IBM strengthens disk platforms

Focuses on partitioning, virtualization

By Heidi Biggar

IBM recently countered disk array announcements from Hitachi Data Systems (and resellers Hewlett-Packard and Sun) with two introductions of its own-the TotalStorage DS6000 and DS8000 families.

Both arrays are part of IBM's new TotalStorage Disk System (DS) family, which also includes the entry-level DS300 and midrange DS400 (previously sold under the FAStT model name). The new systems are expected to become generally available next month. IBM introduced the TotalStorage DS brand in September.

IBM officials say that while there is some overlap between the existing DS400 platform and the new DS6000 family, contrary to rumors the company has no plans to do away with the DS400.

The DS8000, however, is intended as a replacement for the Enterprise Storage Server (ESS), or "Shark," product line.

While the new DS6000 and DS8000 disk systems boast impressive performance and capacity improvements over previous IBM disk systems, analysts say that the real news is not "speeds and feeds" but features and functionality.

"The big story is the future potential of IBM's logical partitioning [LPAR] capability [for some DS models] and the shared storage microcode between the two DS platforms," says Dianne McAdam, senior analyst and partner at the Data Mobility Group consulting firm.

Logical partitions are designed to give administrators greater flexibility in deploying the capacity they have purchased for these high-end systems and to make better use of overall disk resources (e.g., capacity, cache, and processing power) by allowing administrators to carve a single box into multiple logical (or separate) storage systems that can then be assigned to various applications according to quality of service (QoS) objectives.

What this means is that users can run separate workloads from the partitions. For example, users could designate one partition for test data and a second partition for production data, explains Leslie Swanson, vice president of IBM Storage Systems. Alternatively, users could take a snapshot of their production data and then move it over to the second partition, where it could be later used for data mining purposes, says Swanson.

LPAR is currently available as an option for the DS8300 model. It is not available with the DS8100 or DS6000 platforms. Initially, users will be able to create two partitions; however, IBM expects to expand this to as many as eight partitions over the next year.

Additionally, IBM says users will eventually be able to run various applications, such as IBM TotalStorage SAN Volume Controller (SVC), Tivoli Storage Manager, or Virtual Tape Server (VTS), from these partitions as an alternative to running applications on intelligent switches or, in the case of EMC, on a router in the fabric.

"We're kicking around many ideas," says IBM's Swanson. "We'll let some of it play out in the marketplace and we'll decide how we want to exploit [the technology] as we see how our customers want to use it."

About IBM's LPAR capability, Randy Kerns, senior partner at The Evaluator Group consulting firm, says: "It creates some new opportunities for IBM. It gives them a great consolidation plan where they can control the resources for specific applications and present multiple storage system images."

As for the potential of running other applications from the LPAR, Kerns says: "It is very intriguing and will be a new competitive wrinkle."

If all this talk of virtualization and partitions sounds vaguely familiar, it is. The buzz surrounding Hitachi's TagmaStore introduction in September also centered on virtualization and partitioning.

Hitachi and IBM have a similar story, according to John Webster, senior analyst and partner at the Data Mobility Group analyst firm. They both believe "intelligence" (or the virtualization engine) belongs in the disk array, not in the fabric, and they both believe partitioning is necessary to support this capability (for scalability and resource allocation reasons), says Webster.

However, while their strategies are similar, their implementations-in particular, how they virtualize, or plan to virtualize, back-end storage-are different.

In a nutshell, Hitachi virtualizes the storage environment through a front-end port on the TagmaStore USP, not through a USP partition. (The USP is not capable of running other software applications from the partitions like the DS8300 can.) IBM, in turn, will do it by running SVC off a controller in one of the system's LPARs, according to Kerns.

It is important to note that the SVC can currently be implemented as an in-band appliance or embedded on a blade (e.g., Cisco's Caching Services Modules) in an intelligent switch (e.g., Cisco's MDS 9000).

Both vendors' implementations allow users to aggregate heterogeneous disk capacity in their SAN environments and then manage them as "pooled" capacity from a single point of control. (For a detailed look at IBM's SVC and TotalStorage SAN File System, see IBM takes 'virtual' steps with SFS )

As for the shared microcode between the two new IBM disk platforms (the company claims that more than 97% of the code is the same), it is significant for several reasons, most notably that midrange users will now not only have access to many of the same features/functions of IBM's high-end models (e.g., direct copy, replication, etc.), but they will also be able to manage and move data on these two platforms.

Additionally, the DS8000 shares microcode with the ESS750/800 platforms, which makes it possible to replicate and copy data between these two platforms, but not the DS6000.

This article was originally published on October 18, 2004