Stratus Computer Inc, an up-scale Unix clustering vendor that’s been waiting for NT to grow up and age before supporting it, is pushing the maturing process along by signing an OEM agreement with Microsoft Crop to sell a massive NT server cluster it calls RADIO fitted out with BackOffice as well, according to our sister publications ClieNT Server News. While Stratus’s Continuum business is still a Unix presence, the RADIO super server will be an NT-only affair. Although it had once planned to support UnixWare on it too, it sold only a few systems to the customers it’d promised it to, then abandoned the operating system and recently retired the UnixWare part number. As the rest of the NT world plays around with mere dual-node offerings, Stratus will bring multi-node clustering to market this quarter, making it the first vendor to cluster more than two NT servers. While RADIO is not as gargantuan as the things Stratus is used to offering the enterprise world, by NT’s standards it’s huge: at release it will ship with up to six clustered nodes, interconnected in one chassis. RADIO marketing and management chief Jane Shurtleff said that Stratus’s working goal is to cluster four fully configured RADIO chassis together, meaning 24 clustered nodes. The product will start at about $66,000, upstream of most NT servers. Short for Reliable Architecture for Distributed I/O, RADIO will offer selectable availability, three options based on user needs ranging from fault tolerance to high and continuous availability. The top-of-the-line continuous availability configuration, which replicates mission-critical applications and data and transparently directs clients to another node if a server node fails, is the first such offering on NT. RADIO Continuous Cluster uses Stratus’s Isis Reliable software to synchronize replicated data across multiple systems. Continuous Cluster will automatically support SQL Server, and Isis software will pave the way for other major databases. Oracle Parallel Server (OPS) is conceptually similar, but it won’t be ready for market when the Stratus product hits and can only handle two nodes anyway but Shurtleff said that RADIO’s flexible structure can run OPS on top. Most NT customers will primarily use the high-availability option, she said, which lets applications failover and restart to another node. While the mid-range High Availability Cluster will let users custom script support for any NT application through Isis Availability Manager, Stratus will write scripts for selected BackOffice apps and bundle them with RADIO in one of three different packages that include varying configurations of NT, IIS, Exchange and SQL Server. The packages will follow a month after RADIO. The third entry-level fault-tolerant configuration replicates data but doesn’t automatically failover or provide continuous availability. RADIO’s structure supports nodes with different degrees of availability so one pair could be highly available and another could be continuously available. Each RADIO cluster contains eight bays, two of which must be occupied by Network nodes while the other six can have a mix of Storage and Compute nodes. Compute contains dual 133MHz Pentiums while Storage has a single 100MHz Pentium. Shurtleff said Stratus will upgrade the Compute chips to Pentium Pros, possibly at the beginning of next year. A latecomer to the Wolfpack clustering discussions – it started attending design reviews in March shortly after it signed the OEM agreement – Stratus says it will support the Redmond API when it emerges and is working with Microsoft now to ensure its offerings will be compatible. Shurtleff said it’s too early to tell what if any of its technology will make its way into the API but figures that because of the group development process, there’s gonna be a little bit of all of us in there. She did say that Stratus will port the ability to reassign IP addresses in a cluster to the Wolfpack architecture.
