I operate and maintain IT infrastructure for organizations where they need me. Usually, I work in a datacenter or a corporate location. I have lived in many places. I have worked with an extensive list of equipment and systems. I record some of my work on this website, for example, my crafts. If you visit this site for the first time, you must press the alt and G keys together to see the keyboard shortcuts.
This website is a thought dictionary. It contains my notes, projects, and ideas. You can scroll down to read the entries, or you can click on a word to read its full page. You can read more about these pages and how the site is made.
January 2025
February 2025
March 2025
April 2025
May 2025
June 2025
July 2025
August 2025
September 2025
October 2025
November 2025
December 2025
February 2026
June 2026
July 2026
August 2026
800-volt DC distribution
Where the Grid Comes In
Aa
-
A Room Built for Air, Asked to Run Water [/ə ˈrum ˈbɪlt fər ˈɛr ˈæst tɪ ˈrən ˈwɔtər/] n - An older air-cooled hall stranded by its own design as rack densities climbed.
-
About these pages [/əˈbaʊt ðiz ˈpeɪʤɪz/] n - How this commonplace is kept: the manner of its pages, its lexicon, and the threads that bind them.
-
Access and role-based controls [/ˈækˌsɛs ənd ˈrole-based kənˈtroʊlz/] n - Managing who can touch which operational system - least privilege, role-based access, and proving it at audit time.
-
AI factory power and cooling [/ˈeɪaɪ ˈfæktəri ˈpaʊər ənd ˈkulɪŋ/] v - Why an AI campus has to be designed as one coupled power-cooling-network system.
-
Alpha Decay and Model Lifecycle [/ˈælfə dɪˈkeɪ ənd ˈmɑdəl ˈlifesykle/] n - Why edges erode, how fast they erode, and the process of rebuilding or retiring a strategy before it destroys capital.
-
Anomaly Detection for Infrastructure Sensors [/əˈnɑməli dɪˈtɛkʃən fər ˌɪnfrəˈstrəkʧər ˈsɛnsərz/] n - Distinguishing real equipment degradation from sensor drift, noise, and benign events using statistical process control and isolation forests.
-
-
ARIMA and Conditional Mean Models [/ˈarima ənd kənˈdɪʃənəl ˈmin ˈmɑdəlz/] n - Forecasting the conditional mean of returns and prices, and where Box-Jenkins methods stop being useful in finance.
-
Autocorrelation Function and Partial Autocorrelation [/ˈautokorrelaʃən ˈfəŋkʃən ənd ˈpɑrʃəl ˈautokorrelaʃən/] n - Using ACF and PACF to read the memory structure of a time series and choose AR and MA orders.
-
Automatic transfer switch [/ˌɔtəˈmætɪk ˈtrænsfər ˈswɪʧ/] n - The switch that moves the building load between utility and generator.
Bb
-
Backtesting and Multiple Comparison Bias [/ˈbakktestɪŋ ənd ˈməltəpəl kəmˈpɛrəsən ˈbaɪəs/] n - Simulating strategy performance, and why a backtest that has been optimized is no longer a backtest.
-
Battery energy storage [/ˈbætəri ˈɛnərʤi ˈstɔrɪʤ/] n - The stored energy a UPS draws on during the ride-through window.
-
Bayesian Updating for Beliefs [/ˈbayesian ˈəpˌdeɪtɪŋ fər bɪˈlifs/] n - Using prior knowledge and new evidence to form posterior beliefs, and why Bayesian methods matter when data is scarce.
-
Busway distribution [/ˈbuswai ˌdɪstrəˈbjuʃən/] n - Overhead busbar that delivers power flexibly along a row of racks.
-
On the Making of Bespoke Devices [/ɔn ðə ˈmeɪkɪŋ əv ˈbespoke dɪˈvaɪsɪz/] n - The full path from a vague operational itch to a tested, documented device in the rack.
Cc
-
Bespoke Devices for the Datacenter [/ˈbespoke dɪˈvaɪsɪz fər ðə ˈdatasenter/] n - Small, purpose-built computers and sensors that fill gaps the vendor catalog does not cover.
-
Calculating SLA credits [/ˈkælkjəˌleɪtɪŋ ˈsla ˈkrɛdɪts/] n - Turning an outage into the right dollar figure - measuring against the contract, validating it, and processing the credit.
-
Clean Tests, a Dirty Connector [/ˈklin ˈtɛsts ə ˈdərti kəˈnɛktər/] n - An optical link looked clean by every instrument until one specific traffic pattern revealed a dirty connector.
-
Co-packaged optics [/ˈko-pakkaged ˈɑptɪks/] n - Putting the optics in the same package as the switch or accelerator to save power.
-
Cointegration and Pairs Trading [/ˈkointegraʃən ənd ˈpɛrz ˈtreɪdɪŋ/] v - Finding long-run equilibrium relationships between non-stationary price series and the decay of classical statistical arbitrage.
-
Cooling Plant Efficiency Prediction [/ˈkulɪŋ ˈplænt ɪˈfɪʃənsi priˈdɪkʃən/] n - Modeling chiller COP, cooling tower performance, and PUE as functions of load and wet-bulb temperature to optimize cooling dispatch.
-
Copulas and Tail Dependence [/ˈkopulas ənd ˈteɪl dɪˈpɛndəns/] n - Modeling dependence beyond correlation, especially when extreme events cluster across assets.
-
Crafts [/ˈkræfts/] n -
-
Cross-Validation in Finance [/ˈkross-validaʃən ɪn ˈfaɪˌnæns/] n - Why k-fold cross-validation fails in time series, and how purged, embargoed, and combinatorial methods fix it.
-
Current transformer [/ˈkɑrənt trænsˈfɔrmər/] n - A sensor that measures current without cutting the conductor.
-
The Fault That Would Not Be Found [/ðə ˈfɔlt ðət ˈwʊd ˈnɑt ˈbi ˈfaʊnd/] n - An intermittent ground-fault alarm that wasn't a fault - it was a schedule.
Dd
-
Data center maintenance tasks [/ˈdætə ˈsɛnər ˈmeɪntənəns ˈtæsks/] n - A running checklist of the maintenance tasks that keep a data center reliable. Most problems do not start with a bang; t...
-
Data Centers on the Moon [/ˈdætə ˈsɛnərz ɔn ðə ˈmun/] n - What it would take to build computing off-world, and why some of the hardest problems are the ones Earth already solved.
-
Data integrity and log retention [/ˈdætə ˌɪnˈtɛgrəti ənd ˈlɔg riˈtɛnʃən/] n - Making sure operational records and system logs are accurate, complete, and kept exactly as long as compliance requires - no more, no less.
-
Direct-to-chip liquid cooling [/ˈdirekt-to-khip ˈlɪkwɪd ˈkulɪŋ/] v - Bringing coolant to the silicon because air can no longer keep up.
-
On the Loop in the Switching Fabric [/ɔn ðə ˈlup ɪn ðə sˈwɪʧɪŋ ˈfæbrɪk/] n - Diagnosing a network switch loop.
-
The Engines of the Hall [/ðə ˈɛnʤənz əv ðə ˈhɔl/] n - The end-to-end chain that carries electrical power from the grid to the server.
Ee
-
Efficient datacenter case studies [/ɪˈfɪʃənt ˈdatasenter ˈkeɪs ˈstədiz/] n - How the largest operators have pushed efficiency, and what smaller sites can borrow from them.
-
EWMA and RiskMetrics [/ˈewma ənd ˈriskmetriks/] n - Exponentially weighted moving averages for covariance and variance, and why simpler models often beat GARCH in fast trading.
-
On Early Detection [/ɔn ˈərli dɪˈtɛkʃən/] n - Spotting failure points before they become outages, using the right mix of sensors, signals, and human judgment.
Ff
-
Factor Zoo and Multi-Factor Models [/ˈfæktər ˈzu ənd ˈmulti-faktor ˈmɑdəlz/] n - Navigating the explosion of equity factors, from Fama-French to q-factors, and avoiding data-mined exposures.
-
Failure Mode and Effect Analysis with Data [/ˈfeɪljər ˈmoʊd ənd ˈifɛkt æˈnælɪsɪs wɪθ ˈdætə/] n - Using historical incident data to quantify failure mode probability and impact, and prioritize predictive monitoring investment.
-
Feature Engineering for Predictive Signals [/ˈfiʧər ˈɛnʤəˈnɪrɪŋ fər prɪˈdɪktɪv ˈsɪgnəlz/] n - Building inputs that are stable, economically meaningful, and robust across market regimes.
-
Field Notes [/ˈfild ˈnoʊts/] n - Small, real problems from the floor and the limited, clever fixes that solved them.
-
-
On Finding the Gaps in One's Tools [/ɔn ˈfaɪndɪŋ ðə ˈgæps ɪn ˈwənz ˈtulz/] n - The continuous-improvement habit - spotting where tools and workflows fall short and turning that into automation, not just complaints.
Gg
-
Google's efficient datacenter design [/ˈguːgle's ɪˈfɪʃənt ˈdatasenter dɪˈzaɪn/] n - How Google designs, measures, and locates datacenters for efficiency.
-
Grounding and bonding [/ˈgraʊndɪŋ ənd ˈbɑndɪŋ/] v - The reference plane that keeps the whole power system safe and stable.
Hh
-
High-Frequency Data and Tick Analytics [/ˈhigh-frequensi ˈdætə ənd ˈtɪk ˌænəˈlɪtɪks/] n - Working with trade-and-quote data, classification, signature plots, and the limits of fine-grained inference.
Ii
-
Information Coefficient and Signal Decay [/ˌɪnˌfɔrˈmeɪʃən ˌkoʊəˈfɪʃənt ənd ˈsɪgnəl dɪˈkeɪ/] n - Measuring the predictive power of a signal, how fast it decays, and what that means for turnover and capacity.
-
IOC operations [/ˈiok ˌɑpərˈeɪʃənz/] n - How an integrated operations center actually runs - the tools, the workflows, and the daily work of keeping them honest.
-
ITSM ticketing systems [/ˈitsm ˈtɪkətɪŋ ˈsɪstəmz/] n - Keeping ServiceNow and Jira tuned so work routes itself, priorities mean something, and the operational record stays clean.
Kk
-
Kelly Criterion and Optimal Bet Sizing [/ˈkɛli kraɪˈtɪriən ənd ˈɑptɪməl ˈbɛt ˈsaɪzɪŋ/] v - Sizing bets to maximize growth, and why full Kelly is usually too aggressive for real trading books.
-
Keyboard shortcuts [/ˈkiˌbɔrd ˈʃɔrtˌkəts/] n - nano-style keystrokes for reading these pages without the mouse.
Ll
-
Linear Factor Models in Trading [/ˈlɪniər ˈfæktər ˈmɑdəlz ɪn ˈtreɪdɪŋ/] v - Decomposing returns into common factors, idiosyncratic bets, and the practical limits of factor exposure.
-
LoRaWAN [/ˈlorawan/] n - A long-range, low-power wireless network for sensors that are too far apart for Wi-Fi.
Mm
-
A Cooling Failure Near the Top-of-Rack Switches [/ə ˈkulɪŋ ˈfeɪljər ˈnɪr ðə ˈtop-of-rakk sˈwɪʧɪz/] n - Managing a cooling failure near top-of-rack switches.
-
Machine Learning Alpha Models [/məˈʃin ˈlərnɪŋ ˈælfə ˈmɑdəlz/] n - Gradient boosting, neural networks, and ensemble methods for return prediction, with a focus on what survives out of sample.
-
Managing monitoring tools [/ˈmænəʤɪŋ ˈmɑnətərɪŋ ˈtulz/] n - Keeping Datadog, Splunk, Prometheus, Grafana, and the rest of the monitoring stack healthy, current, and trusted so the operations center can rely on them.
-
Market Impact and Optimal Execution [/ˈmɑrkɪt ˌɪmˈpækt ənd ˈɑptɪməl ˌɛksəˈkjuʃən/] n - Modeling the cost of trading size, and the algorithms used to minimize the gap between paper return and realized return.
-
Market Microstructure and Execution [/ˈmɑrkɪt ˈmikrostruktʃər ənd ˌɛksəˈkjuʃən/] n - Bid-ask spreads, order flow, liquidity, and why the price you backtested is rarely the price you get.
-
Medium-voltage switchgear [/ˈmedium-voltage ˈswitkhgiːr/] n - The protective switching assembly between the utility and the building transformers.
-
Meta's efficient datacenter design [/ˈmeta's ɪˈfɪʃənt ˈdatasenter dɪˈzaɪn/] n - How Meta uses open hardware, free cooling, and building-level integration to drive efficiency.
-
Modbus [/ˈmodbus/] n - A simple, old, and everywhere industrial protocol for reading registers over serial or Ethernet.
-
Modern Portfolio Optimization [/ˈmɑdərn pɔrtˈfoʊliˌoʊ ɑptɪmɪˈzeɪʃən/] n - From mean-variance to robust and shrinkage methods, and why constraints often beat elegant solutions.
-
Monte Carlo Methods in Quantitative Research [/ˈmɑnti ˈkɑrloʊ ˈmɛθədz ɪn kˈwɑntɪˌteɪtɪv ˈrisərʧ/] n - Using simulation to handle path dependence, non-linearity, and uncertainty in portfolios and strategies.
-
Moving Pictures [/ˈmuvɪŋ ˈpɪkʧərz/] n - Horror films from the classics through the present, rated by someone who takes the genre seriously.
-
On Moving Warm Air Instead of Cold Water [/ɔn ˈmuvɪŋ ˈwɔrm ˈɛr ˌɪnˈstɛd əv ˈkoʊld ˈwɔtər/] n - Reusing server exhaust to pre-heat a small water loop sounded elegant; the controls made it brittle.
-
The Datacenter Beneath the Sea [/ðə ˈdatasenter bɪˈniθ ðə ˈsi/] n - What Project Natick tested by putting a datacenter on the seafloor for two years.
Oo
-
On-site power generation [/ˈon-site ˈpaʊər ˌʤɛnərˈeɪʃən/] n - When the grid can't deliver fast enough, the data center brings its own megawatts.
-
Onboarding and operational docs [/ˈonboardɪŋ ənd ˌɑpərˈeɪʃənəl ˈdɑks/] n - The runbooks, guides, and training that let any analyst run the operations center without the tribal knowledge living in one person's head.
-
Operational metrics and reporting [/ˌɑpərˈeɪʃənəl ˈmɛtrɪks ənd rɪˈpɔrtɪŋ/] v - Turning the operations center's raw activity into metrics, incident summaries, and reports leadership can actually act on.
-
Optical interconnects in AI clusters [/ˈɑptɪkəl ˈinterkonnekts ɪn ˈeɪaɪ ˈkləstərz/] n - How light-based links shape the topology, power, and cooling of large GPU fabrics.
Pp
-
Power distribution unit [/ˈpaʊər ˌdɪstrəˈbjuʃən ˈjunɪt/] n - The floor-level transformer and breaker panel that feeds the data hall.
-
Power Load Forecasting in Data Centers [/ˈpaʊər ˈloʊd ˈfɔrˌkæstɪŋ ɪn ˈdætə ˈsɛnərz/] n - Forecasting aggregate IT power draw from workload signals, historical patterns, and weather data across multiple horizons.
-
Power over Ethernet [/ˈpaʊər ˈoʊvər ˈiθərˌnɛt/] n - Delivering both power and data over a single Ethernet cable.
-
Power usage effectiveness [/ˈpaʊər ˈjusɪʤ ˈifɛktɪvnəs/] n - PUE - the ratio everyone quotes to describe data center efficiency.
-
Predictive Analysis for Data Center Operations [/prɪˈdɪktɪv æˈnælɪsɪs fər ˈdætə ˈsɛnər ˌɑpərˈeɪʃənz/] n - A hub for applying forecasting mathematics to cooling, power, and maintenance in a live data center environment.
-
Predictive Generator Maintenance [/prɪˈdɪktɪv ˈʤɛnərˌeɪtər ˈmeɪntənəns/] n - Oil analysis, vibration spectra, and run-hour models for assessing standby generator health before a real outage tests it.
-
Predictive Maintenance for CRAC and CRAH Units [/prɪˈdɪktɪv ˈmeɪntənəns fər ˈkrak ənd ˈkrah ˈjunɪts/] n - Detecting compressor degradation, fan bearing wear, and filter fouling before they cause a cooling failure using time series anomaly detection.
-
Principal Components and Eigenportfolios [/ˈprɪnsəpəl kəmˈpoʊnənts ənd ˈeigenportfolios/] n - Reducing dimensionality with PCA, building eigenportfolios, and spotting the market factor in covariance structure.
-
Problems We Could Only Work Around [/ˈprɑbləmz wi ˈkʊd ˈoʊnli ˈwərk ˈəraʊnd/] n - Big constraints that no clever fix could remove - only manage.
-
Quantitative Predictive Analysis [/kˈwɑntɪˌteɪtɪv prɪˈdɪktɪv æˈnælɪsɪs/] n - A working map of the mathematics, models, and habits used to predict prices, risk, and execution outcomes in institutional trading.
Rr
-
A Small Alarm for the UPS [/ə ˈsmɔl əˈlɑrm fər ðə ˈəps/] n - A small custom device that turns a Raspberry Pi into a physical alarm panel for UPS events. It uses the UPS network mana...
-
On a Server That Would Not Wake [/ɔn ə ˈsərvər ðət ˈwʊd ˈnɑt ˈweɪk/] n - Resolving a server boot failure.
-
On Changing a Supply While the Load Ran [/ɔn ˈʧeɪnʤɪŋ ə səˈplaɪ ˈwaɪl ðə ˈloʊd ˈræn/] n - Replacing a faulty power supply unit under load.
-
On Running the Root-Cause Inquiry [/ɔn ˈrənɪŋ ðə ˈruːt-kause ˌɪnkˈwaɪˌri/] n - Owning the root-cause-analysis lifecycle so incidents turn into durable lessons - on time, with quality, and with the record intact.
-
On the Battery Behind the RAID Controller [/ɔn ðə ˈbætəri bɪˈhaɪnd ðə ˈreɪd kənˈtroʊlər/] n - Repairing a RAID controller battery failure.
-
On the Changing of a Motherboard [/ɔn ðə ˈʧeɪnʤɪŋ əv ə ˈmotherboard/] n - Replacing a failed server motherboard.
-
Rack power distribution unit [/ˈræk ˈpaʊər ˌdɪstrəˈbjuʃən ˈjunɪt/] n - The power strip inside the cabinet that feeds individual servers.
-
Random Matrix Theory [/ˈrændəm ˈmeɪtrɪks ˈθɪri/] n - Separating signal from noise in large covariance matrices using the Marchenko-Pastur distribution.
-
Raspberry Pi [/ˈræzˌbɛri ˈpaɪ/] n - A small, ARM-based Linux computer that is cheap enough to deploy by the dozen.
-
Raspberry Pi in the datacenter [/ˈræzˌbɛri ˈpaɪ ɪn ðə ˈdatasenter/] n - Using cheap, Linux-capable single-board computers as monitoring, bridging, and control nodes in a data center.
-
Redundancy topologies [/rɪˈdəndənsi ˈtopologies/] n - N, N+1, 2N and what "concurrently maintainable" really means.
-
Regime Switching and Hidden Markov Models [/rəˈʒim sˈwɪʧɪŋ ənd ˈhɪdən ˈmarkov ˈmɑdəlz/] n - Detecting changes in market behavior, and why knowing the regime can matter more than knowing the forecast.
-
Remaining Useful Life Estimation [/rɪˈmeɪnɪŋ ˈjusfəl ˈlaɪf ˌɛstəˈmeɪʃən/] n - Applying Weibull distributions, degradation models, and survival analysis to estimate when data center equipment will need replacement.
-
Remote power panel [/rɪˈmoʊt ˈpaʊər ˈpænəl/] n - A breaker panel that extends a PDU's circuits closer to the racks.
-
Risk Parity and Equal Risk Contribution [/ˈrɪsk ˈpɛrəti ənd ˈikwəl ˈrɪsk ˌkɑntrəˈbjuʃən/] n - Allocating capital by risk contribution rather than dollar weight, and why this avoids concentration in equities.
-
The Right Sensor in the Wrong Air [/ðə ˈraɪt ˈsɛnsər ɪn ðə ˈrɔŋ ˈɛr/] n - A temperature sensor that looked wrong was actually right; the problem was the rack it was sitting in.
Ss
-
Seasonal ARIMA and Exponential Smoothing [/ˈsizənəl ˈarima ənd ˌɛkspoʊˈnɛnʃəl sˈmuðɪŋ/] v - Modeling seasonality in prices, volumes, and volatility with SARIMA and Holt-Winters methods.
-
Silicon photonics [/ˈsɪləkən ˈfotoniks/] n - Moving data with light on a chip, and why it has become a data-center power story.
-
Single-board computer [/ˈsingle-board kəmˈpjutər/] n - The category of full computers built on a single circuit board, from hobby boards to industrial embedded PCs.
-
SNMP [/ˈsnmp/] n - The Simple Network Management Protocol, still the default way to ask network and server gear for its status.
-
State Space Models and the Kalman Filter [/ˈsteɪt ˈspeɪs ˈmɑdəlz ənd ðə ˈkælmən ˈfɪltər/] n - Filtering and smoothing latent states, from dynamic betas to hidden regimes in financial time series.
-
Static transfer switch [/ˈstætɪk ˈtrænsfər ˈswɪʧ/] n - A millisecond solid-state switch that protects a single dual-corded load.
-
Stationarity and Unit Root Tests [/ˈstationariti ənd ˈjunɪt ˈrut ˈtɛsts/] n - Testing whether a series is stable over time, and why non-stationarity ruins forecasts and inference.
-
Stochastic Calculus in Finance [/stoʊˈkæstɪk ˈkælkjələs ɪn ˈfaɪˌnæns/] n - Ito's lemma, stochastic differential equations, and the boundary between derivative pricing and predictive trading.
-
Stress Testing and Scenario Design [/ˈstrɛs ˈtɛstɪŋ ənd sɪˈnɛrioʊ dɪˈzaɪn/] n - Building adverse scenarios, measuring portfolio damage, and going beyond correlation matrices when markets break.
-
The Standby Engine [/ðə ˈstændˈbaɪ ˈɪnʤən/] n - The engine-driven source that carries the site through a prolonged outage.
Tt
-
On the Camera That Drifted on the Hot Aisle [/ɔn ðə ˈkæmərə ðət ˈdrɪftɪd ɔn ðə ˈhɑt ˈaɪəl/] n - A monitoring rig slowly started lying; the fix was a reference target, not a recalibration trip.
-
On the Packet That Never Arrived [/ɔn ðə ˈpækɪt ðət ˈnɛvər ˈəraɪvd/] n - Tracing a dropped network packet issue.
-
On the Router That Slowly Forgot [/ɔn ðə ˈrutər ðət sˈloʊli fərˈgɑt/] n - Troubleshooting a core router memory leak.
-
Temperature sensor [/ˈtɛmpərəʧər ˈsɛnsər/] n - How temperature sensors work, what they measure, and how to choose one for a data center.
-
The Black-Scholes Framework [/ðə ˈblakk-skholes ˈfreɪmˌwərk/] n - The benchmark option pricing model, its assumptions, and why experienced traders treat it as a quoting convention rather than truth.
-
The Breaker Never Moved [/ðə ˈbreɪkər ˈnɛvər ˈmuvd/] n - A breaker alarm kept firing; the breaker was fine, but the shunt trip wire was not.
-
The Current That Comes in Threes [/ðə ˈkɑrənt ðət ˈkəmz ɪn ˈθriz/] n - Why data centers distribute power as three phases rather than one.
-
The Generator Was Not Overheating [/ðə ˈʤɛnərˌeɪtər wɑz ˈnɑt ˈoʊvərˌhitɪŋ/] v - A monthly test kept aborting; the culprit was a sensor and a setpoint, not the engine.
-
The Greeks in Practice [/ðə ˈgriks ɪn ˈpræktɪs/] n - Delta, gamma, vega, theta, and the higher-order Greeks used to manage option books.
-
The Harmonics We Had to Live With [/ðə hɑrˈmɑnɪks wi ˈhæd tɪ ˈlɪv wɪθ/] n - A power-quality problem we could contain at the edges but never truly remove.
-
The New Buildout [/ðə ˈnu ˈbuildout/] n - The handful of big shifts reshaping how data centers are powered, cooled, and connected.
-
The Quantitative Research Workflow [/ðə kˈwɑntɪˌteɪtɪv ˈrisərʧ ˈwərkˌfloʊ/] n - The end-to-end process from idea generation to production trading, with checkpoints that keep research honest.
-
Thermal Forecasting in Data Centers [/ˈθərməl ˈfɔrˌkæstɪŋ ɪn ˈdætə ˈsɛnərz/] n - Using sensor data, workload signals, and time series models to predict rack inlet temperatures and thermal margin before a hotspot forms.
-
Things That Did Not End Up Working Out [/ˈθɪŋz ðət ˈdɪd ˈnɑt ˈɛnd ˈəp ˈwərkɪŋ ˈaʊt/] n - Promising ideas, small experiments, and pilot projects that looked good on paper but failed somewhere in reality.
-
Time Series Decomposition and Forecasting [/ˈtaɪm ˈsɪriz ˌdikəmpəˈzɪʃən ənd ˈfɔrˌkæstɪŋ/] v - Separating a price or return series into components, and why a forecast is only as good as the decomposition that produced it.
-
Transaction Costs and Slippage Models [/trænˈzækʃən ˈkɔsts ənd sˈlɪpɪʤ ˈmɑdəlz/] n - Building explicit, implicit, and opportunity-cost models so backtests stop lying about profitability.
-
Tuning alert thresholds [/ˈtunɪŋ əˈlərt θˈrɛˌʃoʊldz/] n - Turning raw telemetry into pages worth waking up for - thresholds, workflows, and automation rules for GPU, network, and facility events.
Uu
-
On Firmware, A Thousand Servers at a Time [/ɔn ˈfirmware ə ˈθaʊzənd ˈsərvərz æt ə ˈtaɪm/] n - Updating server firmware at scale.
-
Uninterruptible power supply [/ənɪnərˈəptəbəl ˈpaʊər səˈplaɪ/] n - The device that carries the IT load through the instant utility power fails.
-
UPS and Battery State Prediction [/ˈəps ənd ˈbætəri ˈsteɪt priˈdɪkʃən/] n - Estimating battery state of charge, available runtime, and capacity fade using electrochemical models and recursive filtering.
-
Uptime Institute tier classification [/ˈuptime ˈɪnstɪˌtut ˈtir ˌklæsəfəˈkeɪʃən/] n - The Tier I–IV scale that grades a facility's redundancy and maintainability.
-
Where the Grid Comes In [/ˈwɛr ðə ˈgrɪd ˈkəmz ɪn/] n - Where grid power physically enters the facility.
Vv
-
A Visit to Lake Moab [/ə ˈvɪzɪt tɪ ˈleɪk ˈmoʊæb/] n - I love the cooling breeze of the lake as I sit at the banks and make my first drawing in a long time, which I did with a...
-
Value at Risk and Expected Shortfall [/ˈvælju æt ˈrɪsk ənd ɪkˈspɛktɪd ˈʃɔrtˌfɔl/] n - Measuring tail risk with quantiles and conditional averages, and the limits of risk metrics in fast markets.
-
Volatility Modeling with GARCH [/ˌvɑləˈtɪləti ˈmɑdəlɪŋ wɪθ ˈgarkh/] n - Forecasting conditional variance, clustering, and the practical limits of ARCH-family models in trading.
Ww
-
The Watts Hidden in Volt-Amps [/ðə ˈwɑts ˈhɪdən ɪn ˈvolt-amps/] n - A metering quirk was quietly stranding capacity; the fix was a poller, not a purchase.
-
When the Fence Would Give No More [/ˈwɪn ðə ˈfɛns ˈwʊd ˈgɪv ˈnoʊ ˈmɔr/] n - Capped at the fence, growth had to come from the demand side instead.