We are literally paranoid about how we run real-time market data infrastructure.
Colocated infrastructure is either a problem you manage or a problem someone already solved. algoseek has spent two decades making sure it’s the latter.
Our infrastructure at a glance
2
decades building trading infrastructure
3
Equinix facilities: NY2/NY4, CH1
0
SLA failures, even under market volatility
4-way
arbitration across redundant feeds
30TB
market data processed per day
2
regions: fully redundant across NJ and Chicago
Trusted by
Two US regulators
Equity SIP feeds and custom OPRA NBBO calculations
Bulge bracket banks
Index TWAPs and intraday QIS calculations
Hedge funds
Raw multicast delivered into AWS across regions
Prop trading firms
Custom-built servers and private network delivery
How algoseek builds trading infrastructure
Things break. The question is whether you built for it.
Two decades of digging into every problem encountered, from data congestion to CPU meltdowns, and preventing it from happening at all.
Custom servers, not off-the-shelf
Manufacturers skimp wherever they can. algoseek uses custom copper coolers, matches motherboards and memory by production year, and puts dedicated-chip network cards in capture servers.
“
We glue the damn fan, so it stays where it belongs. And not just the fan. We glue cables so they never wander into a fan’s spinning blades.
Testing: twice before it touches a rack
Every component tested individually, every server tested at the office, the full suite run again at the data center. One failure there and the unit goes back.
Preventive monitoring, not failure detection
algoseek’s own tools read low-level OS logs for early indicators: an SSD going from 5 bad sectors to 20 in a week, a switch issuing pause frames before packets drop.
“
When you monitor for failure, you already failed.
Changes only on Saturdays
No changes during the production week: touching one server can dislodge a cable on the next. Visits are scripted and logged to the port; mid-week failures get a pre-tested spare.
Data integrity as the goal
Listener servers catch dropped packets and compare exchange timestamps against server timestamps. algoseek over-monitors rather than explain a gap.
No trust in commercial products
Commercial motherboards lock you into their upgrade path; commercial monitoring can’t see congestion on individual ports. algoseek builds its own because the alternative is paying more for less visibility.
Market data delivery to AWS, Google Cloud and Azure
Real-time market data into the cloud, at the latency your strategy requires.
Public clouds cannot accept multicast; algoseek is one of the only firms delivering it raw, in production for years. TCP direct for latency-sensitive workloads.
Cloud delivery options
Real-time market data pipeline, capture to delivery
Infrastructure came first. Then the data.
Every exchange publishes an A feed and a B feed. algoseek receives both, twice over, through two independent routes. Mercury arbitrates across all four copies of the same exchange data in real time and writes the consensus.
Mercury is algoseek’s proprietary ticker plant, written in C++ and Assembly with zero external dependencies, running on custom real-time low-latency Linux with specialized low-latency network cards. The same handler serves real-time clients and writes the historical archive. Now in its 3rd generation.
All feeds pass through a single normalization layer: TAQ and bars, with up to 90 quantitative fields per bar at second or minute timeframe. The same normalization serves real-time and historical.
One pipeline, three outputs: real-time, delayed, and the historical archive. The real-time stream your trading consumes is the same data that appears in the historical archive the next morning. Delivered however suits your latency needs: co-location cross-connect, cloud, or internet.
Your servers sit in algoseek’s own cage, with a fiber cross-connect direct from the Mercury ticker plant. Choose standardized small, medium, or large pre-built servers, bring your own server, or a full custom enterprise build, at Equinix NY2 and NY4 in New Jersey and CH1 in Chicago.
Redundancy runs end to end: duplicate capture, two regions, arbitration, automatic failover. A feed, a route, or an entire region can drop without touching what reaches you. No client downtime since 2015 through redundancy.
Exchanges
Raw multicast feeds
SIP (CTA/UTP) · OPRA
CME · CBOT · NYMEX · COMEX
OTC Markets · CBOE Indices · CFE
New Jersey · NY2 / NY4
A feed
Lossless
capture
B feed
Lossless
capture
Regional failover
Chicago · CH1
A feed
Lossless
capture
B feed
Lossless
capture
Mercury ticker plant · 3rd gen
Four-way
arbitration
4 copies in
1 consensus out
Normalization
TAQ · Bars
Up to 90 fields per bar
One source
Real-time
Streaming feed
Delayed
15-minute
Historical archive
Same handler
Since 2007
By latency need
Colocation
Cross-connect
direct from Mercury
Cloud
AWS · Google · Azure
Internet
API · bulk download
Your
applications
Colo · Cloud
On-prem
Market data infrastructure case studies
The hard problems we’ve already solved
US Regulator: Custom OPRA NBBORegulatory
A US regulator needed its own NBBO for equity options, calculated from the standard OPRA feed to the regulator’s own specification, resulting in a daily volume of up to 30 terabytes per day of uncompressed data. algoseek built a Mercury compute grid scaled to manage the fast growing OPRA data set, with the extra compute and bandwidth to handle large volatility spikes. To date it has not missed an SLA.
Hedge Fund: Raw Multicast into AWSBuy Side
A large fund wanted raw multicast equity SIP and multicast OPRA into AWS, which cannot accept multicast. Mercury encapsulates the raw UDP feed into TCP and delivers it over dark fiber into algoseek-controlled AWS accounts, making the data available over AWS PrivateLink. Full regional redundancy is provided by duplicating the infrastructure across Equinix New Jersey and Chicago and us-east-1 and us-east-2, on both the data center and cloud sides, so the fund can pull any channel, A or B feed, from either region.
Prop Trading: Custom Servers to DenmarkBuy Side
A Danish fund’s quant and prop teams run on algoseek-built custom servers, receiving multicast SIP, minute bars, and Nasdaq TotalView over a dedicated connection to Denmark. An enterprise ArdaDB instance serves multiple teams.
Global Bank: Index TWAPsSell Side
A bulge bracket bank needed intraday one minute TWAP equity and futures pricing for index calculations under strict SLAs, for the bank’s internal teams and the external calculation agents who publish the official prices. The bank and its agents pull from algoseek’s GSS file server using simple HTTP, any contract or ticker needed, with thousands of requests running in parallel and no special libraries or other integration work.
Common questions about colocation and market data infrastructure
-
Where are the co-location facilities?
Equinix data centers in New Jersey (NY2/NY4) and Chicago (CH1). algoseek maintains its own cages with dedicated cabinets and infrastructure for maximum security and reliability.
-
Can I bring my own servers?
Yes. Tier 3 is designed for this, including custom GPU builds. algoseek handles hardware installation, networking, power and rack setup, and remote access. Fiber cables run directly from the Mercury ticker plant into your server(s).
-
Which feeds are available?
Normalized feeds from Mercury and raw exchange feeds for: CTA/UTP (equity SIP), OPRA (options), CME/CBOT/NYMEX/COMEX (futures), CFE (Chicago Futures Exchange), CBOE Global Indices, and OTC Markets. Both normalized and raw feeds can be delivered into cloud environments.
-
Can I connect to my cloud environment?
Yes. You can connect over the internet, or for dedicated bandwidth, lower latency and more reliability algoseek recommends your cloud provider’s dark fiber network connection endpoint, which for AWS is AWS Direct Connect. algoseek supports all major clouds including AWS, Google Cloud, Azure and other public clouds. Many clients run a hybrid setup with some components in the data center and some in the cloud.
-
What is the uptime record?
Zero client downtime. Individual exchange feeds can go down, but 4-way arbitration across redundant networks means clients still receive data.
-
Is algoseek a consulting firm?
No. No hourly rates, no standalone engagements. algoseek builds and operates infrastructure as part of solving a market data problem for a client.
-
Which standard feeds are available?
Normalized and raw feeds for CTA/UTP, OPRA, CME/CBOT/NYMEX/COMEX, CFE, CBOE Global Indices, and OTC Markets. Both normalized and raw feeds can be delivered into cloud environments, in co-location, or over the internet.
The rest of the pipeline
Colocation
Your servers in our cage at Equinix, on the same network as Mercury. Three tiers from hosted apps to full custom builds.
Data Customization
Custom pipelines, computed feeds, data onboarding, and cloud migration.
Data Supplier Solutions
Tools and managed services for data vendors to build, sell, and deliver data products.
ArdaDB
Subsecond SQL queries on the full algoseek historical archive. Included with every data package.
We built the real-time infrastructure and custom feeds behind two US regulators. Let’s talk about yours.
You talk to the engineers who built it. No sales layer, no hourly billing. One conversation, and we scope the solution.