Finding the right storage fit
At a time when business metrics such as revenue, profit and customer-win rates are heading south, there is another corporate indicator that is stubbornly heading in the other direction – data storage volumes.
Data storage growth has been one feature of organisational life that has remained utterly predictable in the face of a turbulent and uncertain trading environment.
“People have become a lot more dependent on storage within their infrastructure and the demand for storage and appetite for storage has remained huge,” said Kevin Swan, head of enterprise solutions with Dell Ireland. “Data growth can be up to 100 per cent year-on-year, depending on the environment.”
The storage bubble within enterprises is being sustained by a number of factors – intensive back-up practices that can see the same piece of data being replicated and stored 25 or 30 times; the availability of low-cost disk-based storage, which is making it cheaper than ever to store data; and the legal obligations imposed on publicly listed firms to retain emails, deleted and undeleted, for several years for ‘e -discovery’ purposes.
But perhaps the biggest driver of all is the evolving nature of the data itself.
“There’s an increase in the amount of data that’s been stored and there’s a change in the character of that data,” said Karl Jordan, country manager, enterprise storage at HP Ireland.
“Much more of it now is unstructured data – it’s the documents, it’s the image files, it’s the sound files – and it’s taking up most of the capacity of what we sell now.”
The changing storage requirements have been reflected in the type of storage solutions seen within organisations. Time was, enterprises stored a lot of data offline, in tape libraries for example, but the tumbling cost of disk based storage combined with increasing disk capacities had caused a huge jump in the amount of data stored online.
“The biggest disk you could buy in the mid-1990s was a four-gigabyte disk,” said Declan van Esbeck, managing director of storage services firm Redstone Technology.” Now we’re looking at terabyte drives and terabyte-and-a half is the next one. It’s just colossal.”
Similarly, interest in disaster recovery (DR) solutions has shot up, fuel led by falling hardware prices.
Swan estimates that whereas five years ago only 10 to 15 per cent of storage projects with customers involved a DR component – customers buying two arrays rather than one, which would be replicated against each other. Today the figure is closer to 75 per cent.
“Our customers are buying a storage solution for today but they’re already thinking about a secondary site – be it their own site or a hosting site – to replicate that information too, and that’s very high on the agenda,” said Swan.
But not only are enterprises using more hardware, they are combining technologies in new ways in order to achieve the best bang for their buck.
“What we’re seeing is horses-for-courses in the storage world,” said van Esbeck.” We’re seeing people who are buying completely different sub-systems for archival data to the subsystems they’re buying for their high-performance requirements. We never saw that before.
“You can go into an enterprise and find they have high-performance Sanbased fibre channel for 10 per cent or 20 per cent of their storage. They have an iSCSI middle tier that’s based on ATA drives. Then they have a third tier that’s based on these massive, scale-out storage farms – petabyte-in-a-box type stuff.”
The trend is partly being driven by budget pressure: organisations have less money to spend and are looking to apply storage technologies that do the job for the lowest outlay.
This can clearly be seen in the connectivity technologies being chosen to move data between servers and storage. Fibre channel is still the default connectivity technology for large enterprises – 80 per cent of the market install base is fibre channel. But iSCSI SANs – which use the ethernet to transfer data between servers and a San – are starting to become a popular alternative.
The drawback of iSCSI is that it is significantly slower than fibre channel. At the same time, it is also considerably cheaper, making it an attractive alternative in non-high performance installations.
iSCSI has been around for the last five or six years but has been slow to build a user base – so much so that some industry wags had rechristened it ‘I-Still Can’t-Sell-It’. But people are buying it now, and the reason is cost.
“For many mid-sized organisations, the performance is fine; it’s fit-for purpose,” said Jordan. “iSCSI-based arrays represent the fastest growing storage market sector at the moment, and we will see this continuing for some time.
“The attraction of simpler connectivity to storage resources, lower cost than fibre channel, and the fact that iSCSI delivers adequate performance for most mid-sized environments means that it will continue to increase in market share.”
One of the vendors pushing iSCSI hard is Dell, which claims number one spot globally in the technology. Swan said its key selling point was that it could match up to fibre channel in terms of capacity while being much cheaper to deploy.
“In terms of cost, it is completely different economics,” he said.” Fibre channel requires a lot of infrastructure, a lot of plumbing.You don’t need that with an iSCSI solution because you’re leveraging off the network you already have in place.
“Moreover, over its life time, iSCSI is much cheaper to manage because you don’t need the same level of expert skills – traditional networking skills suffice – plus the tools available on the newer arrays give a better operational efficiency. So, in the right place, iSCSI provides a much better TCO [total cost of ownership].”
Another, newer connectivity technology being pushed hard by the likes of EMC and Cisco is fibre channel over ethernet (FCoE). FCoE switches allow both Lan and storage traffic to travel on a single 10 gigabit ethernet link, eliminating the need for separate San and Lan adapters and cables. Reducing the number of adapters and simplifying cabling helps lower equipment acquisition costs as well as the operational costs associated with powering and cooling servers.
Currently, the technology is only at early adopter stage and global standards have yet to be agreed, but Jordan is in little doubt that it’s the connectivity technology of the future.
“Be sure, once the standards have been ratified it will become a mainstream switching technology. Because what it allows you to do is, instead of having two networks in your environment – one for IP traffic and a second, usually fibre channel, to drive your highspeed San – you can converge that into a single fabric.
“It will allow fibre channel to be a’loss-less’ protocol over ethernet – ie, no packets of data wil l be lost in transit – which is not currently the case.” Concurring with Jordan’s positive assessment of FCoE is Redstone’s van Esbeck,who felt that the technology “had the potential to kill iSCSI’‘.
Software options
Advances in storage hardware have given enterprises lots of options with regard to deploying storage systems, but it has also increased the complexity of their environments.
One consequence of this is that more and more software is required to make sense of all this complexity; that can manage the storage across an entire system consisting of heterogeneous components. In recent years, as a result, there has been a dramatic increase in storage software spend.
“In 2005, we were selling 85 per cent hardware and 15 per cent storage for a typical storage enterprise solution,” said van Esbeck.
“Now it’s 50/50 – and that 50 for software will become far higher.”
He also predicted that, as well as final ly witnessing “the end of tape’‘, the next few years would see the average enterprise storage infrastructure expand to at least five or six tiers.
This being the case, having good software that can manage this complex storage environment wil l be a necessity.”The key won’t be hardware; it’ll be software, particular software management tools,” he said. Investment in storage software is currently being funneled into a range of technologies, including:
* Enterprise storage management
“Enterprise storage management software (eg HP Storage Essentials) is now very advanced, and can provide full, end-to-end discovery, alerting, management and reporting across local and distributed heterogeneous SANs,” said Jordan.”There is no reason why a single administrator cannot have full visibility of the status of the ITenvironment from application to disk, from a single pane of glass.”
* Virtualisation
Virtualisation technologies such as EMC’s VMware have automated the management of virtual servers, meaning that server capacity is automatically allocated to an application as and where needed. By extension, the storage capacity is also automated and becomes virtual.
“The storage is made available to the application by the configuration of the servers and we don’t have to know where the storage is,” said Conor Sexton, associate director of IT architecture at Fujitsu Consulting.
Sexton pointed out that virtualisation was also a key technology underpinning the high-availability, fault-tolerant computer systems demanded by many modern businesses.
“Applications have to be highly available or 24/7, even in businesses where you wouldn’t normally have expected it. It’s just the way business has gone.”
He said that VMware played a particularly important enabling role in high-availability computing, particularly its latest version, V-Sphere 4.
“This promotes fault tolerance rather than failover,” he said.” This means there is no loss of service if one server goes down. It used to be the case that if server A went down there would be a 90-second delay before server B kicked in. Today, there is no loss of service, it is completely automated.”
A notable development within the virtualisation arena is a new technology known as ‘thin provisioning’, which can dynamically distribute the storage capacity as and when needed.
“Thin provisioning is one of the most popular of the new storage virtualisation technologies and one that has the potential to significantly reduce the costs of acquisition, maintenance and management of data storage,” said Jordan.
“Thin provisioning can be described as the ability to automatically present sufficient storage capacity to applications to meet the immediate needs of those applications, and to automatically increase the capacity presented at any time with no disruption to the applications. Additional physical disks can be installed as required on the fly, transparent to the applications. It is the enabler of true storage-as-a-utility service.”
Virtualisation software is also allowing smaller enterprises to experience the benefits of San technology. They can do this by deploying a virtual or ‘soft’ San. This is achieved by implementing a software layer that pools the capacities of several storage units directly attached to servers. According to Jordan, many growing organisations are deploying virtual Sans as a stepping stone to putting in place a full San infrastructure which involves physical San switches.
* Information lifecycle management (ILM) Information lifecycle management (ILM) tools manage the life of a piece of data through the various storage layers, from tier 1 ultra-high-performance, high-accessibility storage down to slower, fatter tier 3 and 4 storage.
“This is still quite a manual process because there aren’t many ILM tools available but we’re starting to see HP and EMC in particular coming out with tools that can manage the lifecycle of storage from tier 1 down to tier 4,’’said van Esbeck.
According to Sexton, organisations have become more aware of ILM and the need to classify data according to its importance so that lower priority data does not continue to sit on expensive high-performance disks.
But while technology can help manage this data lifecycle, he believes that corporate policy also has a role to play.
“Basic data management starts with deleting the dead wood,” said Sexton, who observed that organisations that tackled their growing storage mountain by simply adding new disks were” just kicking the problem down the road’‘ when it would only become a bigger problem to manage.
* De-duplication
In recent years, technologies such as data de-duplication have emerged to help organisations eat into their data storage mountain.” Data de-duplication works by looking at commonality between files,” said Swan.” If there are 250 of exactly the same 10mb file on the network, it stores the file once rather than 250 ten megabyte files and if there are changes to those 250 files, it only updates the changes.
What you’re really seeing is that for 6TB of data – which is probably the average in Ireland for an enterprise – you’ll need another 6TB of storage to back that up. But with techniques such as data de-duplication this 6TB can be reduced down to about 300GB – a huge level of compression.”
Over the coming years, as the storage volumes continue to expand, the industry will continue to introduce new technologies – hardware and software – that can help enterprises manage their storage inventory more flexibly, more efficiently and, perhaps, more cheaply than ever before.
“The bottom line is that storage is not going away,” said Swan.” If anything, year-on-year growth is going to be double.”