Using Extended Vhosts for Unified HA/DR in Financial Services
A Purpose-Built Solution for a Global Enterprise
Global financial services corporations manage some of the largest SQL Server environments in existence. End-user services aren’t an easy thing to differentiate in this industry either, so these massive enterprises depend on their ability to keep services available and data protected as their primary means to sustaining a reputable brand and loyal customer base. This means a massive emphasis on high availability (HA) and disaster recovery (DR) capability, because it only takes one wrong step to lose customers by the thousands.
This use case describes the HA/DR conundrum one of our largest financial services customers faced that prompted DH2i to create and roll-out Extended Vhost technology for the first time.
Key Customer Challenges
- Achieve Seamless Business Continuity: Needed to support both intra-site high availability (HA) and cross-site disaster recovery (DR) simultaneously, despite limited options available in the market.
- Optimize Storage Replication Costs: Needed to minimize or eliminate storage replication costs across heterogeneous datacenters, given constrained budgets that limited investments in new technologies.
- Streamline Infrastructure and Licensing Expenses: Needed to reduce Windows Server Failover Clustering (WSFC) footprint and leverage Linux to lower overall infrastructure and licensing expenses.
not available
Solution: Unified HA/DR with ExVhosts on Linux
DH2i provides platform-agnostic HA for SQL Server with DxEnterprise (DxE) Smart High Availability Clustering software (a standalone replacement for solutions like WSFC and Pacemaker). The flexibility of this solution allowed this customer to drastically reduce licensing costs by moving to Linux.
DxE utilizes proprietary Virtual Host (Vhost) and extended Vhost (exVhost) technology to allow HA instances and Availability Groups (AGs) to be combined in a gainful way, creating an AG across HA instances rather than standalone SQL Server instances.
The DxEnterprise concept of a Vhost is very similar to a WSFC cluster, and has three kinds of things attached to it:
- Members: Cluster nodes/VMs/hosts where the Vhost can be run
- Applications that run as part of the Vhost, e.g. SQL HA, SQL AG
- Shared disks that belong to the Vhost, which must be available to start the applications
An exVhost is different. Instead of only having nodes/VMs/hosts as members, it can have entire Vhosts as members. This is how we represent an AG that is created between HA instances. Each HA instance is described in a regular Vhost. The AG is described in a parent exVhost, with the HA Vhosts as its members.
A typical exVhost setup for this customer looks like this:
- Two sites, two VMs per site, for a total of four VMs
- Shared storage at each site, for a total of two sets of storage
- One HA instance at each site. The instance can failover between the two VMs at that site, and is attached to the storage at that site.
- One AG between sites, with each site’s HA instance as a member of the AG
Tiered High Availability Approach
This financial services corporation has hundreds of application teams, each with their own specific SLA/uptime requirements. For this reason, they implemented a tiered HA approach in which teams with the strictest requirements have both HA and DR capabilities, and teams with lower requirements deploy smaller, less complex environments to avoid extra costs and management. Their specific tiered model looks like this:
Tier 1 (highest): Multi-site HA+DR
- HA instances at two different sites (each site has two VMs, shared storage, failover if the active VM fails)
- AG replication between sites
Tier 2: Multi-site AG, without HA
- Plain (non-HA) instance at two different sites
- AG replication between sites
Tier 3: Single-site HA
- HA instance at a single site
Customer Benefits of Extended Vhosts
Unlocked huge licensing savings by moving SQL Server to Linux and leveraging DxEnterprise for unified HA/DR
Leveraged exVhost technology so that they could rely on built-in SQL Server local HA and included AG replication
Created a tiered approach to organizational HA to optimize resource use and minimize management costs
Eliminated VM and physical server sprawl by letting HA instances and AGs all reside in one single DxE cluster