GB2526573A

GB2526573A - Controlling a server

Info

Publication number: GB2526573A
Application number: GB1409453.6A
Authority: GB
Inventors: Martin Green; Duncan Forsyth
Original assignee: Blueberry Consultants Ltd
Current assignee: Blueberry Consultants Ltd
Priority date: 2014-05-28
Filing date: 2014-05-28
Publication date: 2015-12-02
Also published as: GB201409453D0

Abstract

The resources associated with providing a service to a user 1 from a server 3 via a network are reduced by automatically deactivating the server when it is determined that the service is not required. When the server is deactivated, any requests addressed to the server are redirected via the network to a machine manager 7. The machine manager determines whether the service is required based on the request, and if so, automatically reactivates the server. The server may be a cloud based or virtual server, or a physical server. The service may only be required intermittently and may be a test website used by developers. The requests may be redirected by a proxy server between the user and server. The machine manager may determine that the service is required by inspecting the request and/or metadata associated with the request. Deactivation may occur when utilisation of the server is below a predetermined threshold for a given period of time.

Description

Intellectual Property Office Application No. GB1409453.6 RTTVT Date:19 November 2014 The following terms are registered trade marks and should be read as such wherever they occur in this document: Microsoft iOS Android Windows Phone Blackberry Windows Mac OS Linux

ARM

Intellectual Property Office is an operating name of the Patent Office www.ipo.govuk

CONTROLLING A SERVER

The invention relates to controlling a server. More particularly the invention relates to reducing the resources associated with providing a service from a server.

Severs arc frequently used to provide a service, for example to make available a website, A cost is associated with the server being in an active state. Although many services (such as live websites) arc intended to be provided without interruption, there arc circumstances where a service provided by a server is not required to be availab'e at all times. An example is where a server is used to provide a development environment in which services are being developed and/or tested, or where a vebsitc is under development or test. This type of service is typically only required to be in an active state when the developers are working, which may be confined to normal working hours in a single time zone, Sonic services niav be only used intensively for short periods of time, and otherwise left idle, A server which remains in an active state, providing services at all times may not be desirable where demand for the services is intermittent. This proMem particularly applies to virtualised or cloud based servers, Such servers may incur unnecessary resource costs when they arc providing a service at a time when it is not required. Stopping such services when the services that they provide are not required saves resources in a number of ways. For the user of the server, the cost associated with the server when it is not required can be saved. For the provider of the server, the amount of power used and/or the amount of computing infrastructure can be reduced if users make more efficient use of resources.

It is known to idcntifr and stop servers that arc not being used. For example. Amazon® provide a mechanism for identifying and stopping unused servers. The Amazon® Cloudwatch service can be used with the Amazon® Web Services Console to identify and stop servers whose services are not being used.

Auto scaling is a technique that is applied to optimise the provision of a service by increasing and decreasing the amount of servers that are providing the service as demand for the service scales. This does not address the problem of how to efficiently provide a service for which there is intermittent demand (i.e. for which there are periods in which the services are simply not required at all).

It is also known to schedule when a server should be deactivated and reactivated. Third party management toots exist for this purpose (e.g. www.rocketpeak.com). However, a pre-defined schedule that defines when servers are deactivated may not be suitable for some applications.

For instance, the service may be required intennittently in a way that is not readily predictable.

For example, flexible working patterns may mean that the demand for a service used in dcvclopment and testiug is hard to prcdict.

It may bc rclativcly difficult to rc-start a server whcn it is deactivatcd. The uscr of the servicc provided by the server may not know how to re-start the server. Re-starting the server may require authorisations that the user may not be allowed. For instance, where a server is a physical server, physical access may bc rcquired to re-start the server. Where the server is a virtual or cloud server access to a management interface may be necessary. Typically, access to such management interfaces is restricted to administrators, There may be good reasons for limiting user/developer access to such a management interface, for instance the management interface may be very complex, and there may be the potential for miseonfiguration of the server.

Figure 1 illustrates a known arrangement for server control, in which a user 1 communicates with a server 3 via a network 5. The server 3 is a virtual server, which is accessed by the user 1 via the networkS. Requests from the user 1 to the server 3 are directed by the networkS to the server 3. An administrator 2 also has access to the server 3 via the network 5, and to a management interface 4, from which administrative functions of the server 3 may be controlled.

When the server is inactive (suspended or switched oil), the user 1 is left without access to the service provided by the server 3. The administrator must restart the server 3 viathe management console 4 for the user ito again have access to the service provided by the server 3.

The need for an administrator to maniLally re-start a server may constitute a barrier to shutting down the server when die service that the server provides is not required. Shutting down and restarting a server based on a fixed scheduk may be too inflexiMe for some users.

I

The present invention seeks to address at least some of the above mentioned problems.

According to the invention, there is provided a method of controlling a server, comprising: automatically dc-activating a server that is providing a service via a network when it is determined that the service is not required; when the server is deactivated: redirecting any requests addressed to the server via the network to a machine manager; determining, at the machine manager, that the service is required, based on a 3 request addressed to the server, and automatical'y activating the server, This method provides a way of reducing the physical resources that are necessary to provide a scrvicc with a scrvcr, by solving thc significant problem of how to rcactivate thc scrvcr when it is required after is has automatically been deactivated. The method automatically reactivates the server when it is required, based on requests directed to the server: the user is not required to interact with a management interface that is intended for administrators, Instead, thc user can simply request the service in the normal way, and the machine manager can detect the demand for the service and reactivate the server, Rcdirccting the requests may be performed by a proxy server between the user and die server.

When the server is activated, the proxy server may direct requests addressed to the server to the server. V/hen the server is deactivated, the proxy sever may direct requests to the machine manager, The server may be addressed using a hosthame, and redirecting die requests may comprise updating a domain name system (DNS) zone record to resolve the hostname to an IP address of the machine manager.

Automatically activating the server may comprise updating the DNS zone record to resolve the hostname to an iF address of the server, Using DNS based redirection may be convenient, for instance, where a single provider hosts both the server and the DNS server, but this is not essential. DNS based redirection is also made convenient when the provider makes an API (application programming interthce) available for modifir'ing the DNS record, Updating die DNS record to resolve the hosbiame to an IP address of the server may be delayed until the server is operable to provide the service.

Updating the DNS record to resolve the hostname to an IP address of the server may comprise flushing a DNS cache of a user (for instaiice a browser DNS cache), Some DNS caching does not respect TTL values of DNS records, and it may be necessary to flush such DNS caches in order to clear data that is no longer live, This probkm arises in particular as a result of DNS 3 caching in some web browsers, Flushing the DNS cache may comprise instructing the cache to bc cleared, clearing a record associated with the server hosmame. or providing sufficient DNS records to the cache to displace the current DNS records.

\\Then the server is operabk to provide the service, the machine manager may forward to the server any requests received at the machine manager that are addressed to the hostname, This overcomes problems that may be caused by the redirect not being fully removed (for example due to local DNS caching).

The method may comprise responding from the machine manager to requests addressed to the server when the server is not operable to provide the service. The machine manager nmy provide an interface to the user which indicates to the user the state of the server, for instance, providing an estimate of the time for the senerto become active.

Determining that a service is not required may comprise comparing the utilisation of the server to a pre-determined threshold.

Determining that a service is not required may comprise determining that the utilisation of the server is below the pre-determined duralion for a predetermined period of time. This helps to avoid unnecessary deactivation of the server as a result ofatemporarv hiatus in demand, The machine manager may determine that the service is required by inspecting the request and/or metadata associated with the request. Inspection of the request nmy determine whether the request indicates demand for the service from a legitimate user, or may determine that the request is not from a egitimate user, and therefore does not indicate demand for the service,

I

The machine manager may determine that the service is required based on at least one of a network address of the origin of thc request, a uscr agent of the requester, and what service is requested.

According to a second aspect of the invention, there is provided a method of controlling a pumlity of servers, comprising controlling each server using the method of the first aspect, wherein a single machine manager controls each server. A single machine manager may be used to reduce the cost of operating a large number of servers. The servers may be operated by a purnlity of different organisations, and the machine manager may be operated by a further organisation (for example as a service). Alternatively, a single organisation with a plurality of servers may also operate a machine manager to reduce the costs associated with the servers.

Each of the puraliW of servers may provide a different service. I0

In anothcr aspect, a computer program is provided for instructing a computer or computer system to perform the method of the first or second aspect. In another aspect, a machine readable medium is provided comprising a computer program according to an aspect of the invention, The invention will now be described, by way of examp'e, with reference to the accompanying drawings, in which: Figure 1 is a schematic of a known server control arrangement; and Figure 2 is a schematic of a server control arrangement according to an embodiment.

Referring to Figure 2, a schematic of a server control arrangement 10 is shown. comprising a user 1, request res&ver 6, server 3, machine manager 7 and management interface 4.

The server 3 provides a service, The server 3 may be a cloud based or virtual server, but this is not essential to the invention (it is equally applicable where the server is a physical server). The service may only be required intermittently, and may for instance be a test website used by developers. When the server 3 is inactive, the service (e,g, test website) is not availaMe,

I

The management interface 4 is operable to control administrative functions of the server 3, and can be used to active the server 3 from an inactive state. and to dc-activate the server 3, for instance when the service it provides is not required. Activation of the server 3 may take some time, and there may be a delay between requesting activation of the server 3 and the service provided by the server 3 being available, The management interface 4 may comprise an application programming interface (or API) or a web based console interface, An API based management interface 4 makes it re'atively straightforward to control selected administrative functions of the server 3 nsing an application.

but this is not essential to the invention.

The request rcsolver 6 resolves requests from the uscr that are addressed to the server 3, In this example the server 3 is addressed using a hoshiame (slLch as acme.com). and the request resolver 6 represents the domain name system (DNS) resolving process. The request resolver 6 therefore translates the hostname by which the server 3 is addressed into an IP address, When the server 3 is active and operable to provide the service, the request resolver 6 is configured to provide the IP address of the server when the hostname of the server is resolved, The machine manager 7 comprises a process external to the server 3, and may be implemented in a further server. The machine manager 7 does not provide the service that is provided by the server 3. In one embodiment the machine manager 7 may be implemented on a virtual or cloud based server. Preferably the machine manager 7 is used to manage a plurality of servers 3. Each of the plurality of servers 3 that is managed by the machine manager 7 may provide a different service.

The machine manager 7 is configured to determine when the service provided by the server is not required. This can be achieved by monitoring the utilisation of the server 3, which may be performed via the management interface 4. The utilisation of the server 3 may be monitored at regular intervals. A management interface provided by cloud server provider may include the ability to monitor server utilisation, An example of resources that may indicate the utilisation of the server 3 are CPU utilisation and network utilisation. In this example the machine manager 7 determines that the service provided by the server 3 is no longer required when the server utihsation falls below a predetermined threshold. Server utilisation may be indicated by central processing unit utilisation, input output requests. network utilisation. or some other measure of activity of the server 3.

I

In some embodiments the server 3 may itself determine when the service it provides is no longer required. and the server 3 may dc-activate itself Dc-activation of the server 3 may comprise storing the state of the server 3 and suspending the processes running on the server 3.

Dc-activating the server 3 may also comprise shutting down the server 3 without storing its state.

The level of the threshold may be adjustable. Preferably it is also necessary for the server utilisation to be below the predetemtined threshold for a predetemiined duration for the machine manager 7 to determine that the service provided by the server 3 is no longer required.

3 Wli the server 3 is deactivated it no longer responds to requests addressed to it by users (but will still respond to the management interface 4, for example if instructed to reactivate), When the server 3 is deactivated, requests are therefore re-directed to the machine manager 7. This can be achieved in any suitable way, for example by domain name service (DNS) based redirection or using a proxy server. I0

Redirecting requests made to the server 3 may comprise updating a DNS record, For example, the service provided by the server 3 may comprise a test website that is undergoing testing or development. In such a case, requests addressed to the server 3 are addressed using a hostname (e.g. www.acme.com). The hostname is be resolved by the request resolver 6, which comprises a DNS server. The request resolver 6 resolves the hosthame into an IP address (e.g. 123.456.123.456). Tn this scenario, it is convenient to redirect requests addressed to the server at www.acme.com by updating the DNS record of the request res&ver 6 that defines the IP address associated with the hostname (i.e. the DNS zone record for the hosthame). it is particularly convenient if the server 3 and request resolver 6 (e.g. DNS sever) are provided by a common entity. Preferably, both the server 3 and request resolver 6 are conuollable via an API.

so that an application can readi'y be written to control their operation via the management interface 4.

In another embodiment, a proxy server may be provided between the server 3 and the network, so that all requests directed to the server 3 are first received and forwarded by the proxy server.

In this case, redirecting requests to the server 3 may comprise configuring the proxy server to forward such requests to the machine manager 7 instead of the server 3.

In some embodiments, a combination of a proxy server and DNS redirection may be appropriate, as will be explained below.

In one embodiment, when it is determined that the server 3 is to be deactivated, the machine manager 7 updates a DNS zone record associated so that requests addressed to the server 3 are res&ved to an IP address of the machine manager 7, When a user tries to request the service (e.g. a test website) from the deactivated server 3, the request will instead go to the machine manager 3.

The machine manager 7 receives requests directed to the server 3, and inspects the requests to determine whether the server 3 is required. For examp'e, inspection of the request (e.g. metadata of the request) may dctermine that it originates from a search engine robot, rather than from a user, in which case the request may not indicate that the service (slLch as a website) provided by the server 3 is required by a legitimate user. The machine manager 7 may employ a whitelist approach to detemrine whether the request is from a legitimate user of the service provided by the server 3. The machine manager 7 may check the originating address of the request, and if the address is within the whitelist, the request may be treated as indicating that the service is required. hi some embodiments the determination of whether the request is from a legitimate user may made based on at least one of: the user agent of the request, the originating address of the request, the content of the request, For example a request for certain resources such as /robots.txt may indicate web crawler traffic. A file extension of the request may also indicate whether it originates from a legitimate user. For exanipe, a request for /index.php on a server running ITS/NET (or vice versa) may indicate that the request does not originate from a legitimate user, and therefore does not indicate demand for the service. Such filtering by requested URL (universal resource locator) may be performed using a configurable blacklist.

The blacklist may be user configurable. An alternative (or additional) method for detennining demand based on the requests is to consider the frequency and/or number of requests. For example, a single resource request from a user (who may be identified by IP address and/or user agent) within a predetennined period (for example 60 seconds) may be considered accidental and ignored. A plurahty of requests within the predetermined period may indicate sufficient confidence to result in the server being instructed to reactivate, This approach may be termed greylisting. A request that conforms to the list is necessary but not sufficient to trigger a response. The pattern of requests over time, in combination with their conformance to the greylist, indicates demand for the service. An approach that uses temporal filtering may exclude some automated crawlers and indexers, which tend to be designed for efficiency, and so do not repeat requests.

When the machine manager 7 has determined that the service is required, the machine manager is operable to re-activate the server 3, for instance by re-starting the server In the case of a cloud based or virtua' server, a management APT may be used to re-activate the server 3.

In some embodiments. when the machine manager 7 receives a request addressed to the server 3 that appears to be from a kgitimate user, the machine manager 7 may present an appropriate re-activation interface to the user. The re-activation interface may comprise information that indicates to the user that their request has resdted in the server 3 being reactivated, with an 3 indicator of the progress of re-starting. For instance the re-activation interface may provide an estimate as to when the server 3 will be available following re-activation.

In some embodiments the re-activation interface may request a username and/or password for re-activating the server 3. The machine manager 7 may only re-activate the server if an authorised username and password are provided.

Once the server 3 is reactivated, requests addressed to the server 3 are no longer redirected. The machine manager 7 may wait until the server 3 is operable to provide the service before removing the redirection, For instance in the case where requests are subject to DNS redirection, the machine manager 7 may wait until the server 3 is operable to provide the scrvice before updating the DNS zone record, There may be a delay associated with removing the redirection, For instance, a DNS record may include a time to live (TTL), The TTL specifies how long the DN S record should be treatcd as live. When the TTL period for a DNS record has expired, the DNS record should be checked again against the nominated domain name server. Where DNS based redirection is emp'oyed, the TTL of the DNS record should preferably be set to a minimum value. When the server 3 is to be reactivated, the DNS record may be updated at a time selected based on the TTL period of the DNS record and of the time the server 3 takes to re-start, such that the redirection is removed at substantially the same time that the server 3 becomes active (and operable to provide the service), The flL period may preferably be set to a relatively small time interval, such as: 100, 30 25, 20, 15, 10, 5. 4, 3. 2 or less; or at the minimlLm time period available.

One potenti& problem with DNS redirection is that of DNS caching that may not respect the TTL v&ue of a DNS record, For example, a number of web browsers cache DNS records in a way that does not respect the TTL associated with the DNS record. Under these circumstances, when a DNS record is lLpdatcd to remove a redirection to the machine manager 7, a user's browser cache may retain the old DNS record, even after the TTL of the DNS record indicates that it has expired. In this case the user would not be aNe to access the service even after the server 3 was reactivated and the redirection removed, until their browser DNS cache was updated.

In some embodiments a means of flushing or dearing such a loc& browser DNS cache may be 3 provided. In one embodiment the machine manager 7 may be configured to provide a large number of DNS records to thc browser, thereby flushing tile browser cache of all previous records. When the user requests a DNS lookup of the hostname of the service, the browser will thcn be forced to looklLp thc DNS rccord from sourecs that rcspect thc TTL value of thc DNS record for the server 3. I0

In addition, or as an alternative to flushing a browser DNS caehc whcn the rcdireet is removed, the machine manager 7 may also include a proxy server service that is operable to transparently redirect requests received by the machine manger 7 that are addressed to the server 3 when the server 3 is active. If a delay in removing the redirect is expected (for example due to browser caching of a DNS record), when thc scvcr 3 is activatcd thc machine managcr 7 can acfivatc a proxy server that directs ally requests arriving at the machine manager 7 that are directed to the server 3. The user may be unaware that their requests are being routed through a proxy server of thc machine managcr 7. because the requests will be passcd straight through to the server 3.

The proxy scrver may rcmain active for a predctcrmincd period of timc that is sclectcd to cnsurc that the redirect has been removed before the proxy server is deactivated, 011cc the redirect has becn fully rcmovcd (including updating DNS records in a brovscr cache, if necessary).

deactivating the proxy will have no effect, because no requests should be received by the machine manager 7 that are directed to the server 3. since the hostname of the service provided by the server 3 should then be resolved to the IP address of the server. In some embodiments the proxy server at the machine manger 7 may remain active at all times that the server 3 is active, so that any requests received at the machine manager 7 that arc intended for the active server 3 are redirected.

In addition to the thnctionality described above, the machine manager 7 may also indude a scheduling function, so that the machine manager 7 is operable to deactivate the server 3 according to a schedule. The scheduling may eonnolled by one or more rules. The rules may be based on a time range, selected from at least one of every day, every weekday, every Saturday and Sunday, on a recurring day during each week, on a recurring day during each fortnight, on a recurring day during each month, on a specific date each year, and on a specific date. The machine manager 7 may be configured to override the schedule if the service provided by the server 3 is required when the schedule calls for the sen'erto be shut down.

The machine manager 7 may also comprise means for delegating control of the state of the 3 server 3 to an authorised user of that server 3, For instance, the machine manager 7 may comprise an interface for manually activating and deactivating the server, and for controlling the scheduling of activation and deactivation.

An interface provided by the machine manager 7 may provides at least one of the following capabilities: displaying the status of servers for which the user has control, providing an interface for a user to manually start or stop each server, providing visual feedback during a server start process (e.g. indicating progress), providing the user with a simple mechanism for launching the rdevaiit client for access to the server (such as a web browser or Microsoft RDP client). The machine manager 7 interface may be implemented for both mobile platforms (e.g. iOS, Android. Windows Phone. Blackberry. etc) and desktop platforms (Windows, Mac OS, Linux, etc).

The machine manager 7 interface may provide a simplified mechanism for accessing a subset of the commands available via the management interface 4. This avoids the need to involve system administrators in routine activation and deactivation of the server 3. and also avoids the need to ifill access to a potentially complex management interface to a user who only needs to use a subset of the commands available.

Although embodiments have been described in which the server 3 is a cloud or virtual server, it will be appreciated that the invention is also applicable to a physical server. The server may for instance be a physical server located at an office that is used by geographically dispersed users via a network on an intermittent basis. The machine manager 7 may be implemented using a physical computer or server operable via a network. The machine manager 7 server may comprise a processor that imp'ements an ARM instruction set or another suitable instruction set that provides a low power architecture. The machine manager 7 computer may activate and deactivate the physical server 3 by controlling a power system of the server 3, Embodiments of the invention significantly reduce a rate of power consumption of servers 3 that provide a service that is only required intermittently. Furthermore, a significant cost saving can be achieved for a user that pays for servers based on the anwunt of time that they are active.

Embodiments of the invention cause a server to run more efficiently, because the server is deactivated when it is not required. In sonic embodiments the ongeviW of a server may be increased by deactivating the server when it is not required. Maintenance of the server may be simplified, because the server can be repaired or maintained when it is inactive, Prior to de-activating, the server may be configured to perform maintenance fttnctions such as defragmenting data on a storage mediunt or checking a file system. Alternative, the server may be configured to perform such maintenance operations when it is deactivated.

A number of other variations are possibk within the scope of the invention, which is defined by the appended claims.

Claims

Claims 1. A method of controlling a server, comprising: automatically dc-activating a server that is providing a service via a network when it is 3 determined that the service is not required; when the server is deactivated: redirecting any requests addressed to the server via the network to a machine manager; determining, at the machine manager, that the service is required, based on a request addressed to the server, and automatically reactivating the server.
2. The method of claim 1, wherein requests are redirected by a proxy server between the user and the server.
3. The method of claim 1 or 2, wherein the server is addressed using a hostoame. and redirecting the requests comprises updating a DNS zone record to resolve the hostname to an IP address of the machine manager.
4. The method of claim 3, wherein the step of automatically reactivating the server comprises updating the DNS zone record to resolve the hostoame to an IP address of the server.
5. The method of claim 4. wherein updating the DNS record to resolve the hostname to an IP address of the server is delayed until the server is operable to provide the service.
6. The method of claim 4 or 5, wherein updating the DNS record to resolve the hostname to an IP address of the server comprises flushing a DNS cache of a user, such as a browser cache.
7. The method of any ofcaims 4 to 6, wherein when the server is operable to provide the service. the machine manager f'onvards to the server any requests received at the machine manager that are addressed to the hostname,
S. The method of any preceding daim. comprising responding from the machine manager to requests addressed to the server when the server is not operable to provide the service.
9. The method of any preceding claim, wherein determining that a service is not required comprises comparing the utilisation of the server to a pre-determined threshold.
10, The method of claim 9, wherein the detemiining that a service is not required comprises 3 determining that the utilisation of the server is below the pre-determined duration for a predetermined period of time.
11. The method of any preceding claim, wherein the machine manager determines that the service is required by inspecting the request andlor metadata associated with the request. I012, The method of claim 11, wherein the machine manager determines that the service is required based on at least one of a network address of the origin of the request, a user agent of the requester, and what service is requested.13. A method of controlling a plurality of servers, comprising controlling each server using the method of any preceding claim, wherein a single machine manager controls each server.14, The method of claim 13, wherein each of the plurality of servers provides a different service.15. A computer program for instructing a computer or computer system to perform the method of any preceding claim.16, A machine readabk medium comprising a computer program according to claim IS,