The Business Case for Storage Networks [Electronic resources] نسخه متنی

Storage Environment

This firm's primary datacenter is hosted in London; secondary datacenters in Chicago and Paris also provide additional capacity and resiliency.

The primary datacenter and the Paris datacenter are connected to each other over a pair of Synchronous Digital Hierarchy (SDH) rings (the European equivalent of a Synchronous Optical Network [SONET] ring). These links are completely independent of each other and share no common infrastructure. Chicago attaches to the optical ring via an IP Point of Presence (POP) in London.

The IT operations group currently supports 60 TB of raw storage across all three locations; the application data is replicated to each datacenter as dictated by business need.

Approximately 20 TB of that 60 TB is for core applications hosted on Virtual Memory System (VMS) platform servers from Digital Equipment Corporation connected to a Brocade SAN infrastructure. The remaining 40 TB of raw storage is split between the main datacenter, Chicago, and Paris and resides on a mix of high-end and mid-tier storage arrays from Hitachi Data Systems™. 10 Cisco MDS 9509 multilayer director switches form the core SAN infrastructure.

The Cisco MDS 9509s are configured with a combination of 16- and 32-port line cards to take advantage of both the line-rate performance of the 16-port cards and the price performance of the 32-port cards. In general, the 32-port cards are used primarily for development and test environments (which require less maximum throughput at peak performance times), whereas the 16-port cards are used for storage and connectivity for production hosts.

Note

This configuration takes advantage of the 3.38 to 1 oversubscription rate of the 32-port cards on the MDS platform.

Each application environment is typically configured for three times the maximum required performance. Although an average day might see 3.6 million trades, the firm must provide headroom for sustained performance of over ten million trades. It is critical for the firm to process all trades in a single day so that trades can be cleared through a clearing house, such as ClearNet, in time to make daily margin calls. In no way can the firm afford to get behind in its transaction processing; the downstream effects of a backlog of trades could be economically disastrous.

Each trade goes through a series of gates from bid to completion as part of its life cycle. The trade registration system (TRS) receives trades through trading desktops and application front-ends. From there, the relevant data for each trade (contract specifications, expiration dates, and so on) is passed to the primary database where the trade data is then logged and stored. When the trade is executed, the trade data is sent back to the TRS. From the TRS, it is sent to the clearing house for reconciliation of margin calls. As the final step of each trade's life cycle, data from the clearing house regarding the client'ss financial standing and risk exposure is then posted back into the primary database.

This type of environment (multiple-staging environments for data transfer between numerous applications) is common to many financial institutions, as is the reliance upon storage networking technologies to provide increased availability and reliability.