Presentation on theme: "Services CPTE 433 Chapter 5 PowerPoint by John Beckett."— Presentation transcript:
Services CPTE 433 Chapter 5 PowerPoint by John Beckett
What is a service? Specified input Specified output Audience Privileges –May be different for user class or individual Target uptime Notice: We have said nothing about implementation!
The SA role Ascertain services needed Provide services as needed How Use server-class equipment Use server-class methods –Avoiding linkages that reduce reliability
The Distribution Dilemma One Big Server Put all your eggs in one basket Take very good care of that basket Simplifies gross support Intra-machine interdependencies Many Small Servers One server won’t take down the system May reduce overall downtime May simplify detailed support Interdependencies remain!
The VM Option Big machine running a Virtual Machine OS Multiple instances of server OS’s under that Advantages of “Big Server” Mitigates disadvantages Encourages consistent server OS environments Allocates resources as needed (better fit)
Bandwidth versus Latency “I hate the trucks on the Interstate!” –But: what would traffic be like if all those items were carried on smaller truck? –But: How much would it cost if we did it that way? Example: TTY/Fortran Today: Client-side packaging of data –However: Don’t give the inmates the keys!
Latency - 2 More bandwidth doesn’t solve a latency problem. –By encouraging development of unneeded services it can actually make the problem worse! “Minimize the number of packets in high- latency nets.” Good reason for client-side data validation –But validate on server-side as well!
Service Names Don’t tie server names to functions Example: device numbers for printers Do nslookup: hw.cs.southern.edu display.cs.southern.edu
What should be centralized? Excellent example: wireless in HSC SOC installs wireless so they can use it. IS switches SOC’s wireless devices to new “wireless” network (beta test). Eventually IS replaces SOC’s wireless devices with what they’ve found best for the task. This requires cooperation and knowledge –Which you won’t always have –Cultivate it if you have or might get it!
Data Backups An excellent service to provide –But perhaps one of the most difficult A needed service –Clients should be imaged –Data should be kept on a data server
Remote Administration Sounds like a way to integrate out-sourced vendors –Do not forget the trust issue –Consider conflicts of interest Serial consoles – on the way out –Most systems work via ssh-implemented telnet or a web server
People and Servers People and servers don’t really belong in the same room –People need a warmer environment –Servers need more consistent temperature, without people coming in and out all the time
Server Appliance A device that performs a network-based service – e.g.: –WebCam –Printer –Network-Attached Storage –SOHO networking device (router, or even a more comprehensive device)
Axentra Rumba VIA C3 800 MHz 256 MB 133 MHz 40GB Hard Drive 2 LAN Ports (10/100 BT) 1 Wan Port (10/100 BT) $599 Internet Gateway: Broadband connectivity (ADSL/Cable modem) DNS, IP filtering, port forwarding NAT firewall UPnP (Universal Plug and Play) Local Networking: Ethernet router/switch Wi-Fi* wireless networking (802.b/g) Network Services: Web Server (http/https) File Server, Web folders (Web DAV) Email Server (IMAP/SMTP) Print Server (SAMBA) Content filtering Spam filtering Web Applications: Web administration Web publishing Web mail Web address book Web bookmarks Photo album publishing Note: This product has been replaced by Net-Box Services
Power Cords Beware – they can work loose Know where they go Relation to UPS Components of the server
Reliability Making it simple makes it more reliable Avoid unnecessary inter-dependencies Fix problems before they get large –“Soft” outages become “hard” outages
Centralization/Standards Just like workstations, server standards save support money But…you don’t buy servers as often and they cost more Look for economies of scale Argument for lots of small servers –Or virtualizing on a large array of servers
Why Monitor? To help you diagnose current problems Capacity planning Look-back on problem history –Perhaps in a criminal investigation –Beware: Logs can be helpful, but must be carefully interpreted
Service Rollout Plan the sequence Work your plan Log actual events Post-mortem on rollout
Redundancy May be servers, services, RAID, spare machines for quick roll-out How quickly is your redundant system brought on-line? How well will it handle the load? How will you get back to normal operation?