1 Comment

Summary:

Dave Wright, who founded the Amazon S3-based Jungledisk storage service, is back in the entrepreneurial saddle with today’s launch of the cloud-focused SolidFire. It’s building a scale-out block storage offering, which will try to address the growing and vexing problem of storing virtual machine images.

Dave Wright, who founded the Amazon S3-based Jungledisk storage service, is back in the entrepreneurial saddle with today’s launch of the cloud-focused SolidFire at our annual Structure conference. SolidFire is building a scale-out block storage offering, which amongst other things will try to address the growing and vexing problem of storing virtual machine images.

Wright has a solid background, having been the first engineer at GameSpy, an online gaming company sold to IGN and then merged into News Corp’s Fox Interactive Media. He later founded JungleDisk, which he sold to RackSpace in late 2008. Wright spend a year with RackScpace before moving on to start SolidFire, which currently has a handful of employees and is expecting its beta product to be available by the end of this year.

Wright sees service providers struggling to keep up with storage for virtual machines. Traditional vendors still want to sell disk-based solutions and current block-based solutions don’t scale to support the onslaught of virtual machine workloads. Wright says that SolidFire’s focus on solid-state media enables a high degree of parallelism within the system to keep up with service provider workloads across hundreds to thousands of physical machines.

While not a one-to-one relationship, cloud computing has benefited directly and indirectly from the virtualization trend that kicked off long before the cloud consumed us. On the cloud provider side, it allowed for the drastic overprovisioning of virtual machines that could be managed on a cost-effective infrastructure, and delivered at a low price per hour. On the end user side, it prepared customers to package entire servers as software images and move them around internally, and ultimately externally to public cloud providers.

But virtualization has created storage and I/O workloads that are different from what most companies and service providers have dealt with in the past. Today almost every storage company touts an ability to integrate more tightly with virtualization solutions, but according to SolidFire, a new approach is required to tackle the problem effectively.

SolidFire’s product is a software solution that runs on physical servers to deliver a highly scalable block storage. This contrasts with traditional block storage that often has limitations in the total capacity or the total number of volumes within a single system. The company also focuses exclusively on solid-state drives, which provide enormous performance advantages, and more importantly, low latency response times across highly varied workloads. The deterministic behavior of solid-state media compared to rotating disks allows allows storage providers to offer a quality-of-service guarantee — critical when dealing with multiple customers on a shared cloud infrastructure.

But solid-state media gets expensive, so SolidFire has kept en eye on efficiency of that storage as well, and is initially targeting workloads such as virtual machines that have a favorable compression ratio. Since most virtual machines share a common operating system, if handled appropriately, virtual machine images can be deduplicated down to a far more efficient storage footprint — which helps keep the cost of a solid-state solution more in line with that of a disk-based solution.

In more ways then one, virtualization is the new file system. If SolidFire can find a way for service providers to cut the cost of their virtual machine cloud offerings and get a deterministic quality of service, they could become a force in cloud computing infrastructure buildout.

Gary Orenstein is host of The Cloud Computing Show

  1. [...] at the Structure conference reception. He explained to me what he is up to. Solidfire has a software solution that uses solid state drives (SSD) as scale-out block I/O. I see huge potential in how this can be used for BI applications. When you put all the pieces [...]

    Share

Comments have been disabled for this post