SATA disks

Enterprise-class storage systems have always been built using enterprise-class disk drives – components that have been engineered to provide the kind of ultra-high reliability required for highly active data.

Such primary storage does not come cheap, of course, which has prevented its use outside of high-performance and mission-critical applications. Certainly there is little justification for using such devices for moderately active data, back-up or non-critical applications.

However, in recent years, an alternative has emerged: the use of SATA (serial advanced technology architecture) disks – the same type of disks that have been fitted in PCs, laptops and other low-end servers since 2001. Not only are these much cheaper, but in many cases they can be fitted within the same storage sub-systems as enterprise-class Fibre Channel or SCSI disks.

Moreover, the growing sophistication of failover software means that organisations are increasingly using SATA-based systems for more than just secondary storage. “Low-cost or economy disks using SATA are moving from the desktop and into the data centre creating new and appearing opportunities," says Fred Moore, president of storage consultancy Horison Information Strategies. As such, SATA is the fastest growing segment of the disk market.

SATA still trails FC and SCSI disk technologies in many ways: mean-time between failure on a SATA disk runs at around half a million hours based on usage of 2 to 4 hours a day; while SCSI and FC disks average 1.2 million hours based on 24/7 running. SATA is also implemented as a point-to-point architecture, requiring cabling for each device, while SCSI uses a bus architecture and FC a loop architecture.

But SATA still has some growing to do. “With a roadmap that calls for 600MB/s, SATA-based disk drives may become a somewhat disruptive force in the disk industry,” says Moore.

Storage resource management

Earlier this year, when Joe Tucci, the CEO of storage market leader EMC, identified what he labelled ‘the three game-changing technologies’ of storage, there was one unexpected item on his list. Along with virtualisation and information lifecycle management was the badly understood and arguably neglected area of resource management.

Although it has been an integral part of storage environments since the dawn of the sector, storage resource management (SRM) has been relatively unsophisticated in its scope. Existing systems often consist of system vendor-provided utilities that offer little more than device-level monitoring of faults, loads, performance and related storage attributes.

However, the complexity and critical position of modern, networked storage environments demand a much more ambitious set of management capabilities that includes the ability to discover, map, report and act upon the changing status of the storage estate, and relate that all the way up to the applications layer it serves. Nowhere is that more evident than when things go wrong. According to Chris Gahagan, senior VP of resource management at EMC, storage managers find it very difficult to understand the business impact of a single fault from the storm of alerts that the incident triggers. Traditionally, such events have kicked off an intensive, largely manual effort right across the infrastructure to diagnose the root cause.

What they really need, he says, is a tool that can “fully automate incident management and triage”, that correlates the cause and presents an action plan. Addressing that requires some fresh thinking, he argues: next-level SRM will be characterised by the move from framework-based to model-based management.

Most SRM  tools today collect a group of events and try to make sense of them; model-based management turns that on its head, by modelling the infrastructure, discovering all the devices and understanding their connections and behaviours. Such systems generate all possible signatures for all possible failures across all storage domains and correlate appropriate responses.

Such growing sophistication, according to analyst group Gartner, will help to generate compound annual growth in SRM software revenues of around 15% between now and 2009, a pace that will take the market to $1.2 billion.

The beneficiaries of that demand, says Gartner, will be vendors with broad SRM suites that span discovery, capacity monitoring, performance, chargeback and quota management, configuration and change management, root-cause analysis, reporting and event management.

Today, vendors can tick some of the components on that list, but that leaves plenty of potential for the emergence of a ‘gorilla’ of this next wave of SRM.

4 gigabit SANS

When storage area networking first sprung on the scene in the late 1990s, its Fibre Channel (FC) backbone could only pump data at around 1 gigabit (Gb) per second. That was the bare minimum to make SANs practical, but it was still a bit too slow to encourage largescale adoption.

That wider take-up began with the arrival of 2Gb FC storage systems, and their switches and host-bus adapters (HBAs) in 2001. But, in recent years for some organisations, that 2Gb has become “a choke point,” says Ian Bond, a data centre consultant at networking company Cisco Systems.

“Two gigs was holding people back,” echoes Paul Phillips, UK sales director at FC switch maker Brocade. He cites the example of a bank with two linked data centres, where the requirement is for a superfast data restore from tape in the event of a failure at one site.

But since the early part of this year, that barrier has been lifted. The launch of 4Gb switches and HBAs, followed by storage arrays that could support those new accelerated data paths has enabled a move towards end-to-end 4Gb storage environments.

The rate of the switchover is being encouraged by the fact that 4Gb devices are compatible with the previous generation, enabling users to plug new 4Gb purchases into existing 2Gb installed products. Typically in mixed environments, systems can automatically detect the ‘link rate’ that each target can support and transmit at maximum speed.

The benefits are clear. Most obviously, a 4Gb device attached to 4Gb switches will run up to four times as fast as its un-upgraded counterparts. But that is not the top speed: eight switches can be combining to deliver up to 32Gb per second on a single ‘trunk group’.

At the same time, 4Gb FC technology can support a much higher ‘fan-in’ ratio of hosts to storage ports and, perhaps surprisingly, runs at considerably lower cooling and power levels than previous systems.

And, true to the crazy economics of the storage industry, that hike in performance and functionality does not come at any extra cost: the price of a 4Gb unit is about the same as its 2Gb predecessor.

But that 4Gb is just a taste of things to come. The specification for 8Gb has already been drawn up and vendors are already planning products for late 2007 or early 2008.

The reason for the aggressive roadmap is that FC is under attack. A new wave of products supporting the use of standard Internet protocol SANs and Ethernet is appearing with promises of turbo speeds. As Praveen Asthana, director of storage and networking at Dell, recently suggested at Information Age’s Future of the Data Centre conference: “10 gigabit on copper is right around the corner”.

The fact that IP SANs are based on a completely different architecture from today’s FC SANs means the transition will not be overnight.

However that battle turns out, the transition to end-to-end 4Gb FC storage will take the brakes off many SAN implementations. “The transition from 1Gb to 2Gb took about 4 years; the switch between 2Gb and 4Gb will be much more rapid,” observes Cisco’s Bond.

Further reading

• More articles from Information Age's Storage Business Briefing.
• For more stories like this, check out the Business Intelligence briefing room.