14 Distributed Computing
619
password which is easily guessed is denominated “weak”. Simple measures exist
to avoid weak passwords: 17 The password should be as long as possible and
contain a mix of numbers and upper and lower case letters. Many installations
require users to change their passwords periodically in order to improve computer
security. A very dangerous practice which should be completely avoided, but
which is unfortunately common practice in the particle physics community, is the
“sharing” of personal accounts (and hence, their passwords) or the use of “service”
accounts with passwords known by dozens of persons, written in documents and
blackboards or even posted on Web sites. Modern operating system features make
these practices completely unnecessary, as they can be configured so that each user
can first authenticate with their personal account and then gain access to a common
environment to perform the tasks required.
Portable devices (“laptops” and “smartphones”) can be deployed so that they are
elements of a cluster and therefore part of the security domain. Since the portable
computer must remain useable when temporarily disconnected from the network,
the operating system will keep a local copy (or “cache” 18 ) of the security data,
including the passwords. This means that a stolen portable computer is a computer
security threat. Care should be taken in configuring the “suspended” or “hibernated”
modes in portable computers to ask for a password when they are turned back on.
Another major issue with authentication based on usernames and passwords is
that each cluster (and each online service such as electronic mail, social networks,
and document and file repositories), being a separate security environment, requires
a separate username and password and has its own policies for requiring password
changes. Users are becoming overwhelmed in keeping track of their usernames
and passwords and their reaction is often to use weak passwords or keep a list of
passwords in a file which can be stolen. In order to reduce username/password
proliferation, many organizations are linking their clusters and online services to
shared authentication servers, thus providing a “single-sign-on” environment.
In the 2000s, organizations, most notably the CERN Large Hadron Collider
experiments, started to use an alternative authentication and authorization frame17 Users should consult their local security rules for specific recommendations on password
choices.
18 The term “cache” is used in computer architecture to describe a local copy of a limited
amount of data which is normally stored elsewhere. The introduction of this local copy serves
to increase the performance and efficiency of use of the element connected to it, by smoothing
out peaks and valleys in storing and retrieving information, or avoiding retrieves altogether if the
same information is needed repetitively. The cache is transparent to the element accessing the
information; if the data needed is in cache, it will be delivered from there (a cache “hit”) whereas
if the data is not in cache, the normal access path to the normal storage place will be used (a cache
“miss”). The key to successful implementation of a cache is the right choice of algorithm to choose
which data to copy and keep in the cache. Examples are high-speed memory caches in CPU chips
which hold local copies of instructions and data, and memory caches in disk controllers which hold
local copies of disk blocks.
619
password which is easily guessed is denominated “weak”. Simple measures exist
to avoid weak passwords: 17 The password should be as long as possible and
contain a mix of numbers and upper and lower case letters. Many installations
require users to change their passwords periodically in order to improve computer
security. A very dangerous practice which should be completely avoided, but
which is unfortunately common practice in the particle physics community, is the
“sharing” of personal accounts (and hence, their passwords) or the use of “service”
accounts with passwords known by dozens of persons, written in documents and
blackboards or even posted on Web sites. Modern operating system features make
these practices completely unnecessary, as they can be configured so that each user
can first authenticate with their personal account and then gain access to a common
environment to perform the tasks required.
Portable devices (“laptops” and “smartphones”) can be deployed so that they are
elements of a cluster and therefore part of the security domain. Since the portable
computer must remain useable when temporarily disconnected from the network,
the operating system will keep a local copy (or “cache” 18 ) of the security data,
including the passwords. This means that a stolen portable computer is a computer
security threat. Care should be taken in configuring the “suspended” or “hibernated”
modes in portable computers to ask for a password when they are turned back on.
Another major issue with authentication based on usernames and passwords is
that each cluster (and each online service such as electronic mail, social networks,
and document and file repositories), being a separate security environment, requires
a separate username and password and has its own policies for requiring password
changes. Users are becoming overwhelmed in keeping track of their usernames
and passwords and their reaction is often to use weak passwords or keep a list of
passwords in a file which can be stolen. In order to reduce username/password
proliferation, many organizations are linking their clusters and online services to
shared authentication servers, thus providing a “single-sign-on” environment.
In the 2000s, organizations, most notably the CERN Large Hadron Collider
experiments, started to use an alternative authentication and authorization frame17 Users should consult their local security rules for specific recommendations on password
choices.
18 The term “cache” is used in computer architecture to describe a local copy of a limited
amount of data which is normally stored elsewhere. The introduction of this local copy serves
to increase the performance and efficiency of use of the element connected to it, by smoothing
out peaks and valleys in storing and retrieving information, or avoiding retrieves altogether if the
same information is needed repetitively. The cache is transparent to the element accessing the
information; if the data needed is in cache, it will be delivered from there (a cache “hit”) whereas
if the data is not in cache, the normal access path to the normal storage place will be used (a cache
“miss”). The key to successful implementation of a cache is the right choice of algorithm to choose
which data to copy and keep in the cache. Examples are high-speed memory caches in CPU chips
which hold local copies of instructions and data, and memory caches in disk controllers which hold
local copies of disk blocks.