ClearSky: Keeping your premises free of unwanted clouds

Allowing you to forklift your silos the hell out of Dodge

Looking at the ClearSky now

El Reg: Can you say anything about pricing?

Laz: We are an "all-in" per-GB, per-month model, which includes the cost of all gear and network links. We aren’t saying too much about pricing other than that we will be roughly 1/3 the cost of a conventional, fully protected enterprise-class storage deployment.

El Reg: What access protocols are used?

Laz: Our first release is iSCSI, but we expect to have support for FC and NFS in the first set of follow-up releases.

El Reg: What geographies/regions have ClearSky PoPs?

Laz: We are launching the company with PoPs in Boston, Philadelphia and Las Vegas. We expect to be in around ten locations in the US and Europe by the end of 2016. Out model is to market locally at each new location around our colocation partners’ facilities.

El Reg: What is on-premises cache/cloud storage gateway hardware/software configuration?

Laz: We will provide customers with a fully redundant, 2U, enterprise-class storage array chassis. The platform leverages the latest-generation Intel processors and is populated entirely with flash devices.

It provides four 10Gb Ethernet interfaces for the customers’ SAN and it also has additional ports which are used to connect to the ClearSky Global Storage Network through our provided network connections.

The platform runs our own system software, which provides de-duplication, optimisation, and encryption and key management services.

El Reg: How does it link to accessing servers?

Laz: It looks like a SAN device to the customer’s servers (iSCSI over Ethernet) and we also have a full range of VMware integration capabilities, so that they can migrate and use us for their virtualised workloads. We will include full VAAI and VVOL support in the first release.

El Reg: What is round-trip network latency from on-premises to PoP?

Laz: It depends on the physical locations of the facilities in each metro, but it is quite low. In our Boston PoP, for example, our current customers are all experiencing sub-1-millisecond round-trip latencies.

El Reg: What is data access latency at the PoP?

Laz: Average is less than 10ms, but the PoP is usually accessed infrequently.

El Reg: What kind of storage exists at PoP? Some kind of array? What media?

Laz: We have a large, multi-tenant, high-endurance flash write back cache that is large enough to handle all of the customers’ writes. In the background this gets copied back to our warm tier and the public cloud. For our warm cache, we have a unique multi-tenant scale-out object layer that is a hybrid of flash and spinning disks.

El Reg: What WAN optimisation is used?

Laz: We have written all of the data management software that runs in the PoP. All of the data management functionality (clones, snapshots, DR, etc.) is our own. This storage software uses a proprietary protocol that incorporates WAN optimisation in its design. Thus, we are able to maximise the utilisation of our network links.

El Reg: How is data protected?

Laz: The network ensures that there are multiple copies of the data across the cloud and the various cache tiers. If a piece of data is not available locally, the system retrieves a copy from the nearest available cache. This obviates the need for managing RAID.

El Reg: How is data backed up?

Laz: Because the cloud has a full copy, including metadata, for the various points in time, it serves as an integrated back-up. Customers can keep these images indefinitely at a very modest cost.

El Reg: What about disaster recovery?

Laz: We have multiple levels of DR across the network. First, the Edge cache on-premise is synchronously replicated to the PoP. This means customers have continuous RPO times if they were to lose their data centre.

Even more interestingly, customers with more than one data centre in a metro area can failover to another data centre with a simple click. We expect to take advantage of metro-cluster solutions to further automate this in the future.

We will have secondary PoPs in metro locations in case we were to have a failure in one of our buildings.

Finally, because the data is backed up continuously to the cloud, it is possible to extract and run a very recent copy if all else fails. This can be done in the cloud provider’s compute cloud as an option.

El Reg Conclusion: This is another storage silo-melding offering, based on fork-lifting your silos the hell out of Dodge. ClearSky said having multiple, on-premises storage silos is a costly nightmare of managing primary, back-up and disaster recovery storage infrastructure.

Wake up, you can be free of this. Gaze up into the ClearSky and see how your data access and storage could be, so to speak. It's ambitious and worth keep on an eye on to see how it develops.

ClearSky Data will premiere the company and exhibit its service at VMworld 2015 at the Moscone Centre, San Francisco, Aug 30-Sep 4. Visit booth #441 and get a demo. ®