The Unthinkable Trade: Disaster Recovery and High Availability in Cloud and Colocation

For a trading firm, downtime is not an inconvenience; it is an existential threat. A system outage during market hours can result in millions of dollars in direct losses, as well as the intangible but equally damaging costs of lost customer trust and reputational harm. A robust disaster recovery (DR) and high availability (HA) strategy is therefore not a luxury, but a non-negotiable requirement of doing business. The choice between cloud and colocation infrastructure has profound implications for how a firm approaches this important challenge. Each environment offers a different set of tools and trade-offs for building a resilient and fault-tolerant trading operation.

The Spectrum of Resilience: From Backups to Active-Active

Before comparing the DR/HA capabilities of cloud and colocation, it is important to understand the spectrum of resilience options available:

• Backups: The most basic form of DR, involving the regular copying of data to a separate location. Recovery from a backup can be slow and may result in some data loss.
• Cold Site: A secondary data center that is equipped with the necessary power and cooling, but no hardware. In the event of a disaster, hardware must be shipped to the site and configured, a process that can take days or even weeks.
• Warm Site: A secondary data center that is equipped with hardware that is ready to be powered on. Recovery is faster than a cold site, but still may take several hours.
• Hot Site: A secondary data center that has a live, running copy of the primary site. Recovery can be almost instantaneous.
• Active-Active: Two or more data centers that are all actively serving traffic. If one site fails, the others can seamlessly take over its workload with no interruption in service.

Colocation: The Fortress with a Single Point of Failure

Traditionally, trading firms have relied on a multi-site colocation strategy for DR/HA. This typically involves a primary data center, where the main trading operations are located, and a secondary data center in a different geographic location. The two sites are connected by a high-speed, private network, and data is replicated between them in real-time.

This approach can be highly effective, but it is also extremely expensive and complex to manage. The firm must bear the full cost of duplicating its entire hardware stack, as well as the ongoing costs of maintaining the secondary site and the private network connection. Furthermore, a traditional colocation setup, even with a secondary site, can still have single points of failure. A major regional disaster, such as a hurricane or an earthquake, could potentially take out both the primary and secondary data centers if they are not sufficiently far apart.

The Cloud: A New Paradigm of Resilience

The cloud offers a fundamentally different approach to DR/HA, one that is based on the principles of virtualization, automation, and geographic distribution. Cloud providers have a global network of data centers, known as "regions" and "availability zones" (AZs). A region is a separate geographic area, such as the US East Coast or Western Europe. An AZ is a distinct data center within a region, with its own independent power, cooling, and networking.

This global infrastructure provides a effective set of building blocks for creating highly resilient and fault-tolerant applications. A firm can deploy its trading application across multiple AZs within a single region, so that if one AZ fails, the application will continue to run in the others. For even greater resilience, the application can be deployed across multiple regions, so that it can withstand the failure of an entire geographic area.

A Cloud-Native DR/HA Strategy: A Practical Example

A cloud-native DR/HA strategy for a trading firm might look like this:

1. Multi-AZ Deployment: The firm’s trading application is deployed across three AZs in a single region. Each AZ has a full copy of the application, and a load balancer distributes traffic between them.
2. Real-time Data Replication: The firm’s trading data is replicated in real-time across the three AZs, so that there is no data loss in the event of an AZ failure.
3. Automated Failover: If one AZ fails, the load balancer automatically redirects all traffic to the remaining two AZs. The failover is seamless and transparent to users.
4. Multi-Region Backup: For added protection, the firm also backs up its data to a different region. In the unlikely event that the entire primary region fails, the firm can restore its application and data in the secondary region.

This approach provides a level of resilience that would be prohibitively expensive and complex to achieve with a traditional colocation setup. The cloud’s pay-as-you-go model means that the firm only pays for the resources it is actually using, and the cloud provider handles all of the underlying infrastructure management.

The Latency Trade-Off

The one major drawback of a cloud-based DR/HA strategy is latency. The geographic distance between regions means that there will always be some latency when replicating data or failing over to a secondary region. For the most latency-sensitive trading strategies, this may be unacceptable. In these cases, a hybrid approach may be the best solution, with the primary trading operations in a low-latency colocation facility and the DR site in the cloud.

Conclusion: Resilience as a Service

The cloud has democratized access to high-end disaster recovery and high availability. What was once the exclusive domain of the largest and most well-funded financial institutions is now available to any firm, regardless of size. By leveraging the global infrastructure and sophisticated automation of the cloud, trading firms can build a level of resilience that was previously unimaginable. In a world of increasing uncertainty and market volatility, this is a effective competitive advantage.