Resource slots#
In scenarios when you have many accounts trading in correlation (e.g. copying the same strategy) the trade or signal latency may increase because the accounts will compete for CPU available.
In order to reduce trade latency in scenarios like this you can allocate extra resource slots to the accounts so that each account have more RAM and CPU allocated to them.
High-level explanation of how resource slots works#
Each server has a fixed number of resource slots. You can think of them as virtual CPUs.
Please be aware about some specific scenarios which might affect your trade or signal latency.
Example 1. Trading uncorrelated strategies#
If you trade uncorrelated strategies on accounts, then at the moment trade is executed only single or a few accounts execute a trade. So that they share a plenty of free CPU and your trade executes with minimal latency even if a single resource slot allocated to the account.
Allocating more resource slots to accounts will not make any difference in this scenario because there is already plenty of resources available.
Example 2. Trading correlated strategies, uneven resource slots allocation#
Suppose now that the server has 16 resource slots available and you created 12 accounts with 1 resource slot and 1 account with 4 resource slots. Suppose the accounts will be placed on the same server. Suppose also you are trading the same strategy on all these accounts.
When you execute the trade, 13 accounts will execute it concurrently, so all CPU power will be shared across 13 accounts.
This means that you will not see any significant improvement in trade copying latency in this scenario for the account which have 4 resource slots assigned.
Example 3. Trading correlated strategies, uniform resource slots allocation#
Suppose now that the server has 16 resource slots available and you created 4 accounts with 4 resource slots each and the system placed them all on the same server. Suppose the account are trading the same strategy.
When you execute a trade, only 4 accounts will execute it concurrently, so that each will get 4x more CPU power.
This means that you will see a significant improvement in trade copying latency in this scenario.