Scaleability scale up and scale out l.jpg
1 / 23

Scaleability Scale Up and Scale Out.


82 views
Uploaded on:
Category: General / Misc
Description
Scaleability Scale Up and Scale Out SMP Super Server Departmental Server Personal System Grow Up with SMP 4xP6 is now standard Grow Out with Cluster Cluster has inexpensive parts Cluster of PCs Thesis Many little beat few big 3 1 MM 10 nano-second ram 10 microsecond ram
Transcripts
Slide 1

Scaleability Scale Up and Scale Out SMP Super Server Departmental Server Personal System Grow Up with SMP 4xP6 is presently standard Grow Out with Cluster has modest parts Cluster of PCs

Slide 2

Thesis Many minimal beat couple of huge 3 1 MM 10 nano-second smash 10 microsecond ram 10 millisecond circle 10 second tape document $1 million $10 K $100 K Pico Processor Nano Micro 10 pico-second slam 1 MB Mini Mainframe 10 0 MB 1 0 GB 1 TB 1 00 TB 1.8" 2.5" 3.5" 5.25" 1 M SPECmarks, 1TFLOP 10 6 timekeepers to mass ram Event-skyline on chip VM resurrected Multi-project reserve, On-Chip SMP 9" 14" Smoking, furry golf ball How to associate the numerous little parts? How to program the numerous little parts? Adaptation to non-critical failure & Management? Dim - Microsoft @ LANL 12/17/98

Slide 3

4 B PC’s (1 Bips, .1GB measure, 10 GB plate 1 Gbps Net, B=G) The Bricks of Cyberspace Cost 1,000 $ Come with NT DBMS High speed Net System administration GUI/OOUI Tools Compatible with other people CyberBricks Gray - Microsoft @ LANL 12/17/98

Slide 4

Kilo Mega Giga Tera Peta Exa Zetta Yotta Computers therapist to a point Disks 100x in 10 years 2 TB 3.5” commute Shrink to 1” is 200GB Disk is super PC! This is now valid for printers and “terminals” Gray - Microsoft @ LANL 12/17/98

Slide 5

Microsoft.com: ~150x4 hubs: a group Building 11 Staging Servers (7) Ave CFG: 4xP6, Internal WWW Ave CFG: 4xP5, European Data Center premium.microsoft.com IDC Staging Servers 512 RAM, www.microsoft.com 30 GB HD (1) MOSWest (3) Ave CFG: 4xP6, Ave CFG: 4xP6, 512 RAM, FTP Servers 512 RAM, SQLNet 30 GB HD Ave CFG: 4xP5, SQL SERVERS 50 GB HD Feeder LAN 512 RAM, SQL Consolidators (2) Router Download 30 GB HD DMZ Staging Servers Ave CFG: Replication 4xP6, Ave CFG: 4xP6, 512 RAM, FTP Router 1 GB RAM, Live SQL Servers 160 GB HD Download Server 160 GB HD SQL Reporting Ave Cost: $83K Ave CFG: 4xP6, (1) MOSWest Switched Ave CFG: FY98 Fcst: 4xP6, 2 512 RAM, Live SQL Server Ave CFG: Admin LAN 4xP6, Ethernet 512 RAM, 160 GB HD 512 RAM, 160 GB HD Ave Cost: $83K 50 GB HD FY98 Fcst: 12 search.microsoft.com msid.msn.com (1) msid.msn.com register.microsoft.com www.microsoft.com (1) (1) www.microsoft.com (2) (4) Ave CFG: 4xP6, Router (4) 512 RAM, search.microsoft.com Ave CFG: 4xP6, 30 GB HD Japan Data Center (3) 512 RAM, SQL SERVERS www.microsoft.com 50 GB HD Ave CFG: premium.microsoft.com 4xP6, (2) (3) 512 RAM, Ave CFG: 4xP6, (1) 30 GB HD home.microsoft.com 512 RAM, Ave CFG: 4xP6, home.microsoft.com Ave CFG: 4xP6, Ave Cost: $28K 160 GB HD FDDI Ring 512 RAM, (3) 512 RAM, FY98 Fcst: (4) 7 (MIS2) 50 GB HD premium.microsoft.com 30 GB HD Ave CFG: 4xP6 (2) msid.msn.com 512 RAM Ave CFG: 4xP6, activex.microsoft.com 28 GB HD 512 RAM, (1) (2) FDDI Ring Ave CFG: 4xP6, 30 GB HD Switched (MIS1) 512 RAM, Ave CFG: 4xP6, Ethernet 30 GB HD 256 RAM, 30 GB HD FTP Ave Cost: $25K cdm.microsoft.com Download Server Ave CFG: FY98 Fcst: 4xP5, 2 (1) 256 RAM, Router (1) HTTP search.microsoft.com 12 GB HD Download Servers (2) (2) Router Internet msid.msn.com Router (1) 2 Primary 2 Router Gigaswitch OC3 Ethernet premium.microsoft.com (100Mb/Sec Each) Internet (100 Mb/Sec Each) Router (1) www.microsoft.com Router (3) Secondary Gigaswitch 13 Router DS3 Router FTP.microsoft.com (45 Mb/Sec Each) (3) FDDI Ring Ave CFG: 4xP5, home.microsoft.com (MIS3) www.microsoft.com msid.msn.com 512 RAM, (2) 30 GB HD (5) (1) Internet register.microsoft.com Ave CFG: 4xP5, FDDI Ring (2) 256 RAM, (MIS4) 20 GB HD register.microsoft.com home.microsoft.com support.microsoft.com (1) (5) register.msn.com (2) (2) Ave CFG: 4xP6, support.microsoft.com 512 RAM, search.microsoft.com (1) 30 GB HD (3) Gray - Microsoft @ LANL 12/17/98

Slide 6

HotMail: ~400 Computers Crowd Gray - Microsoft @ LANL 12/17/98

Slide 7

DB Clusters (swarms) 16-hub Cluster 64 cpus 2 TB of plate Decision support 45-hub Cluster 140 cpus 14 GB DRAM 4 TB RAID circle OLTP (Debit Credit) 1 B tpd (14 k tps) Gray - Microsoft @ LANL 12/17/98

Slide 8

Windows NT Versus UNIX Best Results on a SMP: SemiLog plot demonstrates 3x (2 year) lead by UNIX Does not indicate Oracle/Alpha Cluster at 100,000 tpmC All these numbers are off-scale colossal (20,000 dynamic clients?) Gray - Microsoft @ LANL 12/17/98

Slide 9

Bottleneck Analysis Drawn to straight scale Theoretical Bus Bandwidth 422MBps = 66 Mhz x 64 bits Memory Read/Write ~150 MBps MemCopy ~50 MBps Disk R/W ~9MBps Gray - Microsoft @ LANL 12/17/98

Slide 10

Bottleneck Analysis Adapter ~70 MBps PCI ~110 MBps Adapter Memory Read/Write ~250 MBps Adapter PCI Adapter NTFS Read/Write 18 Ultra 3 SCSI on 4 strings (2x4 and 2x5) 3 PCI 64 ~ 155 MBps Unbuffered read (175 crude) ~ 95 MBps Unbuffered compose Good, yet 10x down from our UNIX brethren (SGI, SUN) 155 MBps Gray - Microsoft @ LANL 12/17/98

Slide 11

Sandia/Compaq/ServerNet/NT Sort 1.1 Terabyte (13 Billion records) in 47 minutes 68 hubs (double 450 Mhz processors) 543 plates, 1.5 M$ 1.2 GB ps system rap (2.8 GBps pap) 5.2 GB ps of plate rap (same as pap) (rap=real application execution, pap= crest promoted execution ) Gray - Microsoft @ LANL 12/17/98

Slide 12

Progress on Sorting: NT now drives both cost and execution Speedup originates from Moore’s law 40%/year Processor/Disk/Network exhibits: 60%/year (this is a product speedup). Dim - Microsoft @ LANL 12/17/98

Slide 13

Compaq AlphaServer 8400 8x400Mhz Alpha cpus 10 GB DRAM 324 9.2 GB StorageWorks Disks 3 TB crude, 2.4 TB of RAID5 STK 9710 tape robot (4 TB) WindowsNT 4 EE, SQL Server 7.0 The Microsoft TerraServer Hardware Gray - Microsoft @ LANL 12/17/98

Slide 14

TerraServer: Lots of Web Hits 35 Total Average Peak 71 30 Hits 1,065 m 8.1 m 29 m 25 Queries 877 m 6.7 m 18 m Sessions 20 Hit Count Page View Images DB Query 742 m 5.6m 15 m 15 Image Page Views 170 m 1.3 m 6.6 m 10 Users 76 k 6.4 m 48 k 5 Sessions 10 m 77 k 125 k 0 7/6/98 8/3/98 9/7/98 6/22/98 6/29/98 7/13/98 7/20/98 7/27/98 8/10/98 8/17/98 8/24/98 8/31/98 9/14/98 9/21/98 9/28/98 10/5/98 10/12/98 10/19/98 10/26/98 Date A billion web hits! 1 TB, biggest SQL DB on the Web 100 Qps normal, 1,000 Qps top 877 M SQL inquiries so far Gray - Microsoft @ LANL 12/17/98

Slide 15

SQL 7 TerraServer Availability Operating for 4 months : 3,133 hrs Unscheduled blackout: 36.5 minutes: 99.98% planned up Scheduled blackout: an hour Availability: 99.95% by and large up No NT disappointments (ever) One SQL7 Beta2 bug No disappointments in Aug, Oct Gray - Microsoft @ LANL 12/17/98

Slide 16

Backup/Restore Gray - Microsoft @ LANL 12/17/98

Slide 17

NCSA Super Cluster National Center for Supercomputing Applications University of Illinois @ Urbana 512 Pentium II cpus, 2,096 plates, SAN Compaq + HP +Myricom + WindowsNT A Super Computer for 3M$ Classic Fortran/MPI programming DCOM programming model http://access.ncsa.uiuc.edu/CoverStories/SuperCluster/super.html Gray - Microsoft @ LANL 12/17/98

Slide 18

Data Rivers: Split + Merge Streams N X M Data Streams M Consumers N makers River Producers add records to the waterway, Consumers expend records from the waterway Purely consecutive programming. Stream does stream control and buffering does segment and converge of information records River = Split/Merge in Gamma = Exchange administrator in Volcano/SQL Server. Dark - Microsoft @ LANL 12/17/98

Slide 19

Generalization: Object-situated Rivers transport sub-class of record-set (= stream of items) record sort and apportioning are a piece of subclass Node transformers are information pumps an article with waterway inputs and yields do late-tying to record-sort Programming gets to be information stream programming indicate the pipelines Compiler/Scheduler does data dividing and “transformer” position Gray - Microsoft @ LANL 12/17/98

Slide 20

NT Cluster Sort as a Prototype Using information era and sort as a prototypical application “Hello world” of appropriated handling objective: simple introduce & execute Gray - Microsoft @ LANL 12/17/98

Slide 21

Remote Install Add Registry section to every remote hub. RegConnectRegistry() RegCreateKeyEx() Gray - Microsoft @ LANL 12/17/98

Slide 22

Cluster StartupExecution MULT_QI COSERVERINFO HANDLE Sort() Sort() Sort() Setup : MULTI_QI struct COSERVERINFO struct CoCreateInstanceEx() Retrieve remote item handle from MULTI_QI struct Invoke strategies as normal Gray - Microsoft @ LANL 12/17/98

Slide 23

Cluster Sort Conceptual Model AAA BBB CCC Multiple Data Sources Multiple Data Destinations Multiple hubs Disks - > Sockets - > Disk - > Disk An AAA BBB CCC B C AAA BBB CCC AAA BBB CCC Gray - Microsoft @ LANL 12/17/