|By Jerry Melnick||
|April 21, 2014 11:00 AM EDT||
Enterprises are moving more and more applications to the cloud. Gartner predicts that the bulk of new IT spending by 2016 will be for cloud computing platforms and applications and that nearly half of large enterprises will have cloud deployments by the end of 2017.1
The far-reaching impact of cloud computing is summarized in a recent McKinsey report on disruptive technologies: "Cloud technology has the potential to improve productivity across $3 trillion in global enterprise IT spending, as well as enabling the creation of new online products and services for billions of consumers and millions of businesses alike."2
For many organizations, moving applications that can tolerate brief periods of downtime to the cloud is a straightforward decision with clear benefits. However, concerns about how to provide high availability and disaster protection in the cloud may make this decision more difficult for business-critical applications such as SQL, SAP, and Exchange. Understanding the facts about HA and DR in the cloud can help you make informed decisions about moving applications to the cloud, while ensuring the important business operations that depend on them are protected from downtime and data loss.
Fact #1: You need high availability protection in a cloud.
Do not assume that your cloud environment provides high availability protection, unless you have specifically configured it for HA. In fact, according to a recent study: "The average unavailability of cloud services is 10 hours per year or more, while the average availability is estimated to be 99.9% far less than the expected availability of business critical applications."3 That is the equivalent of more than a day of downtime. In fact, in 2014, Microsoft Windows Azure, Google, and Amazon Web Services all had some measure of service interruptions or downtime ranging from 4 minutes to several hours.4
For business critical applications, the redundancy that you can get with some cloud solutions, such as Windows Azure, is not enough. When you consider the cost of a minute of downtime for applications, such as SQL Server, Oracle, and SAP that may run many of your key business processes, it becomes clear that you need true high availability and disaster recovery protection. You need to ensure that end users have immediate access to data and applications in the event of a local failure, a regional disaster or anything in between.
However, the traditional way of providing high availability protection is to build a cluster using two identical servers - a primary server and a standby server - with shared (typically SAN) storage. If the primary server fails, the application operation is moved to the standby server, which has immediate access to the same storage. The problem is that SANs are not only expensive to buy, manage, and maintain, they are simply not an option in public cloud offerings. There are, however, high availability solutions that can be used in a cloud that do not require a SAN.
Fact #2: You can build a cluster in a cloud.
Even though you cannot have a SAN in a cloud, you can build a cluster for high availability protection. In a Windows cloud, you simply add SANLess cluster software to your Windows Server Failover Cluster (WSFC). The SANLess software uses real time, block level replication to keep local storage in two geographic regions of the cloud synchronized. If there is an outage, the application operation is automatically moved to the remote instance, which has immediate access to current data. The synchronized storage looks to the WSFC like a traditional shared storage so there is no added complexity or specialized skills needed to build or manage a SANLess cluster. In fact, a SANLess cluster is easy to manage and has the added benefit of eliminating the single point of failure risk of a SAN. SANLess clusters also provide complete configuration flexibility, allowing you to replicate between physical, virtual, cloud, and hybrid cloud environment as well as between SAN and SANLess clusters.
Fact #3: You can have geographically separated nodes for DR in a cloud.
While providing high availability within the cloud will protect you from normal hardware failures and other unexpected outages within an availability zone (Amazon) or fault domain (Azure), you still need to protect against regional disasters. The easiest solution is to configure a multisite (geographically separated) cluster.
One effective method is to build a SANLess cluster within a cloud and extend it for disaster recovery by adding another node(s) in an alternate data center or a different geographic region within the cloud. Unlike traditional clusters that require you to have identical hardware and software in every node, a SANLess cluster allows you to mix physical, cloud and hybrid cloud configurations. The benefits of a DR configuration are clear. For example, simply adding a third, geographically separated node to your SANLess cluster in a Windows Azure cloud can give you a recovery point objective (RPO) of near zero data loss and a recovery time objective (RTO) of just about one minute.
Fact #4: You can create a cluster that mixes cloud and on-premises nodes.
You can use your on-premises data center as your primary location with a failover cluster to provide high availability protection and use the cloud as your hot standby DR site. This is a very cost-effective alternative to building out your own DR site, or renting rack space in a business continuity facility. In this case, the on-premises servers can be your choice of traditional SAN-based clusters, SANLess clusters, or even single servers not currently participating in a cluster.
The objective of having a "hot" standby DR site is to have standby servers up and running as quickly as possible in the DR site with access to a copy of the most recent application data. In the event of a disaster, recovery is automatic and immediate. A multisite cluster is an effective way to implement a hot standby DR site. In this case, the SANLess date. In the event of a forecasted disaster, such as a storm or a flood, applications can be moved to the cloud before potential disaster strikes. In the event of an unexpected disaster, applications can be recovered manually or in some cases automatically, depending upon the quorum configuration. This mix of cloud and on-premises nodes gives you an excellent RTO and RPO with minimal investment in infrastructure.
Fact #5: HA and DR in a cloud can be easy and highly cost-effective.
If you choose a SANLess software that provides an intuitive configuration interface, you can create a standard WSFC in a cloud in minutes without specialized skills. A SANLess cluster can help you realize significant cost savings in several ways. First, in a Microsoft SQL Server environment a SANLess cluster can give you high availability with SQL Server Standard Edition software licenses without requiring you to upgrade to costly SQL Server Enterprise Edition.
Second, you can realize hundreds of thousands of dollars in savings with a SANLess by eliminating the total cost of ownership (TCO) associated with a SAN. The savings in TCO include the SAN hardware acquisition costs; the power, cooling, and data center floor space costs; and the ongoing labor cost of specialized SAN administration.
If you are thinking about moving your important applications to the cloud, you need to consider how you will protect those applications from downtime and data loss. While traditional SAN-based clusters are not possible in these environments, SANLess clusters can provide an easy, cost-efficient alternative. These clusters not only provide high availability protection, but also enable significantly greater configuration flexibility and potentially dramatic savings in both licensing costs and SAN TCO.
2 Manyika, James and Michael Chui, et al, "Disruptive technologies: Advances that will transform life, business, and the global economy," McKinsey Global Institute (May 2013)
3Whittaker, Josh, "Amazon Web Services Suffers Outage, Takes Out Vine, Instagram, Others with it," ZDNet, (August 26, 2013)
4Mackay, Martin, "Downtime Report: Top Ten Outages in 2013," Business2Community.com, (December 2013)
Aug. 29, 2016 08:30 PM EDT Reads: 2,435
Aug. 29, 2016 08:15 PM EDT Reads: 1,866
Aug. 29, 2016 08:00 PM EDT Reads: 908
Aug. 29, 2016 07:45 PM EDT Reads: 1,658
Aug. 29, 2016 07:00 PM EDT Reads: 1,978
Aug. 29, 2016 06:45 PM EDT Reads: 828
Aug. 29, 2016 06:15 PM EDT Reads: 309
Aug. 29, 2016 06:15 PM EDT Reads: 790
Aug. 29, 2016 05:03 PM EDT Reads: 252
Aug. 29, 2016 04:30 PM EDT Reads: 3,564
Aug. 29, 2016 02:15 PM EDT Reads: 3,741
Aug. 29, 2016 02:00 PM EDT Reads: 993
Aug. 29, 2016 02:00 PM EDT Reads: 2,482
Aug. 29, 2016 12:45 PM EDT Reads: 2,026
Aug. 29, 2016 12:15 PM EDT Reads: 834