Low Latency
Latency's impact on business performance is increasing by the day. We’ve designed our colocation footprint to provide broad facility coverage with optimal physical paths.
-
Proximity
Physical location is both strategic and critical in reducing overall latency. We provide clients with desirable space in the same facilities that the house exchange matching engines to provide the shortest path possible.
-
Easy Deployment
Optimizing physical footprints often requires geographically dispersed facilities, making service a challenge. We provide facility services like equipment installation and hot hands to eliminate the need for clients to be on site.
-
Secure
Critical business processes are served by equipment in our care. We use industry "best practices" to ensure that equipment is physically secured and that only authorized personnel have access to our space.
Cost-Intelligent
Often clients have applications that do not require ultra low latency. For these applications, we provide high quality colocation space in more cost-effective facilities to reduce a client’s overall cost.
-
CORE NETWORK
Our cost-intelligent facilities are connected into our core network which provides the same broad access as our low latency space.
-
HIGH QUALITY
Ultra low latency facilities command higher rates due to their proximity to venues. We are able to offer more cost-effective space while maintaining an uncompromisingly high level of quality.
-
SAME SERVICES
Since our services are consistent across both our low latency space and our cost-intelligent space, a client can choose to use either or both models.
Shared Model
Many firms mistakenly think they must purchase and manage their own networking equipment. In our colocation venues, we provide managed, shared switching, wherein a customer's traffic remains segregated to their own environment.
-
FULLY MANAGED
Managing network equipment can be difficult, time consuming, and sensitive to change. We have a highly qualified team that effectively manages our shared environments to ensure security, reliability, and speed.
-
PERFORMANCE
While it may seem counterintuitive, shared environments can actually increase performance. Since this is a core offering of Deep Systems, our engineers focus on optimizing shared switching solutions.
-
COST-EFFECTIVE
Firms often deploy large amounts of capital on dedicated network equipment that is not utilized anywhere near its capacity. In a shared environment, clients only pay for the networking resources they need.
Dedicated Model
There are clients that prefer to control and manage their own network switching. We accommodate their chosen switching solutions without imposing limitations.
-
FULL CONTROL
Clients that prefer this model can control the entire network stack. This includes but is not limited to the hardware deployed, how said hardware is managed, as well as the network model and protocols.
-
CORE SWITCHES
While a client can choose to manage the networking in their environment, they may still want connectivity to our core network. In most cases we can accommodate access to our core switches.
-
CONSULTING
For clients in need of assistance architecting or managing their customized network switching environment, we offer consultative and ongoing maintenance services to provide expertise or even helping hands.
Analysis Framework
The Big Data era has created new opportunities for firms that discover ways to monetize huge data sets. We’ve designed an agnostic way to view, store, and process key information to better analyze real-time and historical data sets.
-
USER-DEFINED EVENTS
To provide useful and accurate analytics, clients define what events they consider to be key information. This information can be any field or groups of fields, rolling samples or data sets, or any other samples that provide value.
-
FILE OR MESSAGING
A client has the choice of storing the data in a well-defined file format and/or broadcasting the information over our messaging systems. Results can be viewed or processed in real-time, or saved for future processing.
-
SIGNAL & TRADE ANALYSIS
Our tools facilitate iterative analysis of large market data sets along with real-time feedback from trading systems. A practical use of these analytics is post trade analysis or in developing and tuning signals.
Beacon Framework
The old paradigm for monitoring systems availability focuses on whether certain system processes are alive or dead. We've architected a much more sophisticated model for managing overall system health and current system capabilities.
-
SERVICE DISCOVERY
Our system is designed to allow a process to determine service characteristics and availability. By doing so, the process can decide whether the necessary components are in the proper state to accomplish an objective.
-
HEALTH STATUS
As with any resilient system, our beacon system provides a mechanism to broadcast the current state of the service. This allows other processes to make informed decisions on how to interact with the overall system.
-
REAL-TIME MONITORING
The breadth of information contained in our beacon facilitates more sophisticated real-time monitoring. Exceptions can be handled in more detailed and pragmatic ways, well beyond the industry’s existing paradigm of basic client notification.
Notification Framework
Notifications are an important component of managing critical production systems, but are often relegated to simple emails. We support an agnostic notification framework that can notify clients of an issue using multiple mediums.
-
SYSTEM NOTIFICATION
By default, our core system will create notifications for any exception to expected operational behavior. These notifications provide the basis for a client to monitor the health and state of their extensions to the Deep Systems platform.
-
USER DEFINED
Clients can create custom notifications for exceptions to the operational behavior of their extensions to the system. These user-defined messages permit a notification model as granular as the client desires.
-
3RD PARTY INTEGRATION
Out of the box, we support multiple models for issuing notifications. The architecture of the notifications system is extensible and allows for easy implementation of a 3rd party notification model.
Profiling Framework
Understanding the performance characteristics of a system is an important part of creating consistent results. Profiling can provide important performance data on production systems or development systems before they are released.
-
INTERVAL SAMPLING
In production systems, it is generally not possible to issue performance feedback on every event. For this reason, we sample all events over a sampling period and issue performance details over the life of the sample.
-
PERFORMANCE BROADCAST
The performance characteristics of a service can provide an indication of the health of that service or the health of the overall system. We broadcast these performance samples for aggregated insight into overall performance.
-
CONFIGURABLE THRESHOLDS
Acceptable performance thresholds vary between systems. Deep Systems enables users to configure appropriate thresholds for their own system performance exceptions.
Leasing
Leased servers are a great option for clients that want to get up and running quickly with minimal upfront cost.
-
FLEXIBLE
We offer a range of lease terms and structures to fit the need of your project.
-
FAST
Deep Systems coordinates the entire startup process which reduces complexity and implementation time.
-
EASY
Our experienced team handles the entire process of sourcing equipment, installs, networking, and any client specific implementation steps.
Purchasing
For clients prepared to make a hardware purchase, we can lend our experience and vendor relationships to the selection process. In many instances, our buying power may yield beneficial pricing.
-
EXPERIENCE
We have developed deep expertise in the cost, benefits and risks associated with a wide array of hardware deployments.
-
COST SAVINGS
We generally have larger buying power than our clients. This may result in beneficial pricing.
-
RELATIONSHIPS
Our relationships among vendors and suppliers provide us direct access to their domain experts, who often provide critical input to a purchase decision.
Raw Feeds
Some firms require direct network access to raw market data disseminated from the venues they interact with. We work with our clients to obtain venue approval for this data and provision it to their environments.
-
LOW LATENCY
Consumers of raw market data are generally latency sensitive. We architect our networks to target the least number of hops between the venue and our clients’ devices.
-
VENUE NATIVE
Unless otherwise specified, we do not modify or shape the raw data feed in any way. This ensures that our clients will be able to receive and process the data feeds in the same way that the venue transmits them.
-
INSIGHT
We process many raw data feeds as part of our Software Services business and a number of our clients consume raw data. This often enables us to provide our customer base with timely, valuable information upfront or during an event.
Vendor Of Record
The landscape of market data is both dynamic and complex making it difficult to stay current. We assist our clients with upfront and ongoing compliance support and help them adapt to frequent rate changes.
-
ENTITLEMENTS
We employ an automated entitlement system that allows our team to efficiently and effectively permission data. This ensures we satisfy venue compliance obligations while remaining fast and reliable.
-
EDUCATION
Our clients value our assistance keeping them abreast of the latest rules and regulations regarding the use of venue data and the numerous, relevant fee and structure changes across the industry.
-
REPORTING
Venues often require users to report which devices are receiving their data. There are frequent opportunities to decrease our client’s reporting burden by applying our automated processes and tools.
Overview
Messaging is a foundational piece of our technical pyramid. It was designed with well known paradigms to create a performant and flexible model for sending and receiving messages between the different constituents of our ecosystem.
-
PUBLISHER & SUBSCRIBER
Our messaging adopts the “Publisher & Subscriber” broadcast paradigm which allows senders to publish messages and interested subscribers to listen. Subscribers can be modified without affecting the publisher.
-
REQUEST & REPLY
Our messaging options also include a more direct “Request / Reply” messaging protocol. This is intended to facilitate specific one-to-one messaging relationships in cases where it is desirable.
-
OPTIMIZED TOPIC RESOLUTION
Our messaging employs a deterministic topic system that enables our optimization of the topic footprint based on the hardware in use. We can then plan the topic space flexibly to accomplish specific goals like message rate balancing.
10G & 40G Ethernet
Deep Systems' messaging supports the use of 10G & 40G ethernet switches and cards. Ethernet solutions are ideal for private and shared environments where a variety of publishers and subscribers can coexist without interference.
-
UDP MULTICAST
UDP multicast is a mature and broadly used protocol. It is widely supported and available on most, if not all, ethernet software and hardware stacks. Our support is intended to provide the broadest coverage on ethernet.
-
SOLARFLARE OPTIMIZED
Solarflare enables us to provide a highly optimized ethernet messaging solution. We utilize native APIs to reduce latency and increase throughput by directly accessing the card and making communication as fast as possible.
-
RESOURCE MANAGEMENT
We employ multiple techniques to balance the load of receiving messages with the computational costs of processing them. This balance helps to prevent buffers from filling up while still processing messages efficiently.
Infiniband
Our messaging supports the use of Infiniband switches and cards. Infiniband solutions are ideal for highly-controlled, private environments, where lossless protocols can be used in a safe and deeply optimized manner.
-
IB MULTICAST
IB multicast allows us to provide broadcast capabilities in Infiniband environments. We can then implement a broad set of paradigms on hardware that is optimized for low latency, lossless messaging protocols.
-
MELLANOX OPTIMIZED
Mellanox enables us to provide a highly optimized Infiniband messaging solution. We utilize native APIs to reduce latency and increase throughput by directly accessing the card and making communication as fast as possible.
-
RESOURCE MANAGEMENT
We employ multiple techniques to balance the load of receiving messages with the computational costs of processing them. This balance helps to prevent buffers from filling up while still processing messages efficiently.
Inter-Process
Inter-Process messaging allows for sending messages between processes on the same machine in a highly optimized manner. In cases where the software solution can stay local on the same machine, it dramatically improves throughput and latency.
-
MACHINE LOCAL
Machine local messaging allows the solutions to be implemented elegantly and appropriately while using APIs consistent with networked messaging.
-
MULTIPLE PUB & SUB
Inter-Process messaging solutions generally provide single queues and single paths. Deep Systems provides full support for multiple publishers and subscribers in a similar fashion to our networked messaging implementations.
-
LOWEST LATENCY
Machine native protocols offer greatly reduced latency and increased throughput relative to their network inter-machine counterparts.
Core Network
The broad coverage of the Deep Systems core network enables multiple points of presence with minimal latency and maximum resilience. This eliminates providers and reduces costs for our clients.
-
LOW LATENCY
Our clients rely on us to ensure that our circuit and fiber paths between points of presence are optimized. We view this as an ongoing effort and monitor our deployments as new paths become available.
-
REDUNDANT
While latency is important, it is critical to have redundant diverse paths throughout the network to prevent downtime. We also look to diversify building ingresses, equipment paths, etc. for added stability.
-
BROAD COVERAGE
Our clients connect to venues, brokers, regulators, 3rd party vendors, and customers. The Deep Systems network has broad coverage across the financial space and can quickly facilitate any necessary connectivity.
Intra-Data Center
This network is responsible for all traffic within a given point of presence. Whether traffic is originating from another point of presence or from within the same data center, it must travel on the intra-data center network.
-
HIGH PERFORMANCE
As with our core network, network latency within a facility is critical to overall performance. We implement an optimized network architecture with high performance network equipment to ensure ultra low latency.
-
DIRECT ACCESS
Often times our clients require the shortest path to a counterparty, such as a trading venue. We offer the ability for clients to access the properties within the point of presence with a direct path for optimal performance.
-
RESILIENT
Both our core network and our intra-data center networks are core to our client’s operations. As such, we design our intra-data center networks to utilize redundant hardware, diverse in-facility paths, and other reliability features.
External Access
Deep Systems customers are operating mission-critical systems that often demand 24/7 accessibility. We provide multiple options to assure that they always have the desired type of access to their environments.
-
Private Lines
For the most demanding client, private lines offer dedicated connectivity with specific uptime and bandwidth SLA guarantees.
-
Software VPN
We offer secure software-based VPN solutions that allow login access the client environment from anywhere with Internet access.
-
Hardware VPN
A site-to-site VPN tunnel is a great solution when convenient, secure, "always on" Internet access is desired for a user or group without login.
Position Management
We provide the technical bookkeeping necessary to load positions at the start of the day and apply position updates from different sources throughout the day. This provides a consistent, reliable, and current source for positions.
-
FULL INTEGRATION
Our position management protocols are deeply integrated throughout our platform fostering consistency across trading, risk management, external views, and any other system that requires positions to accomplish its objective.
-
MANUAL ADJUSTMENTS
While our system automatically handles most situations, we support manual position adjustments. Examples of this would include trade breaks, brokered transactions or account movements.
-
DECENTRALIZED
Position management is among the most critical components of a financial system and must always be available. We've architected our platform to decentralize the management of positions, thus making the system more resilient.
Order Recovery
Of the many critical components in a financial technology platform, the order management system is near the top of the list. We recognize the the risks inherent in an outage and have architected the Deep Systems platform with robust and reliable order recovery processes.
-
FULL ORDER LIFECYCLE
Order recovery should ensure that a full order state can be reinstated. Our order recovery mechanism restores the entire lifecyle of orders, allowing the recovering process to rebuild its full state.
-
REDUNDANT STORAGE
Order recovery is only possible if you have all order data stored in an accessible and reliable way. We’ve built a reliable, high speed, redundant storage system to support a high volume of order traffic utilizing multiple storage locations.
-
QUERY API
Clients may want to recover or access a subset of information among a larger order data set. We provide a mechanism to query the larger data set and return only the targeted data.
Clearing Integration
Larger clients may utilize multiple clearing/prime broker relationships. We integrate with a number of clearing entities thus decreasing the technical dependency on a specific clearing firm.
-
POSITION LOADING
The Deep Systems platform manages the start of day process of loading and reconciling positions from one or more clearing firms. The process is designed to handle various secure file dropping models or proprietary APIs.
-
TRADE BREAK DETECTION
We help prevent costly errors with reconciliation tools to compare system data with clearing records in order to detect trade breaks or other accounting issues.
-
EXCEPTION NOTIFICATIONS
Timely awareness can often mitigate large risk exposures. Our notification system can expedite the resolution of exceptions by notifying interested parties.
Virtual Accounts
Financial operations have varying levels of complexity in their strategies, groups and accounts. We've designed a virtual account system that can model the intended structure of an operation without any compromises.
-
GRANULAR ASSOCIATIONS
A client can design their virtual accounting structure to be as granular or as coarse as they desire. A trade can be attributed to any logical entity in the operation including trader, strategy, group, or account.
-
FLEXIBLE GROUPING
Clients can design trade groups in any way that makes sense for their operation to create a tailored virtual account structure. Trades can be attributed to one or many groups with no restrictions in the way they are grouped.
-
BROAD INTEGRATION
Our virtual account system is deeply integrated throughout the Deep Systems platform. This allows for consistency across clearing, trading, risk management and any other account-based system.
Directory Services
Active Directory is the basis for managing the access and policies of your employees. We provide the facilities to manage directory instances locally, in our data centers, or in a distributed manner.
-
USER MANAGEMENT
Deep Systems assists with the on-boarding and off-boarding of employees while maintaining a full audit history. We coordinate with your team to schedule any additions, removals, or changes to your user directory.
-
ACCESS CONTROL
Active Directory requires a user to provide credentials to gain access to your corporate environment. We will help you design and implement the permissions, roles, and responsibilities that are appropriate for your environment.
-
POLICY MANAGEMENT
Active Directory is a centralized location that provides the capacity to define various levels of policy. We'll help you define policies to ensure consistency with organizational goals.
Single Sign-On
Increasingly employees are responsible for performing day to day tasks in multiple systems. Single Sign-on allows users to securely login once, granting secure access to all their systems without separate login.
-
DASHBOARD
One of the productivity boosting benefits of our single sign-on solution is a dashboard which acts as a central location where users are able to login once then, seamlessly and securely, access all their applications.
-
INTEGRATION
Our single sign-on solution seamlessly integrates with the client’s existing directory services. This enables the solution to provide ease of use for employees without adding additional management complexity.
-
INTUITIVE
Often times security solutions can add additional complications for employees and decrease overall productivity. Our single sign-on solutions are intuitive to use and make accessing multiple systems easy.
Multi-Factor Authentication
In today's workforce, productivity happens at home, on the road, or any number of places outside of the office. Multi-factor authentication adds an important layer of safety, increasing confidence that remote access to your system is done in a secure manner.
-
PROTECTION
To gain access to your secured applications a user provides a second form of identification that is only known to them. This greatly reduces the chance that a compromised password alone can overcome your security.
-
POLICY-BASED
While multi-factor authentication can greatly increase your overall security, it does require an additional step by the employee. Setting policies allows the client to decide what conditions should require the additional step.
-
ADAPTIVE
In additional to the credential-based security provided by multi-factor authentication, our solution provides additional adaptive profiling for added security. This includes monitoring threat feeds and suspicious IPs.
Order Gateways
We provide normalized access to regulated venues, broker dealers, and any other entity that accepts orders regardless of the protocol. The system is architected to be ultra-low latency without sacrificing reliability.
-
FIX SUPPORT
Our Fix implementation supports all major versions as well as standard and non-standard implementations. We also support additional venue specific optimizations to further decrease overall latency to Fix venues.
-
BINARY SUPPORT
Where available, we provide support for a venue’s native binary protocol, which generally provides the most optimized and lowest possible latency for order entry.
-
LOAD BALANCING
We support the ability to spread orders over multiple active gateways in order to load balance order traffic. This increases throughput and reliability while decreasing latency and reducing gateway concentration.
In-Process Gateways
In-Process gateways permit orders to be encoded and sent in the same process as your code. This reduces the number of network hops and overall latency.
-
LOWEST LATENCY
By being in the same process, the client benefits from encoding and processing orders with no additional hops. This provides a model to solve the most latency sensitive problems without adding additional complexity.
-
OPTIMIZED STACK
In-Process gateways are part of an optimized stack including in-process feeds and inter-process messaging. These complementary components provide the basis for implementing the lowest latency solutions.
-
SAME COVERAGE
In-Process is a deployment model as opposed to a different software stack. Unlike other systems, the same features and gateway coverage are available for the in-process gateway that are suppored in other models.
Order Routing
In addition to supporting a variety of venues, we support a model to aggregate multiple venues under a single routing interface. This allows clients to access multiple venues in a greatly simplified format.
-
FIX INTERFACE
We enable clients to offer access to multiple venues through a standard normalized Fix interface. A unified interface in the most common protocol decreases obstacles to adoption.
-
CUSTOM VENUES
We support order routing to internal or external systems in client-specified ways. This can be used to normalize access to a venue, a crossing mechanism, custom algorithms or other unique end points.
-
CUSTOM ROUTING
Routing may be configured in a variety of ways; by tags, by customer, by session, by account, or any combination of information available.
Smart Routing
Often, routing decisions require more than basic policy. In these cases, it is necessary to listen to real-time data or regularly update data sets to make more sophisticated routing decisions.
-
MULTIPLE ALGORITHMS
We support the implementation of multiple algorithms to permit diversity in smart routing offerings. A client may want to provide a variety of algorithms, each targeting a different routing objective.
-
OUT OF THE BOX
We have out-of-the-box implementations of many commonly requested algorithms. Clients can use and extend our algorithms, which reduces the time to develop and test while facilitating the desired customization.
-
LOW LATENCY
Our proprietary ultra-low latency framework reduces the overhead of routing logic which increases the probability of achieving a routing objective.
Drop Copy
Drop copies provide a real-time mechanism for importing or exporting order data to and from external systems.
-
INBOUND
We support inbound drop copies via the Fix protocol. These drop copies can be normalized and forwarded over our messaging systems as well as update positions and internal risk management systems.
-
OUTBOUND
Generate outbound drop copies via the Fix protocol, including full order traffic or only fill data. We can also aggregate inbound drop copies to provide a unified feed of internal and external order activity.
-
HIGH THROUGHPUT
As with all of our order-related systems, the drop copy system is designed to handle high throughput, which allows for aggregating or disseminating large amounts of order traffic to and from external systems.
Risk Checks
In evolved trading environments, risk checks are necessary to trade responsibly and satisfy regulatory obligations. We provide a distributed risk system designed to enforce seamless, pre-trade risk checks.
-
EXTENSIVE CHECK COVERAGE
We support a large number of risk checks out-of-the-box. These include fat finger checks, rate limiting, credit-based checks, regulation specific checks, restricted lists, asset-specific checks, and many others.
-
REAL-TIME ALERTING
Understanding the state of risk checks is critical to managing them. We provide configurable notifications for order rejections, thresholds, and other relevant events.
-
EXTENSIBLE
Our risk checks are built on top of a distributed risk framework that enables speedy development of custom risk checks. Clients can add checks specific to their systems to ensure risk management is tailored to their specific use case.
Overview
Many systems access data and services from a remote location. For these systems, we provide mechanisms to access data, services, or other functionality in a broad number of contexts.
-
TCP/IP
TCP/IP is a widely supported protocol that operates in many technology environments. Our unicast access relies on TCP/IP as a way to achieve the broadest coverage of use cases.
-
PUBLISHER & SUBSCRIBER
The publisher & subscriber model is used throughout our messaging protocols and other core systems. We are able to deliver performance and reliability while keeping our unicast access system consistent across the platform.
-
REQUEST & REPLY
The request & reply model is also used broadly in our overall platform. By supporting a request & reply model, unicast access clients can issue one-off requests in cases where the publisher & subscriber model would be excessive.
Service Director
When dealing with unicast protocols, each client has a separate session to send and receive data. To manage the load balancing and performance of this model, the service director acts as a mechanism to optimally direct traffic.
-
SERVICE DISCOVERY
The service director acts as the arbiter of all the service offerings accessed by clients. As such, it can provide the ability for a unicast client to discover which services are currently offered by the director.
-
LOAD BALANCING
Improperly designed Unicast Services can experience load issues when required to send data to each subscriber separately. Our director is able to spread the load over all available handlers for a particular service.
-
FAULT TOLERANCE
Due to the session based nature of unicast services, disconnects can happen when a service handler becomes unavailable. Our service director provides a mechanism to automatically redirect the client to a new service handler.
Service Handlers
Our service handler framework allows for quickly developing methods to interact with the requests from a unicast client. The framework reduces the complexity of implementation while allowing each handler to interact with requests uniquely.
-
PREBUILT LIST
We have built sophisticated coverage for most of the core systems enabling out-of-the-box deployment. This includes delivering market data, order entry, position management, reference data, and other core offerings.
-
DIRECTOR INTEGRATION
The service handler framework manages the interaction between the handler and the service director. This provides a mechanism to notify the directory of current availability and the current state of each handler instance.
-
EXTENSIBLE
The service handler system is architected to focus development on the data or service being offered. This enables our clients to build and maintain a custom service handler in an efficient manner.
Entitlements
As with our core system, entitlements are fundamental for data security and to meet regulatory requirements. We architected an entitlements system designed to solve the problems specific to the session-based nature of unicast.
-
SECURE AUTHENTICATION
To ensure that security is being enforced and to confirm the recipients of data and services are entitled, we implement a user & password login system. At the time of login, all user and password information is encrypted for security.
-
TOPIC-BASED
Post authentication, access is controlled by our topic-based permissions system. A user can be provisioned for various levels of access to a specific topic or group of related topics.
-
RESOURCE-BASED
In addition to topic-based controls, a user can be controlled based on the amount of resources they have access to. Limiting a user's consumption of valuable resources benefits the health of the overall system.