Introduction to Information Retrieval

Introduction to
Information Retrieval
Information Retrieval and Web Search

Basic inverted index construction

Introduction to Information Retrieval Ch. 4

Index construction
 How do we construct an index?
 What strategies can we use with limited main
memory?
Introduction to Information Retrieval Sec. 4.2

Recall index construction

 Documents are parsed to extract words and these
are saved with the Document ID.

Doc 1 Doc 2

I did enact Julius So let it be with

Caesar I was killed Caesar. The noble
i' the Capitol; Brutus hath told you
Brutus killed me. Caesar was ambitious
Introduction to Information Retrieval Sec. 4.2

Key step
 After all documents have been
parsed, the inverted file is
sorted by terms.

We focus on this sort step.

Introduction to Information Retrieval Sec. 4.2

RCV1: Our collection for this lecture

 As an example for applying scalable index
construction algorithms, we will use the
Reuters RCV1 collection.
 This is one year of Reuters newswire (part of 1995
and 1996)

 The collection isn’t really large enough, but it’s

publicly available and is a plausible example.
Introduction to Information Retrieval Sec. 4.2

Reuters RCV1 statistics

 symbol statistic value
 N documents 800,000
 L avg. # tokens per doc 200
 M terms (= word types) 400,000
 avg. # bytes per token 6
(incl. spaces/punct.)
 avg. # bytes per token 4.5
(without spaces/punct.)
 avg. # bytes per term 7.5
 non-positional postings 100,000,000
4.5 bytes per word token vs. 7.5 bytes per word type: why?
Introduction to Information Retrieval Sec. 4.2

Sort-based index construction

 As we build the index, we parse docs one at a time.
 The final postings for any term are incomplete until the end.
 At 8 bytes per (termID, docID), demands a lot of space for large
collections.
 T = 100,000,000 in the case of RCV1
 So … we can do this in memory today, but typical collections
are much larger. E.g., the New York Times provides an index
of >150 years of newswire
 Thus: We need to store intermediate results on disk.
Introduction to Information Retrieval Sec. 4.2

Scaling index construction

 In-memory index construction does not scale
 Can’t stuff entire collection into memory, sort, then write
back
 How can we construct an index for very large
collections?
 Taking into account hardware constraints. . .
 Memory, disk, speed, etc.

 Let’s review some hardware basics

Introduction to Information Retrieval Sec. 4.1

Hardware basics
 Servers used in IR systems now typically have several
GB of main memory, sometimes tens of GB.
 Available disk space is several (2–3) orders of
magnitude larger.
 Fault tolerance is very expensive: It’s much cheaper
to use many regular machines rather than one fault
tolerant machine.
Introduction to Information Retrieval Sec. 4.1

Hardware basics
 Access to data in memory is much faster than access
to data on disk.
 Disk seeks: No data is transferred from disk while the
disk head is being positioned.
 Therefore: Transferring one large chunk of data from
disk to memory is faster than transferring many small
chunks.
 Disk I/O is block-based: Reading and writing of entire
blocks (as opposed to smaller chunks).
 Block sizes: 8KB to 256 KB.
Introduction to Information Retrieval Sec. 4.1

Hardware assumptions (circa 2007)

 symbol statistic value
 s average seek time 5 ms = 5 x 10−3 s
 b transfer time per byte 0.02 μs = 2 x 10−8 s
 processor’s clock rate 109 s−1
 p low-level operation 0.01 μs = 10−8 s
(e.g., compare & swap a word)
 size of main memory several GB
 size of disk space 1 TB or more
Introduction to Information Retrieval Sec. 4.2

Sort using disk as “memory”?

 Can we use the same index construction algorithm
for larger collections, but by using disk instead of
memory?
 No: Sorting T = 100,000,000 records on disk is too
slow – too many disk seeks.
 We need an external sorting algorithm.
Introduction to Information Retrieval

Introduction to
Information Retrieval
CS276: Information Retrieval and Web Search

External memory indexing

Introduction to Information Retrieval Sec. 4.2

BSBI: Blocked sort-based Indexing

(Sorting with fewer disk seeks)
 8-byte records (termID, docID)
 These are generated as we parse docs
 Must now sort 100M such 8-byte records by termID
 Define a Block ~ 10M such records
 Can easily fit a couple into memory
 Will have 10 such blocks to start with
 Basic idea of algorithm:
 Accumulate postings for each block, sort, write to disk
 Then merge the blocks into one long sorted order
Introduction to Information Retrieval Sec. 4.2
Introduction to Information Retrieval Sec. 4.2

Sorting 10 blocks of 10M records

 First, read each block and sort within:
 Quicksort takes O(N ln N) expected steps
 In our case N=10M

 10 times this estimate – gives us 10 sorted runs of

10M records each.
 Done straightforwardly, need 2 copies of data on disk
 But can optimize this
Introduction to Information Retrieval Sec. 4.2

How to merge the sorted runs?

 Can do binary merges, with a merge tree of log210 = 4 layers.
 During each layer, read into memory runs in blocks of 10M,
merge, write back.
brutus d1,d3,d6,d7

brutus d1,d3 brutus d6,d7 caesar d1,d2,d4,d8,d9

caesar d1,d2,d4 caesar d8,d9 julius d10

noble d5 julius d10 killed d8

with d1,d2,d3,d5 killed d8 noble d5

with d1,d2,d3,d5
Postings lists
to be merged Merged
postings list
Disk
Introduction to Information Retrieval Sec. 4.2

How to merge the sorted runs?

 But it is more efficient to do a multi-way merge, where you
are reading from all blocks simultaneously
 Open all block files simultaneously and maintain a read
buffer for each one and a write buffer for the output file
 In each iteration, pick the lowest termID that hasn’t been
processed using a priority queue
 Merge all postings lists for that termID and write it out

 Providing you read decent-sized chunks of each block into

memory and then write out a decent-sized output chunk,
then you’re not killed by disk seeks
Introduction to Information Retrieval Sec. 4.3

Remaining problem with sort-based

algorithm
 Our assumption was: we can keep the dictionary in
memory.
 We need the dictionary (which grows dynamically) in
order to implement a term to termID mapping.
Introduction to Information Retrieval Sec. 4.3

SPIMI:
Single-pass in-memory indexing
 Key idea 1: Generate separate dictionaries for each
block – no need to maintain term-termID mapping
across blocks.
 Key idea 2: Don’t sort. Accumulate postings in
postings lists as they occur.
 With these two ideas we can generate a complete
inverted index for each block.
 These separate indexes can then be merged into one
big index.
Introduction to Information Retrieval Sec. 4.3

SPIMI-Invert

 Merging of blocks is analogous to BSBI.

Introduction to Information Retrieval

SPIMI in action
Sorted
Input token Dictionary dictionary
Caesar d1 brutus d1 d3 brutus d1 d3

with d1 caesar d1 d2 d4
Brutus d1 with d1 d2 d3 d5
noble d5
Caesar d2 noble d5
with d1 d2 d3 d5
with d2
Brutus d3 caesar d1 d2 d4

with d3
Caesar d4
noble d5
with d5

23
Introduction to Information Retrieval Sec. 4.3

SPIMI: Compression
 Compression makes SPIMI even more efficient.
 Compression of terms
 Compression of postings
 More on this later …

Original publication on SPIMI: Heinz and Zobel (2003)

Introduction to Information Retrieval

Introduction to
Information Retrieval
CS276: Information Retrieval and Web Search

Distributed indexing
Introduction to Information Retrieval Sec. 4.4

Distributed indexing
 For web-scale indexing (don’t try this at home!):
must use a distributed computing cluster
 Individual machines are fault-prone
 Can unpredictably slow down or fail
 How do we exploit such a pool of machines?
Introduction to Information Retrieval Sec. 4.4

Web search engine data centers

 Web search data centers (Google, Bing, Baidu) mainly
contain commodity machines.
 Data centers are distributed around the world.
 Estimate: Google ~1 million servers, 3 million
processors/cores (Gartner 2007)
Introduction to Information Retrieval Sec. 4.4

Massive data centers

 If in a non-fault-tolerant system with 1000 nodes,
each node has 99.9% uptime, what is the uptime of
the entire system?
 Answer: 37% - meaning, 63% of the time one or
more servers is down.
 Exercise: Calculate the number of servers failing per
minute for an installation of 1 million servers.
Introduction to Information Retrieval Sec. 4.4

Distributed indexing
 Maintain a master machine directing the indexing job
– considered “safe”.
 Break up indexing into sets of (parallel) tasks.
 Master machine assigns each task to an idle machine
from a pool.
Introduction to Information Retrieval Sec. 4.4

Parallel tasks
 We will use two sets of parallel tasks
 Parsers
 Inverters
 Break the input document collection into splits
 Each split is a subset of documents (corresponding to
blocks in BSBI/SPIMI)
Introduction to Information Retrieval Sec. 4.4

Data flow
assign Master assign
Postings

Parser a-f g-p q-z Inverter a-f

Parser a-f g-p q-z

Inverter g-p

splits Inverter q-z

Parser a-f g-p q-z

Map Segment files Reduce

phase phase
Introduction to Information Retrieval Sec. 4.4

Parsers
 Master assigns a split to an idle parser machine
 Parser reads a document at a time and emits
(term, doc) pairs
 Parser writes pairs into j partitions
 Example: Each partition is for a range of terms’ first
letters
 (e.g., a-f, g-p, q-z) – here j = 3.
 Now to complete the index inversion
Introduction to Information Retrieval Sec. 4.4

Inverters
 An inverter collects all (term,doc) pairs (= postings)
for one term-partition.
 Sorts and writes to postings lists
Introduction to Information Retrieval

Example for index construction

Map: Caesar conquered

d1 : C came, C c’ed.
d2 : C died.
→
<C,d1>, <came,d1>, <C,d1>, <c’ed, d1>, <C, d2>, <died,d2>

Reduce:
(<C,(d1,d1,d2)>, <died,(d2)>, <came,(d1)>, <c’ed,(d1)>)
→
(<C,(d1:2,d2:1)><died,(d2:1)>, <came,(d1:1)>,<c’ed,(d1:1)>)
34
Introduction to Information Retrieval Sec. 4.4

Index construction
 Index construction was just one phase.
 Another phase: transforming a term-partitioned
index into a document-partitioned index.
 Term-partitioned: one machine handles a subrange of
terms
 Document-partitioned: one machine handles a subrange of
documents
 As we’ll discuss in the web part of the course, most
search engines use a document-partitioned index …
better load balancing, etc.
Introduction to Information Retrieval Sec. 4.4

MapReduce
 The index construction algorithm we just described is
an instance of MapReduce.
 MapReduce (Dean and Ghemawat 2004) is a robust
and conceptually simple framework for distributed
computing …
 … without having to write code for the distribution
part.
 They describe the Google indexing system (ca. 2002)
as consisting of a number of phases, each
implemented in MapReduce.
Introduction to Information Retrieval Sec. 4.4

Schema for index construction in

MapReduce
 Schema of map and reduce functions
 map: input → list(k, v) reduce: (k,list(v)) → output
 Instantiation of the schema for index construction
 map: collection → list(termID, docID)
 reduce: (<termID1, list(docID)>, <termID2, list(docID)>, …) →
(postings list1, postings list2, …)
Introduction to Information Retrieval

Introduction to
Information Retrieval
CS276: Information Retrieval and Web Search
Dynamic indexing
Introduction to Information Retrieval Sec. 4.5

Dynamic indexing
 Up to now, we have assumed that collections are
static.
 They rarely are:
 Documents come in over time and need to be inserted.
 Documents are deleted and modified.
 This means that the dictionary and postings lists have
to be modified:
 Postings updates for terms already in dictionary
 New terms added to dictionary
Introduction to Information Retrieval Sec. 4.5

Simplest approach
 Maintain “big” main index
 New docs go into “small” auxiliary index
 Search across both, merge results
 Deletions
 Invalidation bit-vector for deleted docs
 Filter docs output on a search result by this invalidation
bit-vector
 Periodically, re-index into one main index
Introduction to Information Retrieval Sec. 4.5

Issues with main and auxiliary indexes

 Problem of frequent merges – you touch stuff a lot
 Poor performance during merge
 Actually:
 Merging of the auxiliary index into the main index is efficient if we
keep a separate file for each postings list.
 Merge is the same as a simple append.
 But then we would need a lot of files – inefficient for OS.
 Assumption for the rest of the lecture: The index is one big
file.
 In reality: Use a scheme somewhere in between (e.g., split
very large postings lists, collect postings lists of length 1 in one
file etc.)
Introduction to Information Retrieval Sec. 4.5

Logarithmic merge
 Maintain a series of indexes, each twice as large as
the previous one
 At any time, some of these powers of 2 are instantiated
 Keep smallest (Z0) in memory
 Larger ones (I0, I1, …) on disk
 If Z0 gets too big (> n), write to disk as I0
 or merge with I0 (if I0 already exists) as Z1
 Either write merge Z1 to disk as I1 (if no I1)
 Or merge with I1 to form Z2
Introduction to Information Retrieval

Logarithmic merge in action

≤n Z0 Z0 Z0 Z0
n I0 I0 I0 I0
2n I1 I1 I1 I1
4n I2
8n
16n

43
Introduction to Information Retrieval Sec. 4.5
Introduction to Information Retrieval Sec. 4.5

Logarithmic merge
 Auxiliary and main index:
 T/n merges where T is # of postings and n is size of auxiliary
 Index construction time is O(T2/n) as in the worst case a
posting is touched T/n times
 Logarithmic merge: Each posting is merged at most
O(log (T/n)) times, so complexity is O(T log (T/n))
 So logarithmic merge is much more efficient for index
construction
 But query processing now requires the merging of
O(log (T/n)) indexes
 Whereas it is O(1) if you just have a main and auxiliary index
Introduction to Information Retrieval Sec. 4.5

Further issues with multiple indexes

 Collection-wide statistics are hard to maintain
 E.g., when we speak of spell-correction: which of
several corrected alternatives do we present to the
user?
 We may want to pick the one with the most hits
 How do we maintain the top ones with multiple indexes
and invalidation bit vectors?
 One possibility: ignore everything but the main index for
such ordering
 Will see more such statistics used in results ranking
Introduction to Information Retrieval Sec. 4.5

Dynamic indexing at search engines

 All the large search engines now do dynamic indexing
 Their indices have frequent incremental changes
 News items, blogs, new topical web pages
 But (sometimes/typically) they also periodically
reconstruct the index from scratch
 Query processing is then switched to the new index, and
the old index is deleted
Introduction to Information Retrieval

Earlybird: Real-time search at Twitter

 Requirements for real-time search
 Low latency, high throughput query evaluation
 High ingestion rate and immediate data availability
 Concurrent reads and writes of the index
 Dominance of temporal signal

48
Introduction to Information Retrieval

Earlybird: Index organization

 Earlybird consists of multiple index segments
 Each segment is relatively small, holding up to 223 tweets
 Each posting in a segment is a 32 bit word: 24 bits for the
tweet id and 8 bits for the position in the tweet
 Only one segment can be written to at any given time
 Small enough to be in memory
 New postings are simply appended to the postings list
 But the postings list is traversed backwards to prioritize
newer tweets
 The remaining segments are optimized for read-only
 Postings sorted in reverse chronological order (newest first)
49
Introduction to Information Retrieval Sec. 4.5

Other sorts of indexes

 Positional indexes
 Same sort of sorting problem … just larger Why?
 Building character n-gram indexes:
 As text is parsed, enumerate n-grams.
 For each n-gram, need pointers to all dictionary terms
containing it – the “postings”
Introduction to Information Retrieval Ch. 4

Resources for today’s lecture

 Chapter 4 of IIR
 MG Chapter 5
 Original publication on MapReduce: Dean and
Ghemawat (2004)
 Original publication on SPIMI: Heinz and Zobel (2003)
 Earlybird: Busch et al, ICDE 2012