Trader Virtual Desktop Infrastructure (VDI)

Overview

A remoted desktop environment with two machines per trader, one keyboard and mouse, and up to 8 screens.

Historically traders had two machines: One (trading machine) ring-fenced for Bloomberg and trading, and a second (non-trading machine) for MS Office, surfing the web, some 3rd party apps etc.. They would swap between the machines using KVM switch boxes. We switched to a VDI style solution where the trader’s machines are moved to the datacentre and accessed via a ‘thin client’ terminal.

TraderVDIArchitecture2

LeoStream was the only connection broker that could broker to both hosted desktops and virtual desktops, and broker one user to many concurrent desktops.  And HP RGS was the only remoting protocol that could position remoted screens side by side.  These two products worked together well.  The “thin client” as a standard HP small form factor desktop with Nvidia Quad or dual video cards to support the number of screens required.

TraderVDIArchitecture

VMware View was used to deploy a pool of non-trading VMs. And HP Blade Workstations were used for the traders’ trading machines.

Leostream can use Active Directory OUs to discover pools of machines, and AD groups to match users too pools of machines. AD groups were also used to manage screen layout configurations until these were standardised. The AD machine account property ManagedBy was used to assign machines in pools to users.

Once configured, support staff could use familiar AD user and computer management tools to change how brokering works for new joiners/leavers.

If the main datacentre was unavailable, Traders could be directed to machines in an alternative datacentre by swapping their AD group membership. At logon a group policy filtered on AD membership replaces the Windows explorer shell with Leostream client directed to the active datacentre.

If the main office was unavailable (power cut for example) traders could go to a standby office and continue their sessions on their machines running in the datacentre.

Trading

Services

/

Project details
  • Date: December 2012