At LSI’s Accelerating Innovation Summit last week in San Jose the CEO’s keynote began with a somewhat futuristic video showing several scenarios that could have been episodes of the current “Person of Interest” television show. People were connected to big data and big compute infrastructures where ever they went in order to keep them safe and make their worlds more secure. This was an interesting depiction and one that’s probably not too far in the future, the NSA’s recent PR problems notwithstanding.
In addition to the opening video, there was still plenty of emphasis on big data, web-scale data centers, etc (see previous blog “Innovation plus Longevity Wins in the Web-scale Market”). The “bigness” of information technology is something that LSI certainly has a lot of impact on. But another area actually caught my interest, one that focuses less on big infrastructure and more on infrastructure that’s “big enough”. Even small to mid-sized organizations are having to handle and process far larger amounts of data than they did a few years ago. But they’re still not in the enterprise market for SANs or large shared storage systems.
Two of the briefings we had we on LSI’s Syncro CS product and an interesting application of their MegaRAID product. In both examples, LSI was giving users in the SMB and mid-market spaces some cost-effective alternatives to handling their (big) data issues with big enough storage solutions – and without traditional fibre channel (FC) SAN infrastructures.
Cluster in a Box
Syncro CS is a pair of PCIe based MegaRAID controllers with failover and shared storage functionality that enables a company to set up an HA clustered application without a SAN or special clustering software. Two rack mounted servers can each connect to a common JBOD array, or a two-node server chassis can be connected to internal JBOD to create a cluster-in-a-box (CiB) solution. Syncro CS provides dual active high availability across server nodes and server storage cluster HA topology support with write-back cache mirroring.
LSI sees this technology as a way for mid-market and even small companies to set up clustered environments for critical applications without expensive and complex SAN implementations. They’re also getting interest from OEMs that want to use this CiB technology to load their software on and create a purpose-built HA appliance. Another use case is providing a dual-server compute or storage node for a scale-out environment, like Storage- or Infrastructure-as-a-Service clouds.
Continuing with this SAN-replacement theme, LSI had a demonstration running with 20 servers and 10 JBOD arrays all connected to a 40 port, 12 Gb SAS switch. Using the new 12Gb SAS MegaRAID controller in each server, they created a ‘de-clustered RAID’, spreading data across all 160 drives in these 10 JBOD arrays.
To be clear, this was a ‘concept car’ type of demo, not a configuration that’s currently available. But LSI did make the point that you can get SAN-like capabilities out of a shared DAS (sDAS) environment. With each server sharing the same physical spindles they could move volumes between servers, migrating data when a failure occurs, or moving VMs, etc.
A similar example of this shared DAS concept is with LSI partner Dot Hill. Their AssuredSAN 4004 allows four servers to redundantly direct connect to the same array using 12Gb SAS. This way a company can support multiple virtual server hosts or high performance application servers without running a SAN. It also makes and ideal solution for HPC use cases.
Storage Swiss Take
Sharing DAS and using the less refined but abundant performance of SAS to do what used to require an FC SAN is a powerful theme. The big data centers have stopped using ‘big iron’ storage infrastructure and instead found a way to leverage commodity server hardware and cheap drives. I think this is an example of that same kind of thinking. Why assume the cost and complexity that we had to yesterday when there’s enough capability in much simpler solutions today. Sure, there will be cases where only a SAN will do, but for every other time why not use some of these SAS technologies and spend the money you save on something else?