- The hybrid model
- The pure model
- Disaster recovery planning and pricing
- Cloud disaster recovery providers
- Redmayne Bentley virtualises in the cloud
- Voice over IP
- Ground Control gets Plan B cloud DR
Before server virtualisation, the standard approach to implementing disaster recovery (DR) was to build a second site that duplicated the primary datacentre’s hardware and software. Of course, duplicating the infrastructure meant doubling costs, so only larger organisations opted for this route.
Now, virtualisation has made servers and apps independent of hardware so there is no need for the duplicate iron. You can even run apps from someone else’s servers, so the need to own any secondary hardware disappears. This is cloud disaster recovery, and it is ideal for small and medium-sized enterprises (SMEs).
It comes in two key varieties –hybrid and pure.
Many cloud DR services use a hybrid model in which an on-site appliance receives your data and stages it to cloud.
Should disaster strike you can work from servers and data held on the appliance or work from the cloud, depending on the severity of the outage.
The key advantage of the hybrid model is that it overcomes a key limitation of working in the cloud, namely lack of bandwidth.
The pure cloud DR model sees data transferred immediately to the cloud without the intermediate staging appliance.
Restoration in this model can be by remote working from the provider’s cloud or by receipt of a data disk from which servers are rebuilt.
Pure cloud models will incur smaller upfront and ongoing costs but are only really suited to smaller organisations.
But, there are key questions to ask your DR provider. You will need to be sure the service provider will transfer data securely in the event of a disaster, authenticate users properly, and ensure compliance with regulatory requirements.
You also need to ensure you have enough bandwidth and network capacity to redirect users to the cloud.
A DR plan should also detail how you’ll restore data, including expected recovery times. And your due diligence should include research into reference sites and case studies of real-world disasters and the provider's response.
Among key cloud DR service providers are:
• Amazon Web Services underpins the cloud DR services of a number of providers, and offers its systems as the basis for a DIY approach to cloud DR.
• Phoenix offers geographically-dispersed workplace as well as IT DR services. The latter includes a ship-to-site service for when the building is intact but equipment has been lost through damage or theft.
• Plan B recovers individual servers every night, just in case they are needed, tests recovered systems overnight, and promises to repay a year's fees if it misses its guaranteed recovery time.
• Rackspace, which offers host-based replication services and support for physical and virtualised environments at the low-cost end of its price range. At the high end, datacentre-to-datacentre replication for mission-critical data on storage arrays is available.
• Savvis offers a cloud DR service that includes cold, warm and/or hot site provision.
Ed Sibley, head of IT at Leeds-based stockbroker Redmayne Bentley, was looking for a partner to help modernise the company's infrastructure, and turned to cloud and DR provider Phoenix for help. Disaster recovery was included in the bundle of cloud-delivered services it provided.
Redmayne Bentley owns and franchises branch offices across the UK, with sites in Leeds, London and Farnborough, and originally had servers on site in London, with traditional tape backup.
Its infrastructure was in need of an upgrade and the plan was to consolidate into a new head office in Leeds, but this would introduce a potential single point of failure.
“A year away from that move, we could see a weakness in our DR provision,” said Sibley. “A disaster affecting the computer room in Leeds would cut off all our 38 branches. We didn't want to build a new computer room, so we set a strategy to move all services into a hosted environment with an IT services partner."
This meant using the cloud for services such as voice over IP (VoIP), virtual desktops and remote access.
"For us, the cloud partner had to look after the whole hosted environment," Sibley said. And when Phoenix merged with ICM and Servo, with whom Redmayne Bentley already had relationships, it made sense to stay with Phoenix and use its hosted cloud facilities.
"All our systems are hosted with Phoenix and all are virtualised," said Sibley. "The benefit is flexibility, so we can increase CPU and memory and servers easily."
Disaster recovery is built into the offering, as Phoenix uses NetApps SnapMirror array-to-array replication. This takes hourly snapshots from the firm’s Wakefield datacentre, close to Redmayne Bentley's Leeds HQ, and pushes it out to the hosting company's datacentre in Farnborough.
"We are also considering using the Farnborough datacentre for warm standby," Sibley said.
Time-saving a primary benefit
The benefits of using DR in the cloud are primarily the time it saves.
"Using traditional recovery methods, if there was an outage in Leeds and we had to fall back to a server in London we would have had to recover tapes to that server," said Sibley.
"This could take half a day or so. We then had to make sure it worked and reconfigure the network infrastructure to make sure it was accessible, all before giving users access. It could easily take a day."
With the new system, it takes 15 minutes to set up the new workplace, whereupon users can work in one of Phoenix's 18 work area recovery facilities, work remotely, or work in the office using a virtual desktop.
"The great thing about this system is that removes the grunt work of doing a DR test," said Sibley.
"Everything continues as normal, the whole system is giving continual service and it's a more flexible basis for continuing basic operations."
Essex-based Ground Control is a UK-wide commercial landscape design and construction company that 18 months ago outsourced DR to provider Plan B.
Previously, the company backed up its four servers to HP LTO-4 tape but had no server recovery technology.
"When the Microsoft Exchange server went down, which it did often, we had to retrieve our email records and SQL Server data from tape. Rebuilding the server out of the box could take half a day or more," said head of IT, Sim Hassall.
When that server crashed and was unavailable for two days, Hassall and his team knew they needed a better solution, especially as the company was experiencing high levels of growth in volumes of business and numbers of employees.
"We decided to replace the servers with new hardware and all the latest software upgrades, applications and databases," Hassall said. "We also decided to see how long it would take to fix them if they went down, and at that point we realised we had no off-site backup, so we looked at cloud DR.
"At the time, we ran the company on four ADSL lines so we didn't want to upload large amounts of data to the cloud during working hours. Our research for a solution found Plan B, who provided a hardware device embedded in our datacentre that backs up the servers and uploads snapshot images of our servers overnight, taking four hours."
Faith in the system based on experience
Hassall said his trust in the system is the result of experience.
"We have had to invoke the DR service. A server went down, so we phoned Plan B and said we needed to rebuild it," Hassall said. "It took less than 20 minutes to connect to an image of that server hosted at their datacentre. We found the experience not too bad over ADSL; it allowed us to remote desktop into it until we were able to rebuild the in-house server."
“We test the DR service once a month and Plan B provides a daily report to let us know of any problems," Hassall said. "They take an image every night and can tell us of any application failures, which is particularly important for critical services such as DNS, domain controllers and so on. They also look to see if there are any replication issues.”
"This means it’s not just assurance that our data is safe if there's a fire, but there's also a proactive monitoring of our servers."
This was first published in December 2012