Remote Procedure Call CS-4513 Distributed Systems Hugh C. Lauer Slides include materials from Modern Operating Systems, 3rd ed., by Tannenbaum, Operating System Concepts, 7th ed., by Silbershatz, Galvin, & Gagne, Distributed Systems: Principles & Paradigms, 2nd ed. By Tanenbaum and Van Steen, and Distributed Systems: Concepts and Design, 4th ed., by Coulouris, et. al. CS-4513, D-Term 2010 Remote Procedure Call 1 Review Definition — Protocol • Formal set of rules that govern the formats, contents, and meanings of messages from computer to computer, process to process, etc. • Must be agreed to by all parties to a communication • May be defined in terms of other protocols • E.g., 7-layer OSI model CS-4513, D-Term 2010 Remote Procedure Call 2 Message-oriented Protocols • Many in widespread use • Traditional TCP/IP and Internet protocols • Difficult to design and implement • Especially with more sophisticated applications • Many difficult implementation issues for each new protocol and each platform • Formatting • Uniform representation of data • Client-server relationships • … CS-4513, D-Term 2010 Remote Procedure Call 3 Alphabet Soup of many, many Protocols • TCP, UDP, IP, NCP, SMTP, SNNP, NNTP, FTP, TFTP, POP, IMAP, HTTP, VMRL, … • Appletalk, Netware, … • Remote Procedure Call, NFS, … • CORBA, GLOBE, JINI, … • Network Streaming, … • … CS-4513, D-Term 2010 Remote Procedure Call 4 Message-oriented Protocols • Many in widespread use • Traditional TCP/IP and Internet protocols • Difficult to design and implement • Especially with more sophisticated applications • Many difficult implementation issues for each new protocol and each platform • Formatting • Uniform representation of data • Client-server relationships • … CS-4513, D-Term 2010 Remote Procedure Call 5 Problem • How to make sense out of many different kinds of protocols • How to design new protocols for new, more sophisticated applications • How to automate the nitty-gritty of protocol design and testing CS-4513, D-Term 2010 Remote Procedure Call 6 Solution — Remote Procedure Call (RPC) • The most common framework for newer protocols and for middleware • Used both by operating systems and by applications – NFS (Network File System) is implemented as a set of RPCs – DCOM, CORBA, Java RMI, etc., are just RPC systems • Seminal reference – Birrell, Andrew D., and Nelson, Bruce, “Implementing Remote Procedure Calls,” ACM Transactions on Computer Systems, vol. 2, #1, February 1984, pp 39- 59. (.pdf) CS-4513, D-Term 2010 Remote Procedure Call 7 Reading Assignment • Tanenbaum, 8.2.4 • Also online tutorials on this slide:– CS-4513, D-Term 2010 Remote Procedure Call 8 Remote Procedure Call (RPC) • Fundamental idea: – – Server process exports an interface of procedures or functions that can be called by client programs • similar to library API, class definitions, etc. • Clients make local procedure/function calls – As if directly linked with the server process – Under the covers, procedure/function call is converted into a message exchange with remote server process CS-4513, D-Term 2010 Remote Procedure Call 9 Ordinary procedure/function call count = read(fd, buf, bytes) CS-4513, D-Term 2010 Remote Procedure Call 10 Remote Procedure Call • Would like to do the same if called procedure or function is on a remote server CS-4513, D-Term 2010 Remote Procedure Call 11 Solution — a pair of Stubs • A client-side stub is a function that looks to the client as if it were a callable function of a local service – I.e., same API as the service’s implementation of the function • Service-side stub looks like a local client calling the service – I.e., like a hunk of code invoking the service function • The client program thinks it’s invoking a local service – but it’s really calling into the client-side stub • The service program thinks it’s called by a local client – but it’s really called by the service-side stub • The stubs send messages to each other to make the RPC happen transparently (almost!) CS-4513, D-Term 2010 Remote Procedure Call 12 RPC Stubs Tanenbaum & Van Steen, Fig 4-7 CS-4513, D-Term 2010 Remote Procedure Call 13 RPC Stubs Not shown in this textbook figure are Steps 7, 8, and 9, where the server process returns a result to Tanenbaum & Van Steen, Fig 4-7 server stub, which packages it and sends it across network to client CS-4513, D-Term 2010 Remote Procedurestub. Call 14 RPC Stubs – Summary • Client-side stub • Server-side stub – Looks like local server – Looks like local client function to the client function to server – Listens on a socket for – Same interface as message from client local function stub – Bundles arguments – Un-bundles arguments into a message, sends to local variables to server-side stub – Makes a local function – Waits for reply, un- call to server bundles results – Bundles result into – returns reply message to client stub CS-4513, D-Term 2010 Remote Procedure Call 15 Result – a very useful Abstraction • The hard work of building messages, formatting, uniform representation, etc., is buried in the stubs • Where it can be automated! • Designers of client and service can concentrate on semantics of application • Programs behave in familiar way CS-4513, D-Term 2010 Remote Procedure Call 16 Questions? CS-4513, D-Term 2010 Remote Procedure Call 17 RPC – Issues • How to make the “remote” part of RPC invisible to the programmer? • What are semantics of parameter passing? – E.g., pass by reference? • How to bind (locate & connect) to servers? • How to handle heterogeneity? – OS, language, architecture, … • How to make it go fast? CS-4513, D-Term 2010 Remote Procedure Call 18 RPC Model • A server defines the service interface using an interface definition language (IDL) – the IDL specifies the names, parameters, and types for all client-callable server functions • A stub compiler reads the IDL declarations and produces two stub functions for each service function – Server-side and client-side CS-4513, D-Term 2010 Remote Procedure Call 19 RPC Model (continued) • Linking:– – Server programmer implements the service’s functions and links with the server-side stubs – Client programmer implements the client program and links it with client-side stubs • Operation:– – Stubs manage all of the details of remote communication between client and server CS-4513, D-Term 2010 Remote Procedure Call 20 RPC Stubs Tanenbaum & Van Steen, Fig 4-7 CS-4513, D-Term 2010 Remote Procedure Call 21 Marshalling and Unmarshalling Arguments • Marshalling – the packing of function parameters into a message • Unmarshalling – the extraction of parameters from a message • Function call:– – Client stub marshals the arguments into message – Server stub unmarshals the arguments and uses them to invoke the service function • Function return:– – Server stub marshals return values into message – Client stub unmarshals return values and returns them as results to client program CS-4513, D-Term 2010 Remote Procedure Call 22 Issue #1 — Data Representation • Big endian vs. little endian Sent by Pentium Rec’d by SPARC After inversion CS-4513, D-Term 2010 Remote Procedure Call 23 Data Representation (continued) • IDL must also define representation of data on network – Multi-byte integers – Strings, character codes – Floating point, complex, … – … • example: Sun’s XDR (eXternal Data Representation) • Each stub converts machine representation to/from network representation • Clients and servers must not try to cast data! CS-4513, D-Term 2010 Remote Procedure Call 24 Issue #2 — Pointers and References read(int fd, char* buf, int nbytes) • Pointers are only valid within one address space • Cannot be interpreted by another process • Even on same machine! • Pointers and references are ubiquitous in C and C++ • Even in Java implementations! CS-4513, D-Term 2010 Remote Procedure Call 25 Pointers and References — Restricted Semantics • Option: call by value – Sending stub dereferences pointer, copies result to message – Receiving stub conjures up a new pointer • Option: call by result – Sending stub provides buffer, called function puts data into it – Receiving stub copies data to caller’s buffer as specified by pointer CS-4513, D-Term 2010 Remote Procedure Call 26 Pointers and References — Restricted Semantics (continued) • Option: call by value-result – Caller’s stub copies data to message, then copies result back to client buffer – Server stub keeps data in own buffer, server updates it; server sends data back in reply • Not allowed:– – Call by reference – Aliased arguments CS-4513, D-Term 2010 Remote Procedure Call 27 Questions? CS-4513, D-Term 2010 Remote Procedure Call 28 Observation — Remote Procedure Call • It should not be possible for a client to tell whether the service is implemented by • The same process • A different process in the same machine • A process on a different machine • It should not be possible for a service to tell whether the client is implemented by • The same process • A different process on the same machine Except with respect• toA failuresprocess on a different machine Client and server processes can fail CS-4513, D-Term 2010 Remote Procedure Call 29 independently Three More RPC Topics • Transport • Connection vs. connectionless protocol • Asynchronous RPC • RPC Binding CS-4513, D-Term 2010 Remote Procedure Call 30 Transport of Remote Procedure Call • Option — TCP • Connection-based, reliable transmission • Useful but heavyweight, less efficient • Necessary if repeating a call produces different result • Alternative — UDP • Unreliable transmission • If message fails to arrive within