Caringo Gives Structure to Unstructured Data
Early last year, CMWare, which creates technology for serving and sharing rich media (music, pictures and video) over the Web and through mobile devices, was in the process of developing a new service for mobile operators called skyTuneZ.
The goal was "to create a centralized hosted facility where people could have their private lockers upload media that they wanted to have mobile access to," said Steven Ott, CMWare co-founder and vice president of business development. To do this, CMWare needed a primary storage solution that could quickly, easily and cheaply scale as people took advantage of the service.
Initially CMWare had considered outsourcing storage to Amazon's Simple Storage Service, but management liked the idea of keeping everything in house. Then, while speaking with a mobile operator about storage issues, CMWare heard about CAStor, a third-generation content addressed storage (CAS) system developed by Caringo, a startup that launched with much fanfare two years ago. With its ability to run on any x86-based architecture (the software is hardware-agnostic) and to scale as needed, CAStor made CMWare's decision to keep its assets in house a whole lot easier.
Plug and Play
CMWare installed CAStor, which typically comes on a USB memory key that just plugs into a server node, late last summer. After a slight hiccup related to the wrong software release, a problem that was quickly resolved, the system has been running smoothly.
In the six months since installation, CMWare has not had to add additional storage, but Ott said the company is prepared to do so and do so quickly after mobile operators start offering skyTuneZ to customers this March and demand significantly increases.
"One of the things that was attractive about CAStor was its ability to scale is very simple," said Ott. "You can just pop in new servers, plug in a new dongle, and boom, you're up. We're a small company. We're conscious of our budgets and we want to scale as necessary, and not be late in doing that but also not build too far ahead of the curve." Caringo's CAStor allows CMWare to do that.
While CAStor is not inexpensive (base price is $1,500 for 1TB), Ott felt that because of its scalability, data protection (CAStor allows you to replicate to a local cluster or across distance for disaster recovery and failover), speed ("When you go into your locker and click on a song and listen to it, you want to be listening to that song streaming to your PC within a second or two," said Ott), reliability (it automatically addresses hardware and software problems) and flexibility (it resides in house on x86 servers), it was the right storage solution for CMWare.
More Than an Archiving Solution
Interestingly, while CAStor was originally designed and developed to help organizations archive unstructured data and meet regulatory compliance requirements, it has garnered a lot of interest of late from companies like CMWare, which are looking for an active storage solution.
"What we've seen over the last nine months or so is that Web 2.0 types of organizations have seen tremendous value in CAS," said Derek Gascon, vice president of marketing at Caringo. And the reasons these Web 2.0 companies are interested in content addressed storage, he said, "has to do with the architecture and the capabilities of the CAStor technology, in [that] we're capable of building very low cost, very robust, high performance storage clusters that give organizations the capability to handle active content, which is what CMWare is doing, as well as provide all the feature functionality for compliance and archiving of content for regulatory purposes."
Unlike traditional storage, CAStor uses object-based storage (OSD) instead of blocks. "We store each individual file contiguously, so it's not broken up and spread out ... like you would see in a RAID array," said Gascon. "We're basically creating and virtualizing storage using commodity hardware. That's one of the reasons we're able to scale and deal with unstructured content so easily."
CAStor also automatically replicates files, so if a disk goes down, there's another copy. And Caringo claims that CAStor clusters can recover data faster than RAID.
As for implementation, both Gascon and Ott said it is simple.
"On average, it's going to be less than two hours to implement it," said Gascon. "We can deliver our software on a USB key and you essentially boot the server with the USB key in place and within 60 seconds, the time it takes to boot that server node, you have a CAStor node."
Adding additional storage is also straightforward. You purchase basically by the terabyte, as much or as little as you need. And again, the software plugs right in. CAStor automatically inserts itself (with a little bit of help from a human) into the cluster and is immediately recognized by the system. No provisioning necessary. No application interruptus.