Experiment-driven System Management Shivnath Babu Duke University Joint work with Songyun Duan, Herodotos Herodotou, and Vamsidhar Thummala Managing DBs in Small to Medium Business Enterprises (SMBs) • Peter is a system admin in an SMB – Manages the database (DB) – SMB cannot afford a DBA • Suppose Peter has to tune a poorly-performing DB Database (DB) – Design advisor may not help – Maybe the problem is with DB configuration parameters Tuning DB Configuration Parameters • Parameters that control – – – – Memory distribution I/O optimization Parallelism Optimizer’s cost model • Number of parameters ~ 100 – 15-25 critical params depending on OLAP Vs. OLTP • Few holistic parameter tuning tools available – Peter may have to resort to 1000+ page tuning manuals or rules of thumb from experts – Can be a frustrating experience Response Surfaces • TPC-H 4 GB DB size, 1 GB memory, Query 18 2-dim Projection of a 11-dim Surface DBA’s Approach to Parameter Tuning • DBAs run experiments – Here, an experiment is a run of the DB workload with a specific parameter configuration – Common strategy: vary one DB parameter at a time Experiment-driven Management Result Mgmt. task Are more experiments needed? Process output to extract information Yes Plan next set of experiments Conduct experiments on workbench Goal: Automate this process Roadmap • Use cases of experiment-driven mgmt. – Query tuning, benchmarking, Hadoop, testing, … • iTuned: Tool for DB conf parameter tuning – End-to-end application of experiment-driven mgmt. • .eX: Language and run-time system that brings experiment-driven mgmt. to users & tuning tools What is an Experiment? • Depends on the management task – Pay some extra cost, get new information in return – Even for a specific management task, there can be spectrum of possible experiments Uses of Experiment-driven Mgmt. • DB conf parameter tuning Uses of Experiment-driven Mgmt. • DB conf parameter tuning • MapReduce job tuning in Hadoop Uses of Experiment-driven Mgmt. • DB conf parameter tuning • MapReduce job tuning in Hadoop • Server benchmarking – Capacity planning – Cost/perf modeling Uses of Experiment-driven Mgmt. • Tuning “problem queries” <2473, 7496> 7496> <2473, <380459, 229739> Sort <100, 187> Hash Aggregate <100, 187> Nested Loop Join <100, 436> Hash Join Index Scan (orders) Hash <65, 309> Hash Join <65, 309> Sequential Scan (lineitem) <1629, 1615> Sequential Scan (supplier) <Estimated, Actual> Cardinality <1, 0.6> Hash <1, 1> Sequential Scan (nation) <1, 1> Uses of Experiment-driven Mgmt. • Tuning “problem queries” Sort Hash Aggregate Hash Join Sequential Scan (orders) Hash Hash Join Sequential Scan (lineitem) Hash Hash Join Sequential Scan (supplier) Hash Sequential Scan (nation) Uses of Experiment-driven Mgmt. • DB conf parameter tuning • MapReduce job tuning in Hadoop • Server benchmarking – Capacity planning – Cost/perf modeling • Tuning “problem queries” • Troubleshooting • Testing • Canary in the server farm (James Hamilton, Amazon) • … Roadmap • Use cases of experiment-driven mgmt. – Query tuning, benchmarking, Hadoop, testing, … • iTuned: Tool for DB conf parameter tuning – End-to-end application of experiment-driven mgmt. • .eX: Language and run-time system that brings experiment-driven mgmt. to users & tuning tools Problem Abstraction • Unknown response surface: y = F(X) – X = Parameters x1, x2, …, xm • Each experiment gives a <Xi,yi> sample – Set DB to conf Xi – Run workload that needs tuning – Measure performance yi at Xi • Goal: Find high performance setting with low total cost of running experiments 6 8 Example 4 y Utility(X) 0 2 Where to do the next experiment? 4 6 8 10 x1 • Goal: Compute the potential utility of candidate experiments 12 iTuned’s Adaptive Sampling Algorithm for Experiment Planning // Phase I: Bootstrapping – Conduct some initial experiments // Phase II: Sequential Sampling – Loop: Until stopping condition is reached 1. Identify candidate experiments to do next 2. Based on current samples, estimate the utility of each candidate experiment 3. Conduct the next experiment at the candidate with highest utility Utility of an Experiment • Let <X1,y1>--<Xn,yn> be the samples from n experiments done so far • Let <X*,y*> be the best setting so far (i.e., y* = mini yi) – wlg assuming minimization • U(X), Utility of experiment at X is // y = F(X) – y* - y if y* > y – 0 otherwise • However, U(X) poses a chicken-and-egg problem – y will be known only after experiment is run at X • Goal: Compute expected utility EU(X) Expected Utility of an Experiment • Suppose we have the probability density function of y (y is the perf at X) – Prob(y = v | <Xi,yi> for i=1,…,n) • Then, EU(X) = v=+1 sv=-1 U(X) Prob(y = v) dv v=y* EU(X) = sv=-1 (y* - v) Prob(y = v) dv • Goal: Compute Prob(y = v | <Xi,yi> for i=1,…,n) Model: Gaussian Process Representation (GRS) of a Response Surface • GRS models the response surface as: y(X) = g(X) + Z(X) (+ (X) for measurement error) – E.g., g(X) = x1 – 2x2 + 0.1x12 (Learned using common techniques) – Z: Gaussian Process to capture regression residual Primer on Gaussian Process • Univariate Gaussian distribution – G = N(,) – Described by mean , variance • Multivariate Gaussian distribution – [G1, G2, …, Gn] – Described by mean vector and covariance matrix • Gaussian Process – Generalizes multivariate Gaussian to arbitrary number of dimensions – Described by mean and covariance functions Model: Gaussian Process Representation (GRS) of a Response Surface • GRS captures the response surface as: y(X) = g(X) + Z(X) (+ (X) for measurement error) • If Z is a Gaussian process, then: [Z(X1),…,Z(Xn),Z(X)] is multivariate Gaussian Z(X) | Z(X1),…,Z(Xn) is a univariate Gaussian y(X) is a univariate Gaussian Parameters of the GRS Model • [Z(X1),…,Z(Xn)] is multivariate Gaussian – Z(Xi) has zero mean – Covariance(Z(Xi),Z(Xj)) / exp(k –k |xik – xjk|k) • Residuals at nearby points have higher correlation • k, ½k learned from <X1,y1>--<Xn,yn> Use of the GRS Model • Recall our goals to compute v=y* – EU(X) = sv=-1 (y* - v) Prob(y = v) dv – Prob(y = v | <Xi,yi> for i=1,…,n) • Lemma: Using the GRS, we can compute the mean (X) and variance 2(X) of the Gaussian y(X) • Theorem: EU(X) has a closed form that is a product of: – Term that depends on (y* - (X)) – Term that depends on (X) • It follows that settings X with high EU are either: – Close to known good settings (for exploitation) – In highly uncertain regions (for exploration) Example • Settings X with high EU are either: – Close to known good settings (high y*-(X)) – In highly uncertain regions (high (X)) 6 8 Unknown actual surface (X) 4(X) 2 4 y y* 0 EU(X) 4 6 8 x1 10 12 Where to Conduct Experiments? Clients Clients Clients Test Platform Middle Tier DBMS Production Platform Test Data DBMS Data Write Ahead Log (WAL) shipping Standby Platform DBMS Data iTuned’s Solution • Exploit underutilized resources with minimal impact on production workload • DBA/User designates resources where experiments can be run – E.g., production/standby/test • DBA/User specifies policies that dictate when experiments can be run – Separate regular use (home) from experiments (garage) – Example: If CPU, mem, & disk utilization < 10% for past 15 mins, then resource can be used for experiments One Implementation of Home/Garage Clients Clients Clients Standby Machine Middle Tier Home Production Platform Apply WAL DBMS WAL shipping Data iTuned Apply DBMS WAL HomeGarage Workbench for experiments DBMS DBMS Data Interface Engine Experiment Planner & Scheduler Copy on Write Overheads are Low Operation in API Time (seconds) Description Create Container 610 Create a new garage (one time process) Clone Container 17 Clone a garage from already existing one Boot Container 19 Boot garage from halt state Halt Container 2 Stop garage and release resources Reboot Container 2 Reboot the garage Snapshot-R DB (5GB, 20GB) 7, 11 Create read-only snapshot of the database Snapshot-RW DB (5GB, 20GB) 29, 62 Create read-write snapshot of database Empirical Evaluation (1) • Cluster of machines with 2GHz processors and 3GB memory • Two database systems: PostgreSQL & MySQL • Various workloads – OLAP: Mixes of heavy-weight TPC-H queries • Varying #queries, #query_types, and MPL • Scale factors 1 and 10 – OLTP: TPC-W and RUBiS • Tuning of up to 30 configuration parameters Empirical Evaluation (2) • Techniques compared Default parameter settings shipped (D) Manual rule-based tuning (M) Smart Hill Climbing (S): State-of-the-art technique Brute-Force search (B): Run many experiments to find approximation to optimal setting – iTuned (I) – – – – • Evaluation metrics – Quality: workload running time after tuning – Efficiency: time needed for tuning Comparison of Tuning Quality iTuned’s Scalability Features (1) • • • • • Identify important parameters quickly Run experiments in parallel Stop low-utility experiments early Compress the workload Work in progress: – Apply database-specific knowledge – Incremental tuning – Interactive tuning iTuned’s Scalability Features (2) • Identify important parameters quickly – Using sensitivity analysis with a few experiments #Parameters = 9, #Experiments = 10 iTuned’s Scalability Features (3) Roadmap • Use cases of experiment-driven mgmt. – Query tuning, benchmarking, Hadoop, testing, … • iTuned: Tool for DB conf parameter tuning – End-to-end application of experiment-driven mgmt. • .eX: Language and run-time system that brings experiment-driven mgmt. to users & tuning tools Back of the Envelope Calculation • DBAs cost $300/day; Consultants cost $100/hr • 1 Day of experiments gives a wealth of info. – TPC-H, TPC-W, RUBiS workloads; 10-30 conf. params • Cost of running these experiments for 1 day on Amazon Web Serv. – – – – Server: $10/day Storage: $0.4/day I/O: $5/day TOTAL: $15/day .eX: Power of Experiments to the People eXL script Language processor .eX Run-time engine Resources • Users & tools express needs as scripts in eXL (eXperiment Language) • .eX engine plans and conducts experiments on designated resources • Intuitive visualization of results Current Focus of .eX • Parts of an eXL script I1 I2 O1 … … 1. Query: (approx.) response surface mapping, search 2. Expt. setup & monitoring 3. Constraints & optimization: resources, cost, time Result Automatically generate the experiment-driven workflow Process output to extract information Are more experiments needed? Yes Plan next set of experiments Conduct experiments on workbench Summary • Automated expt-driven mgmt: The time has come – Need, infrastructure, & promise are all there • We have built many tools around this paradigm – http://www.cs.duke.edu/~shivnath/dotex.html • Poses interesting questions and challenges – Make it easy for users/admins to do expts – Make experiments first-class citizens in systems