that you use to watch how your applications currently run internally on your data center; information from your running apps is used to package up those applications, which are stored in a repository, so they can be onboarded to various clouds. You keep this repository up to date, and when you want to jump to a new cloud, you let CloudCenter do the migration.
On top of this runs another set abstraction layer called App Orchestrator, which provides specific templates for each type of application architecture – is it high volume batch work, a Web application, a Hadoop or other kind of cluster, multi-step workflows, or another kind of parallel environment, and so on.
These templates are designed so that each kind of application is pre-tuned for optimal performance on a specific cloud type. So, for instance, it knows what dirty tricks you need to deploy a Hadoop cluster on Amazon versus Rackspace based on the hardware and configuration settings on those two different clouds.
The App Orchestrator templates are written with a mix of software images and scripts written in a mix of Ruby, Google Go, and Java, while Cloud Orchestrator is written with tools specific to each public cloud and its API stack.
When Cliqr says it can onboard any app to any cloud, that is a bit of a stretch, of course. Manglik says that in most cases, customers do not need access to their source code to onboard their applications, but he admits that they have not yet done a big ERP suite to prove this works on real enterprise-class apps that are both large and hairy.
Also, the CloudCenter tool only works applications that run on x86-based servers and only Linux or Windows applications can be puffed out onto public clouds. "There's nothing in the architecture that limits us to x86 or any particular operating system," says Manglik. "In fact, we have some early customers that are looking for us to help them support legacy applications."
The CloudCenter software it itself a SaaS product, and it does, of course, run on a bunch of different public clouds that have been hardened and cross-coupled for high availability. The software gives you a single pane of glass to manage all of the apps running on various clouds, including graphical dashboards.
The price/performance broker in CloudCenter V1 allows you to upload a sample workload onto a cloud or the full suite and benchmark and tune it, but in the V2 release Cliqr will allow predictive pricing and performance for the public clouds that it supports.
Cliqr is being a bit cagey about pricing for the CloudCenter tool, saying that production environments will have license fees related to the compute-hours used, with volume discounts. Exactly what price per unit of compute-hour is a mystery. But what Cliqr is being absolutely upfront about is that it will do application onboarding for free. "We are removing all barriers to adoption," boasts Manglik.
And giving customers the option of doing a migration to the cloud quickly. One customer, who spent two months moving an application to the Amazon cloud installed CloudCenter and was able to do a port to in 15 minutes.
If CloudCenter, which is having its formal unveiling at the Google I/O event in San Francisco tomorrow, works as advertised, it would not be surprising to see Hewlett-Packard, Dell, or IBM snap it up – or VMware, Citrix Systems, or Microsoft to do so to tie the technology to their respectively cloudy fabrics and sell it as a service rather than a product. ®