1 / 18

Research Focus

Research Focus. STM Systems. Dependability of STM. Performance Modelling of STM. Roberto Palmieri (PhD Student) “ Sapienza ” University of Rome Italy. Pierangelo Di Sanzo (PhD Student) “ Sapienza ” University of Rome Italy. Dependability Issues of STMs.

thea
Télécharger la présentation

Research Focus

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Research Focus STM Systems Dependability of STM Performance Modelling of STM Roberto Palmieri (PhD Student) “Sapienza” University of Rome Italy Pierangelo Di Sanzo (PhD Student) “Sapienza” University of Rome Italy EURO-TM | 1st Plenary Meeting | Paris 2011

  2. Dependability Issues of STMs • In STM systems, manipulated data and related statements are not natively logged on stable storage • Data-audit loss in case of crashes • Periodic logging (check-pointing) could be employed, however with time-granularity not adequate to the proper operation grain REPLICATION EURO-TM | 1st Plenary Meeting | Paris 2011

  3. Active Replication + OAB • Active Replication (AR) is a common replication scheme • In AR each replica keeps the entire shared data-set and executes the same transactions in the same order • Optimistic Atomic Broadcast protocol (OAB) is a group communication system involved (example of software architecture replica) EURO-TM | 1st Plenary Meeting | Paris 2011

  4. Overlapping Processing • High delays (1/2msec) is in conflict with the growth of available resources in each nodes and with small transaction execution time (typical of STMs) • Solution could be optimistically overlap local processing with replica coordination's: WITHOUT OVERLAP (AB) WITH OVERLAP (OAB) PROCESS m COORDINATION PHASES PROCESS m COORDINATION PHASES to-broadcast (m) to-delivery (m) to-broadcast (m) to-delivery (m) opt-delivery (m) EURO-TM | 1st Plenary Meeting | Paris 2011

  5. STR FRAMEWORK • Speculative Transactional Replication Framework • Until the arrive of TO-Delivery, set of optimistically delivered (unordered) transactions could be processed in speculative way • Key Idea -> STR Framework on-line identifies all and only transaction serialization orders that would cause optimistically executed transactions to exhibit distinct outcomes. • Properties: Consistency, Non-redundancy, Completeness • Graph based Concurrency Control: • Speculative Polygraph EURO-TM | 1st Plenary Meeting | Paris 2011

  6. STR FRAMEWORK EURO-TM | 1st Plenary Meeting | Paris 2011

  7. AGGRO • AGGRessivelyOptimistic replication protocol • Tailored for network with spontaneous order (Opt-Delivery order matches TO-Delivery order) • Optimistic processing aimed to follow the optimistic delivery order • The key idea-> Uncommitted data item versions are aggressively made visible to other transactions independently of whether the creating transactions will be eventually committed • Transactions abort/retry materializing a history compliant with optimistic delivery order EURO-TM | 1st Plenary Meeting | Paris 2011

  8. AGGRO EURO-TM | 1st Plenary Meeting | Paris 2011

  9. OSARE • Opportunistic Speculation in Actively REplicated Transactional Systems • The snapshot miss event is used to opportunistic exploring additional serialization orders • The activation of new transaction instance involves any transaction originally serialized after that (like a wave on a different Speculative serialization order) EURO-TM | 1st Plenary Meeting | Paris 2011

  10. Transactional systems performance models: why? Applications: • system performance analysis • scalability analysis • identifying performance bottlenecks • …. • performance comparison among different concurrency control algorithms (CCAs) • what-if analysis • what would happen if I add one more thread • what would happen if I change CCA • …. • evaluation of new CCAs • …. EURO-TM 1st Plenary Meeting Paris 2011

  11. Building transactional systems performance model… • In transactional systems performance is affected by two factors: • queuing and processing delay in accessing hardware resources (CPU, shared bus, …) • data conflict in accessing shared data items mutual dependence data conflict abort rate transaction response time hardware resources usage EURO-TM 1st Plenary Meeting Paris 2011

  12. Example (with optimistic concurrency control): Suppose that at some point data conflict increases …data conflict increases… data conflict (data items utilization increases) (abort probability increases) …transaction response time increases… abort rate transaction response time …abort rate increases… hardware resources usage (queuing time increases) (transactions re-run many times) …hardware resources usage increases… EURO-TM 1st Plenary Meeting Paris 2011

  13. hardware resources model data conflict model Performance modelling approach two separated modelling layers: Iterative approach to estimate system performance indicators 1) hardware resources model 2) data conflict model different approaches queue network model EURO-TM 1st Plenary Meeting Paris 2011

  14. Performance modelling approach: data conflict model thread model tbackoff back-off non-transactional code block transaction tntcb transaction model code block read/ write code block code block read/ write code block begin . . . . commit tbegin ttcb tcommit twrite/tread t_: expected completion time (input into the model) EURO-TM 1st Plenary Meeting Paris 2011

  15. Threads execution is modeled via a Continuous Time Markov Chain (CTMC) State (i, j): i is the number of threads which are running transactions and j the number of threads in back-off. CTMC transition rates depend on: λ = 1/ tntcb : transaction arrivals rate μi:(=1/ rt,i) txs run service rate in state (i,*) pc,i: probability to successfully commit in state (i,*) State transition diagram of CTMC with k = 3 - μi and pc,I are input from the transaction modelling layer EURO-TM 1st Plenary Meeting Paris 2011

  16. For each state (i,*) of CTMC: initial settings: iterations: concurrency control algorithm model CCA modellingequations CCA modellingparameters update Iterations end when the difference between two consecutive values of pc is < є EURO-TM 1st Plenary Meeting Paris 2011

  17. Model Validation: Model vs. discrete event simulation CCA: lazy locking + read validation (TL2) • testing workload: • three transactional classes • - uniform data accesses EURO-TM 1st Plenary Meeting Paris 2011

  18. Thank you EURO-TM 1st Plenary Meeting Paris 2011

More Related